William L. Weaver, Ph.D. - Recently Updated Pages

Affiliations

williamlweaver — Mon, 02 Jun 2008 22:01:11 CDT

Faculty Member
La Salle University
Philadelphia, PA

Contributing Editor
Editorial Board Member
Scientific Computing
Rockaway, NJ

Homebound Tutor
The Renaissance Academy Edison Charter School
Phoenixville, PA

Member
Class of 2000
Project Kaleidoscope Faculty for the 21st Century
Washington, D.C.

Member
Sigma Pi Sigma
National Physics Honor Society
American Institute of Physics
College Park, MD

Member
Augustus Lutheran Church
Trappe, PA

Research Collaborations

Technical Consulting Services

ALZA Corporation

University of Florida
Aerospace Engineering, Mechanics & Engineering Science Department

Visteon Corporation

NoThankYou

williamlweaver — Mon, 02 Jun 2008 21:59:45 CDT

"If you want to succeed, double your failure rate."

- Thomas J. Watson, IBM

The quote above is attributed to one of the most successful American industrialists of the 20th century, Thomas J. Watson. He began his career in 1892 as a bookkeeper for his home town market at age 18 and later sold pianos and sewing machines. At age 21, he became a salesman for National Cash Register (NCR) and rose to the position of General Sales Manager.

In 1912, he and 30 other NCR executives were indicted for violating antitrust statutes, fired from NCR, fined $5,000 (over $100,000 in today’s dollars) and sentenced to one year in jail. His jail sentence was later overturned on appeal. Soon after, he joined the Computing Tabulating Recording Corporation (CTR) in 1914 as General Manager and after the appeal was promoted to the position of President in 1915.

In 1924 he renamed the company International Business Machines (IBM) and worked to transform the company into the household word it is today. In 1937 he became the president of the International Chamber of Commerce and passed away in 1956 at the age of 82. A career altering event, such as a job loss, hefty fine or jail time would certainly cause many folks to withdraw into obscurity. As you are faced with disappointments and rejections in your career, take some time to evaluate your goals and then formulate alternate paths to achieve them. You can’t win the lottery if you don’t by a ticket.

1987 - 1988 Graduate School Rejections

The George Washington University at NASA Langley Research Center
Joint Institute for the Advancement of Flight Sciences
Hampton, VA

The Pennsylvania State University
Department of Aerospace Engineering
State College, PA

The Pennsylvania State University
Department of Chemistry
State College, PA

1987 - 1988 Graduate School Acceptances

University of Pittsburgh
Department of Chemistry
Pittsburgh, PA

The Ohio State University*Department of Chemistry
Columbus, OH

*Entered Program 1988

1991 - 1993 Job Search Rejections after Ph.D. Degree

3M Corporation
St. Paul, MN

Abbott Laboratories
Abbott Park, IL

Advanced Micro Devices
Sunnyvale, CA

The Aerospace Corporation
Los Angeles, CA

Air Products and Chemicals, Inc.
Allentown, PA

Aldrich Chemical Company, Inc.
Milwaukee, WI

Allied-Signal Aerospace Company
Aerospace Technology Center
Columbia, MD

Allied-Signal Aerospace Company
AiResearch Los Angeles Division
Torrance, CA

Allied-Signal
Bendix Guidance Systems Division
Teterboro, NJ

Allied Signal, Inc.
Research and Technology
Des Plaines, IL

American Cyanamid Company
Wayne, NJ

Amoco Research Center
Naperville, IL

Analytic Services, Inc.
Arlington, VA

Analog Devices, Inc.
Norwood, MA

ANALOGIC Corporation
Wakefield, MA

Applied Research Laboratories
Austin, TX

Aristech Chemical Corporation
Pittsburgh, PA

Applied Research Laboratories
University of Texas at Austin
Austin, TX

ARC Professional Services Group
Rockville, MD

ARCO Oil and Gas Company
Dallas, TX

Argonne National Laboratory
Argonne, IL

ARINC Incorporated
Annapolis, MD

ARMCO, Inc.
Middletown, OH

AT&T
Morristown, NJ

AT&T Bell Laboratories
Holmdel, NJ

BASF Corporation
Parsippany, NJ

Battelle
Pacific Northwest Laboratories
Richland, WA

BDM International, Inc.
McLean, VA

Bendix Field Engineering Corporation
Columbia, MD

Bell Communications Research
Piscataway, NJ

Big Sky Laser Corporation
Bozeman, MT

The Boeing Company
Seattle, WA

Burle Industries, Inc.
Lancaster, PA

Burleigh Instruments, Inc.
Fishers, NY

Burroughs Wellcome Company
Triangle Park, NC

Center for Naval Analysis (CNA)
Alexandria, VA

Chevron Corporation
San Francisco, CA

The Clorox Company
Oakland, CA

Computer Sciences Corporation
Falls Church, VA

COMSAT Laboratories
Clarksburg, MD

Corning Incorporated
Corning, NY

Coulter Electronics, Inc.
Hialeah, FL

Data Translation
Marlboro, MA

Digital Equipment Corporation
Manard, MA

The Dow Chemical Company
Midland, MI

Dynetics, Inc.
Huntsville, AL

Eastman Kodak Company
Rochester, NY

E.I. duPont de Nemours & Co., Inc.
Wilmington, DE

Eli Lilly Company
Indianapolis, IN

E-Systems
Garland Division
Dallas, TX

General Atomics
San Diego, CA

GEO-CENTERS, Inc.
Fr. Washington, MD

W.L. Gore & Associates, Inc.
Newark, DE

Grumman Corporation
Bethpage, NY

GTE Laboratories
Waltham, MA

Harris Corporation
Melbourne, FL

Hercules Incorporated
Wilmington, DE

Hewlett Packard
Microwave Semiconductor Division
Optical Communication Division
Optoelectronics Division
Components Group
San Jose, CA

Honeywell, Inc.
Minneapolis, MN

HRB Systems
State College, PA

Hughes Aircraft Company
Malibu, CA

ICF International, Incorporated
Fairfax, VA

ICI Americas, Inc.
Wilmington, DE

Intel Corporation
Folsom, CA

Intergraph Corporation
Huntsville, AL

Intermetrics, Inc.
Warminster, PA

International Business Machines Corporation
White Plains, NY

ITT Research Institute
Annapolis, MD

Lawrence Livermore National Laboratory
Livermore, CA

Laser Diode, Inc.
Edison, NJ

Lincoln Laboratory
Lexington, MA

Lockheed Missiles & Space Company
Sunnyvale, CA

Lockheed Sanders, Inc.
Nashua, NH

Locus, Inc.
Sate College, PA

LTV Aerospace and Defense Company
Missles Division
Dallas, TX

Martin Marietta Corporation
Advanced Development & Technology Operations
Baltimore, MD

McDonnel Douglas
Electronics Systems Corporation
Hazelwood, MO

Microsoft Corporation
Redmond, WA

Motorola, Inc.
Schaumburg, IL

National Aeronautics and Space Administration
Goddard Space Flight Center
Greenbelt, MD

National Security Agency
Fort Meade, MD

Naval Air Development Center
Warminster, PA

Naval Surface Warfare Center
Dahlgren, VA

NEC Electronics, Inc.
Mountian View, CA

New England Telephone
A NYNEX Company
Boston, MA

Nicolet Instrument Company
Madison, WI

Northrop Corporation
Electronic Systems Division
Norwood, MA

Philips Laboratories
Briarcliff Manor, NY

Phillips Petroleum Company
Bartlesville, OK

Princeton Instruments, Inc.
Trenton, NJ

The Procter & Gamble Company
Miami Valley Laboratories
Cincinnati, OH

Radian Corporation
Austin, TX

The RAND Corporation
Santa Monica, CA

Raychem Corporation
Menlo Park , CA

Rockwell International Corporation
Seal Beach, CA

L.B. Russell Chemicals
Long Island City, NY

Sandia Naitonal Laboratories
Albuquerque, NM

Syracuse Research Corporation
Syracuse, NY

Siemens Medical Systems, Inc.
Iselin, NJ

Silicon Systems
Tustin, CA

SmithKline Beecham
Philadelphia, PA

SRA Corporation
North Arlington, VA

SUN Microsystems
Mountian View, CA

Systems Research Laboratories, Inc.
Dayton, OH

The Analytic Science Corporation (TASC)
Reading, MA

Teledyne Brown Engineering
Huntsville, AL

TEXACO Inc.
Houston, TX

Texas Instruments Incorporated
Dallas, TX

United Technologies Research Center
East Hartford, CT

University of Dayton Research Institute
Dayton, OH

Varian Associates
Palo Alto, CA

Vitro Corporation
Silver Spring, MD

The Williamsburg Group, Inc.
East Brunswick, NJ

Xerox Corporation
Rochester, NY

1991 - 1993 Job Offers after Ph.D. Degree

The Boeing Company
Seattle, WA

Description: Onsite contractor at NASA John F. Kennedy Space Flight Center, Cocoa Beach FL

Duties: Develop gas detection sensors for International Space Station assembly building

Systems Research Laboratories, Inc.*
Dayton, OH

Description: Onsite contractor at Wright-Patterson Air Force Base, Dayton, OH

Duties: Apply ultrafast laser spectroscopy to the development and deployment of advanced turbine-engine combustion and aviation-fuel diagnostics

*Accepted Position 1993

1998 - 1999 Academic Position Search Rejections

East Stroudsburg University
Chemistry Department
East Stroudsburg, PA

La Salle University
Department of Chemistry and Biochemistry
Philadelphia, PA

Lebanon Valley College
Department of Chemistry
Annville, PA

University of Scranton
Chemistry Department
Scranton, PA

1998 - 1999 Academic Position Offers

La Salle University*
Integrated Science, Business and Technology Program
Philadelphia, PA

*Accepted Position 1999

1999 - Present Publication & Presentation Rejections

Handal, J.; Weaver, W.L.; Frisch, R.; Williams, E.A.; DiMatteo, D.O. "Physical Analysis of Seven Drug-Bone Cement Composites," North American Spine Society, 22nd Annual Meeting, October 23-27, 2007 Austin, TX. (not accepted).

Weaver, W.L.; Timmerman, M.W.; Jones, N.L. "A Systematic Framework for Interdisciplinary Education," Journal of College Science Teaching, 2006. (not accepted).

Weaver, W.L."A Virtual Control Room for Production and Operations Management Education," National Instruments NIWeek 2005, August 16-18, 2005, Austin, TX. (not accepted).

Back to top

March 2006 Editorial

williamlweaver — Mon, 02 Jun 2008 21:49:04 CDT

Contributed editorial appearing in
Scientific Computing & Instrumentation 23:4, March 2006, pg. 14.

Going Deep
A new algorithm for robotic vision

Bill Weaver, Ph.D.

Cruising around in my Nissan 350Z with the ground-effect neon thumping to the soundtrack, or careening around curves at over 170 MPH in my V12 Aston Martin DB9, are high on the short list of things that assuage my daily commute amid snarled traffic in the real world. In a relatively short period of time, the hefty cathode ray tube in our family room has been transformed from the keyhole view of lack-luster situation comedies and reality shows into a rich interface between my family's senses and the minds of video game designers. The graphics displayed by our latest holiday acquisitions are visually stunning. The industry has moved beyond the stage of "almost-real" into "hyper-real" — a world wherein the most impossible camera angles are commonplace, the lighting is always perfect and the frame acquisition rate is just right. There is plenty of science buried in this experimental setup and researchers are developing the tools to uncover and leverage it.

One such scientific mystery is how the binocular vision we use to navigate through our three-dimensional (3-D) environment has no difficulty extracting volumetric information from a two-dimensional (2-D) pattern emitted from the glowing phosphors coated on a piece of glass. I see two identical flattened images when using both eyes to look at the screen, yet I can schuss around obstacles at high-scale velocity with ease. Even this simple test reveals that 3-D depth processing is more than optics; it must include some high-level image processing.

Professor Andrew Y. Ng and his research group at Stanford University are asking similar questions of robotic vision. Autonomous vehicles equipped with a phalanx of cameras, sensors, lasers and radar are making their way through cluttered environments; however, Professor Ng is investigating lightweight agile solutions formed around a solitary color video camera. It appears the key to extracting depth from a single, 2-D monocular image involves the same techniques artists use to inject depth into their pieces, namely, texture, perspective and focus. The Renaissance masters skillfully detailed the stitching and folds on the clothing of near subjects while purposefully reducing the scale, focus and detail of objects in the distant background to produce life-like vistas on flat canvas. Short of developing a thinking machine that recognizes objects and their common size in perspective, Ng's method extracts generic features from the digital image and transforms them into depth information.

The algorithm is based on a popular 2-D version of the Naïve Bayesian Classifier (NBC), known as a Markov Random Field (MRF) whose goal is to classify combinations of image pixel attributes into a range of depth values. Bayesian analysis permits an observed outcome to be related statistically to a collection of input observables. For example, atmospheric visibility can be related to input observables of temperature, humidity, time-of-day and atmospheric pressure. Even though an exact deterministic equation connecting the input to the observed outcome may not be known, the NBC can generate the probability of a specific outcome given the known input values. The statistics are generated by "training" the NBC with a set of input/output pairs and are validated by comparing the output to a test set of additional input/output pairs. If there is no true relationship, the NBC performs very poorly on the test set; however, high-quality results can be obtained if a relationship does exist, even if it is not explicitly known.

Ng's group collected image/depth pairs using a small 1704 x 2272 pixel color digital camera and a one-dimensional laser range finder mounted on a translation stage to find the true depth of the image at a resolution of 86 x 107. The MRF was trained using 75 percent of the pairs, and validated using the remaining 25 percent. The digital images were segmented into small pixel cells and correlated with filter patterns designed to classify texture variations, texture gradients, haze and edge orientation resulting in 34 unique local input observables for each cell. The cells are also compared to their nearest neighbors at multiple resolutions to extract global information of 19 additional features, resulting in a set of 646 input observables for each cell. The trained MRF was used to predict depth in test images of both indoor and outdoor locations and was determined to have an average error of 35 percent, meaning the image of an obstacle 10 meters away would appear between six and 14 meters away to the algorithm. At a 10-Hz frame rate, an autonomous robot would have adequate time to avoid the obstacle even with this uncertainty.

The one-camera system has dramatically reduced the amount of hardware required to provide depth information, and can also determine distances five to 10 times further away than the dynamic range of many triangulating two-camera systems. The algorithm has been used by a small radio-controlled car to navigate through a cluttered, wooded area autonomously for several minutes before crashing. Further enhancements may one day enable the development of autopilot systems for automobiles. But that would drastically reduce my enjoyment of video games.

William L. Weaver is an Associate Professor in the Department of Integrated Science, Business and Technology (ISBT) at La Salle University in Philadelphia, Pennsylvania, USA. As a Contributing Editor for Scientific Computing magazine, his latest column can be found at www.ScientificComputing.com.

2006 Columns

williamlweaver — Mon, 02 Jun 2008 21:48:56 CDT

I am a Contributing Editor and Editorial Board member of Scientific Computing, a trade magazine published by Advantage Business Media. I write a monthly editorial column and an occasional feature article on the broad topic of Data Acquisition and Data Analysis. A list of current and past columns and articles appears below. I am providing the text of older columns in the wiki with the hope they will generate discussion.

January 2006

It's About Time
A new data acquisition paradigm

February 2006

Focus on Knowledge
Knowledge-enhanced electronics logic programming

March 2006

Going Deep
A new algorithm for robotic vision

February 2006 Editorial

williamlweaver — Mon, 02 Jun 2008 21:41:05 CDT

Contributed editorial appearing in
Scientific Computing & Instrumentation 23:3, February 2006, pg. 12.

Focus on Knowledge
Knowledge-enhanced electronic logic programming

Bill Weaver, Ph.D.

Figure 1: KEEL dynamic graphic language used to adjust weights visually. Relative heights of the outputs labeled One, Action1, and Action2 indicate their importance to the overall decision (the weight). The blue fill of each output indicates the accumulated support for the output. The thin, black wires represent relationships between the outputs, and the sliders along the bottom of the window permit the user to simulate activating (green) and inhibiting (red) input into the IBIS decision tree. Courtesy Compsim LLC, used with permission.

Valley Forge, Pennsylvania earned its reputation for nasty winters during the Revolutionary War. By the time we roll into February around here, the phrase"warm and fuzzy" elicits thoughts of dry mittens, cozy fleece jogging suits, and slippers being heated by the fireplace. During the 2000 U.S. Presidential election, the term "fuzzy" was used as a pejorative to describe the "fuzzy math" used by one of the candidates to predict the effect of tax cuts on the economy — a complex system with many inputs, some of which are hidden or counter-intuitive. While "traditional math" is well-suited for simple or "tame" problems that have a "correct" answer, it is often inadequate for the modeling of "wicked" problems whose "best available" answers lie in a gray area between good and bad.

Wicked problems were studied by Professor Horst Rittel at the University of California (UC) — Berkeley in the 1970s, where he developed the Issue-Based Information System (IBIS) to augment the vocabulary of traditional math. IBIS decomposes complex problems into the branches of a decision tree having pro and con arguments as inputs for the coordination and planning of political decisions. When formulated, an IBIS tree looks similar to the layout of an artificial neural network (ANN); the leaves serve as inputs and the final decision as the output. The solution of both the IBIS tree and the ANN requires the inputs to be weighted as a function of their relative importance to each other and to the final decision. The weights connecting the neurons of an ANN are commonly adjusted by presenting the network with a set of matched inputs and outputs in conjunction with a supervised or unsupervised algorithm that trains it to "learn" the relationships between each neuron. The final values of the trained ANN weights are typically incidental as long as the network performs correctly. This training approach does not consider that the relative importance (weights) of issues is often known by the human decision makers. While it is difficult for a human expert to explain exactly why they "feel" one issue is of greater importance over another, they "know" it to be true given their education, experience and knowledge of the situation. A more efficient training algorithm would incorporate this expert heuristic knowledge of the weights.

One method used to connect a collection of known inputs to a known output using heuristically-known weights is the application of classical logic. The weights take the form of True (100%) or False (0%), and their combinations are processed using IF/THEN statements coupled with the logical operators of AND, OR, NOT and XOR. This results in the creation of a von Neumann-style expert system produced by traditional programming languages; however, problems arise when the differences between True and False are not clear or sharp. Professor Lofti Zadeh, also at UC Berkeley, introduced "fuzzy" logic (FL) in 1965, enabling the use of real-valued weights in IF/THEN statements. For instance, if an employee misses one day of work in a 30-day month, an FL system can use the weight of "1/30th-True" for the input named "MISSED WORK." This allows the system to work with inputs that blur the line between True and False, making them "fuzzy." While FL continues to be an important paradigm for processing real-valued inputs, the design of the FL algorithms can be tedious. A graph of geometric shapes having known angles and widths must be adjusted to tune each FL algorithm. Human experts often envision the weights of competing inputs to a decision as just that — weights on a balanced scale that tips toward the more important factor, not a complex geometric pattern.

With the desire to utilize the ability of FL to incorporate expert heuristic-knowledge of competing inputs into the creation of layered IBIS decision trees, Compsim has developed a dynamic graphical language called Knowledge Enhanced Electronic Logic (KEEL) that permits the developers to define the number and identity of inputs to a complex decision and, more importantly, adjust the normalized weight of each input by moving a virtual element between the values of 0-100 percent (as in Figure 1). Using KEEL technology, the expert human can supervise the education of the IBIS through direct manipulation of the weights, rather than handing the process over to the blind learning algorithm of an ANN. When a satisfactory IBIS functionality is obtained, all of the settings can be exported to an XML file to be used as documentation for an audit trail, the creation of reusable components, and for importing the values into applications written in traditional programming languages for incorporation into devices and controllers. KEEL technology enables the expert user to focus on the known importance of the input factors to a decision, and to teach an automated system to use the same reasoning. Knowing that an automated controller is using the identical reasoning as my way of thinking gives me a warm and fuzzy feeling.

William L. Weaver is an Associate Professor in the Department of Integrated Science, Business and Technology (ISBT) at La Salle University in Philadelphia, Pennsylvania, USA. As a Contributing Editor for Scientific Computing magazine, his latest column can be found at www.ScientificComputing.com.

2005 Columns

williamlweaver — Mon, 02 Jun 2008 21:27:42 CDT

January 2005

Thinking about Thinking
A new model for algorithmic intelligence

January 2005 Feature Article

Change for the Better
Leading edge technology to watch

February 2005

Turning up the Heat
Melanophila acuminate and biomimetic sensors

March 2005

Creeping Crawlers
Autonomous adaptive data mining agents

April 2005

Aim High
The HIAPER Project for environmental research

May 2005

One Giant Leap
Measuring the small with precision

June 2005

Can You Read My Mind?
Acquiring signals for neuroprosthetics

July 2005

When Push Comes to Shove
Quantum imaging with magnetic resonance force microscopy

August 2005

Look Who's Talking
Spectral learning algorithm for speech separation

September 2005

No SPAM for You!!!
Unified e-mail filtering technology

October 2005

Abuzz About ZigBee
Low-power self-organizing data networking

November 2005

Say Hello to ADIOS
Unsupervised learning of linguistic structures

Columns in Scientific Computing

williamlweaver — Mon, 02 Jun 2008 21:27:28 CDT

2008 Columns
2007 Columns
2006 Columns
2005 Columns
2004 Columns
2003 Columns
2002 Columns
2001 Columns
2000 Columns

January 2006 Editorial

williamlweaver — Mon, 02 Jun 2008 21:26:32 CDT

Contributed editorial appearing in
Scientific Computing & Instrumentation 23:2, January 2006, pg. 14.

It's About Time
A new data acquisition paradigm

Bill Weaver, Ph.D.

"Happy Nu (ν) Year!" Yeah, that's an oldie and not very goodie. I guess the years spent in a darkened laser lab often increase the perceived humor of low wit. Upon reaching the end of our unit on optics each year, I try out the hilarity of "What's new?" ― Answer: "c/λ" on my class and I get the same crickets I am hearing now. All groaning aside, the relationship between frequency, nu (ν) , wavelength, lambda (λ), and the propagation speed of a radiating energy wave, (c) is the foundational model for a large number of processes.

Other models of force, absorption, voltage and resistance permit the design of robust transducers and sensors for the measurement of many physical quantities. Unfortunately, as these quantities are affected by multiple processes simultaneously, data acquisition is often limited by selectivity. If many people are conversing loudly in a conference hall, it is difficult to listen to a single speaker from across the room. Sensitivity and detection limit are of concern also, as it doesn't matter that everyone else is silent if our ears are not able to detect a soft-spoken whisper. Measuring the presence and identity of a chemical compound can be achieved by any number of laboratory instruments, but our current technology often requires the use of separation science, such as chromatography or extraction, so that other signals do not overwhelm our target measurement.

Another method used for signal isolation is to transform it into a less populated, and thereby, a more quiet, dimension. These are not fictional screenwriter dimensions, but physical engineering dimensions such as energy, space and time. In the case of optical absorption spectroscopy, the absorption of a single component contained in a sample of many colored compounds can be isolated by the careful selection of the frequency or wavelength of light used for the analysis. With the aid of an optical chopper, the frequency of the absorption measurement can be moved to a quieter region of the noise spectrum. Cavity ring-down (CRD) absorption spectroscopy transforms intensity-modulation into a measurement of signal decay time. Rather than dealing with the noise inherent in a system used to measure a miniscule decrease in a bright light due to sample absorption, CRD uses an intensity detector that measures the amount of light that leaks out of a resonant optical cavity as a function of time. The natural decay time of the cavity depends on the efficiency of the reflective end mirrors, and is in a separate measurement dimension than the intensity noise of the light source. When an absorption sample is placed into the CRD cavity, the observed decay time is decreased due to the additional loss of energy to the sample.

Arik Ariav, a mechanical engineer and founder of Nexense, has made a large splash in the data acquisition community with the introduction of his recently patented measurement methodology known as frequency change by wavelength control (FCWC). Analogous to the CRD technique, the FCWC uses a resonance cavity embedded into a digital circuit. Instead of optical mirrors, the cavity is formed between a transmitter and receiver, and the transit time of a propagating energy wave is measured continuously. Rather than attempting to measure the transit time directly at a resolution of a few picoseconds, the cavity serves as a delay line in the feedback loop of an operational amplifier. Similar to an optical laser cavity, a resonant wave is established within the electronic circuit as the energy leaves the transmitter, transverses the cavity medium, enters the receiver and is subsequently amplified. The frequency and phase of the resulting standing wave depend on its propagation time through the medium, thereby converting transit time and distance into frequency. The circuit frequency is compared to that of an internal reference clock and the differences are integrated over a specified time to increase the signal-to-noise ratio (SNR).

If a sample is placed in the cavity (or if the transmitter and receiver are placed on the sample to form a cavity), any change in cavity length or speed of energy wave propagation will be detected by the ensuing frequency change of the circuit. Nexense reports that physical changes to the sample induced by (in alphabetical order) acceleration, angular velocity, density, distance, displacement, elasticity, flow, mass, pressure, shape, strain, temperature, tilt, torque, vibration, viscosity, volume and weight can be accurately measured by FCWC with a SNR approaching 190 dB. This is quite a hefty claim in an industry that utilizes specialized sensors having unique response and noise characteristics. Also, Nexense does not specifically address problems of selectivity that plague other highly sensitive techniques. Regardless of the selectivity issue, recent Nexense patent applications describe sensors for measuring the weight of an automobile passenger to determine airbag deployment and an apparatus for non-invasively monitoring blood glucose concentration in vivo by using FCWC to monitor the photoacoustic effect.

This new technology sounds overly optimistic of its ease of application; however, with recently announced partnerships with Timex, Fujitsu and others, only time will tell.

William L. Weaver is an Associate Professor in the Department of Integrated Science, Business and Technology (ISBT) at La Salle University in Philadelphia, Pennsylvania, USA. As a Contributing Editor for Scientific Computing magazine, his latest column can be found at www.ScientificComputing.com.

November 2005 Editorial

williamlweaver — Mon, 02 Jun 2008 21:15:54 CDT

Contributed editorial appearing in
Scientific Computing & Instrumentation 22:12, November 2005, pg. 12.

Say Hello to ADIOS
Unsupervised learning of linguistic structures

Bill Weaver, Ph.D.

"I brang it to school yesterday" is the most recent conversation starter with our fourth-grader. Like many parents, I would like to have the phrase "Use brought; 'brang' isn't a word" emblazoned on a shirt so I can simply point to it several times a day. My correction is always followed by the retort "But it's 'ring-rang-rung' and bring sounds like ring." Explaining 'there are just about as many exceptions as there are rules' can make teaching a language almost as difficult as learning it.

Programming computers is often equated with teaching children. Once a specific rule or task is learned or programmed, it is performed exactly as it was taught, but frequently applied as the solution to the wrong problem. With more education and experience, children learn when to use the appropriate solution by following the examples of their teachers or peers. As they continue to mature and develop, they also synthesize novel solutions using knowledge of the foundational rules. Traditional computer applications typically do not evolve with experience or learn new rules; however, that is beginning to change.

Researchers at Cornell and Tel Aviv Universities have been tackling the task of infusing the rules of grammar into a computer. Rather than creating a massive database of rules and exceptions for the computer to follow, Cornell professor Shimon Edelman and his colleagues have developed an unsupervised algorithm that scans text containing properly-phrased sentences and infers the underlying rules of grammar. The algorithm, named Automatic Distillation of Structure (ADIOS) finds complex patterns and phrases by repeatedly aligning sentences and discovering portions that overlap.

Professor Edelman describes the following example: The sentences, "I would like to book a first-class flight to Chicago," "I want to book a first-class flight to Boston" and "Book a first-class flight for me, please" may give rise to the pattern "book a first-class flight" — if this candidate pattern passes the novel statistical significance test that is the core of the algorithm. If the system also finds the sentences, "I need to book a direct flight from New York to Tel Aviv" and "I would like to book an economy flight," it may infer that the phrases "first-class," "direct" and "economy" are equivalent in the context of the new pattern.

ADIOS is based on two components: (1) a Representational Data Structure (RDS) graph, and (2) a Pattern Acquisition (PA) algorithm that learns the RDS graph by detecting patterns in the text in an unsupervised manner that does not require the classification of phrases into a predetermined taxonomy. In the initial phase, the text is segmented into the smallest possible components (modifiers such as -ed for past tense and -ing for present tense are removed through a statistical analysis of the text to discover root words that appear in multiple tenses). Since these modifiers are not pre-programmed, the algorithm works on any written language, as diverse as English and Chinese, individual characters, and phonemes if applied to speech. The initial set of unique components forms the vertices of the preliminary RDS graph. When adjacent components are found in the text, a directed edge is inserted between the vertices to indicate the transition from one word to the next.

In the second phase, the PA algorithm recursively scans the body of text for "significant patterns" that consist of a sequence of graph edges (words that follow each other to form common phrases, such as "book a" in the example), an equivalence class of vertices (first-class, direct and economy), and a terminating sequence of graph edges (flight). At each level of recursion, the PA updates its collection of equivalence classes to form more and more complex patterns that ultimately represent a hierarchy for the text. The discovered relationships between the distilled patterns can be viewed as a classification tree. For English text, the leaves of the tree represent parts of speech such as nouns, verbs, adjectives and adverbs, while the order of the branches indicate patterns of their use, as in Subject - Conjunction - Subject - Verb - Article - Adjective - Predicate. The algorithm can ultimately synthesize new, grammatically-correct sentences by selecting leaves in branch order to generate a sentence such as "George and Pam have a fast car," following the example pattern described.

ADIOS was applied to a collection of phrases selected from the Child Language Exchange System (CHILDES) consisting of 9665 transcribed sentences (containing 74,500 words) produced by parents addressing their pre-school children. The algorithm discovered 317 patterns and 404 equivalence classes and is currently being used to investigate how toddlers learn their native languages. ADIOS has been tested on the full text of the Bible in several languages and on artificial context-free languages with thousands of rules. In addition to text analysis, it is being extended to other pattern-rich data including musical notation, computer vision and biological data. When used in protein analysis, ADIOS distilled common amino acid sequence patterns that were correlated with the functional properties of the proteins. Perhaps the ADIOS algorithm will help to uncover all kinds of human-made and natural patterns; but for now, "brang" isn't a word because Dad says so.

William L. Weaver is an Associate Professor in the Department of Integrated Science, Business and Technology (ISBT) at La Salle University in Philadelphia, Pennsylvania, USA. As a Contributing Editor for Scientific Computing magazine, his latest column can be found at www.ScientificComputing.com.

October 2005 Editorial

williamlweaver — Mon, 02 Jun 2008 21:06:01 CDT

Contributed editorial appearing in
Scientific Computing & Instrumentation 22:11, October 2005, pg. 14.

Abuzz About ZigBee
Low-power self-organizing data networking

Bill Weaver, Ph.D.

I recently purchased a 2006 Toyota Corolla with the goal of reducing the frequency of trips to my local filling station. In time, I may begin to miss my close friends behind the counter, but one thing I was startled to miss right away was the throttle cable to the Corolla's VVT-i engine. Similar to the "fly-by-wire" transformation experienced by the U.S. Air Force, "drive-by-wire" technology has trickled down to entry-level automobiles. Rather than adjusting the fuel flow mechanically, a sensor in the gas pedal communicates with the electronic throttle control (ETC) unit, which subsequently follows its own internal logic to meet the new set point. This can be disquieting to folks with control issues, but the pedal does not "tell" the ETC to accelerate the car, it simply "asks" for the new speed and the ETC responds. The concept of "control" is evolving into the concept of "re-goal. "

In order to control a device completely, it must be given its every move. If there is a lag between the device and the controller, it either becomes dormant or, worse, it fails to react to its dynamic environment and develops into a hazard to itself or others. My son and I have lost multiple remote-control cars to high-speed collisions with the curb when the cars traveled beyond the range of their transmitters. One solution is to use a wired communication system, but the tether severely limits the range and location of the device. Increasing the range of the transmitters and their communication rate so there is enough bandwidth to accommodate error correction can help as well. This is easily stated, but it requires a large amount of power. Wireless WiFi laptops, cell-phones and Bluetooth-enabled handhelds transmit and display voice and data at high rates, but their batteries must be recharged daily.

Enter ZigBee technology. Based on the IEEE 802.15.4 standard, ZigBee is a wireless communication specification designed with smart devices in mind that do not have to be controlled continuously, just re-goaled occasionally. This is tailor-made for industrial sensors and control units that already know how to perform their function and need to communicate small packets of measurement values when they are available, and then accept new set points. Since there is only occasional traffic on a ZigBee network, it is based on a Carrier Sense Multiple Access with Collision Avoidance (CSMA/CA) protocol such that the sensor waits for the network to be available before it transmits its packet, thereby reducing the amount of bandwidth and power required to coordinate handshaking. Because a ZigBee sensor can spend most of is life in deep sleep, as in the case of a proximity sensor, it can operate on one set of commercial AAA batteries for months or years depending on its task.

A ZigBee Network Coordinator (ZNC) broadcasts a beacon signal to all of the devices in the network, such that each device knows when the time is proper to transmit or receive. When a new ZigBee node enters the network, the ZNC assigns it one of more than 18 x 1018 unique addresses (IEEE 64-bit addressing). Even though each device has a typical range of 50 meters, the network can grow to any size provided each ZigBee device is within range of another, as the specification supports the relaying of messages along the nodes of the network back to the ZNC.

Patrick Kinney, Chair of the IEEE 802.15.4 Task Group, presents a compelling comparison of ZigBee with current 802.11 WiFi technology. Consider a near-future home containing 100 smart sensors and controllers providing security, energy optimization and convenience. As each WiFi transmitter requires 667 mW of power, a city of 50,000 homes would require 3.33 megawatts of power for the sensors. Equivalent ZigBee devices only require 30 mW to remain active, dropping the requirement to 150 kilowatts. Add in the ability of each ZigBee device to sleep until needed at a typical duty cycle of 0.1 percent, and all five million sensors in the entire city can be powered at a total cost of 150 watts.

ZigBee devices require one percent of the power required to operate Bluetooth (IEEE 802.15.1) devices and have an advantage in latency. It typically takes 20 seconds for a Bluetooth device to join the network, while typical join times are 30 ms for ZigBee. A sleeping ZigBee sensor can wake up and communicate its data within 15 ms; 200 times faster than the three seconds typically required by Bluetooth. As both specifications are subsets of the same working group, they are not direct competitors. Rather, they are addressing different demands for wireless personal area networks (WPANs). While Bluetooth continues to connect cell phones, handhelds, computer peripherals and other remote controlled devices, ZigBee is an excellent choice for sensors and controls that go about their duties autonomously and only occasionally need to re-goal.

William L. Weaver is an Associate Professor in the Department of Integrated Science, Business and Technology (ISBT) at La Salle University in Philadelphia, Pennsylvania, USA. As a Contributing Editor for Scientific Computing magazine, his latest column can be found at www.ScientificComputing.com.

September 2005 Editorial

williamlweaver — Fri, 30 May 2008 23:02:11 CDT

Contributed editorial appearing in
Scientific Computing & Instrumentation 22:10, September 2005, pg. 14.

No Spam for You!!!
Unified e-mail filtering technology

Bill Weaver, Ph.D.

My e-mail to our magazine staff containing installments of this contributed column stopped being delivered a while back. There was no returned mail or out-of-the-office notification, just — well, just nothing. It took some time to realize our line of communication had been broken, and we had to regress to the good old handshaking technique of phone calls to make sure they got through. After a bit of snooping, it was discovered that I, or rather my exuberance, was to blame. Considering e-mail as a less formal mode of communication than post mail, I fell into the habit of signing off with the phrase, "Thanks!!!" — three exclamation marks and all. Unfortunately, to an unwanted e-mail or "spam" filter, that closing closely resembles "Viagra!!!" or "Get Rich!!!" and thus my e-mail was quickly shuttled off to the nether regions of ether space.

Filtering unwanted signals or noise has long been an important topic in data acquisition and instrument design. Noise can mask the signal, making it difficult to find, and noise filtering can attenuate or remove the signal altogether. Commercial e-mail providers and software companies have developed or adopted differing methods of spam filtering and they each have their strengths and weaknesses. Dr. William S. Yerazunis, Senior Research Scientist with Mitsubishi Electronic Research Laboratory (MERL) in Cambridge, MA, recently addressed the different filtering methods at the 2005 MIT Spam Conference. Along with his colleagues at University of California Riverside, Freie Universitat Berlin, and Embratel, Brazil, he classifies the current methods into three primary types and proposes that they are simply special cases of a common, unified approach to e-mail filtering.

One e-mail filtering method is simply to block all e-mail from an address contained on a blacklist. After a server has been determined as a source of spam, its address is added to a local blacklist or to a larger list maintained by a third party. This filter method is 100 percent effective against spam from the site; however, in the logical endgame all servers are blocked and no mail is delivered. A second filtering method is heuristic filtering, wherein a human examines spam and non-spam e-mail and determines "likely features" that are used to trigger or mark a message as spam; much like my "If text contains (!!!), then mark as spam." A third method is statistical filtering. Similar to heuristic filtering, a human classifies a group of messages as spam or not spam, but the rules of thumb are generated by an optimization algorithm based on a statistical analysis of the training set, such as Bayesian classification.

Dr. Yerazunis suggests these filtering methods follow a proposed six-step filtering pipeline and simply employ different versions of the common components:

Initial Transformation: This first step may include forcing exotic characters into a basic character set, unpacking MIME encodings into a common representation, and HTML de-obfuscation by removing nonsense tags that are invisible to the human reader, but can be inserted to break up "spammish" key words.

Tokenization: A regular expression (regex) is used to segment the message into text strings that are converted into unique values using a look-up method.

Feature Extraction: The tokens are grouped into meaningful finite sequences (tuples) based on the words they contain or the order in which the words appear in the message.

Feature Weighting: This step is based on the prior training of the filter to rank the importance of the tuple found in the message. The weight can be determined by how often the tuple has been found in spam messages, how closely the tuple resembles a known spam feature, and the size of the training set.

Weight Combination: The weights of the found features are then combined to determine the overall likelihood of the message being spam. This can be a simple linear addition of weight values, or a sophisticated nonlinear method such as one that considers the relative strengths of the sorted weights, a Bayesian combiner that considers the probability of the message being spam before and after a weight is considered, or a chi-squared method that compares the observed number of spammish tuples with an expected or acceptable number of rogue tuples.

Final Thresholding: After the weights are combined, a final "spam/not spam" decision is made based on the final value. For the statistical methods, the final threshold is often 0.5 (50 percent); however, the actual value can be adjusted by the filter designer to tune it for optimal results.

In addition to unifying the description of many current spam filters, the proposed filtering pipeline closely resembles the design of a McCulloch Pitts neural network that connects multiple inputs (the tuple values) through weighted connections into layers of artificial neurons whose values are thresholded to compute "yes/no." Perhaps recent developments into artificial neural networks can add a bit more intelligence into the spam filtering process and permit me to keep my exuberance!!!

William L. Weaver is an Associate Professor in the Department of Integrated Science, Business and Technology (ISBT) at La Salle University in Philadelphia, Pennsylvania, USA. As a Contributing Editor for Scientific Computing magazine, his latest column can be found at www.ScientificComputing.com.

Vitae

williamlweaver — Thu, 10 Jan 2008 21:52:57 CST

Publications

Education

Affiliations

Employment History

A Swing and a Miss

Home

williamlweaver — Thu, 10 Jan 2008 21:52:40 CST

This is a wiki for posting my comments and thoughts on the professional topics I cover as an Associate Professor in the Department of Integrated Science, Business and Technology (ISBT) at La Salle University in Philadelphia, Pennsylvania, USA.

I contribute to public wikis such as Wikipedia and Citizendium; however, these collaborative wikis are consensus building by nature. Everyone needs a place to put their thoughts out there, no matter how far out. I would like to use the "quick" nature of the "wiki" to rapidly post my own thoughts and musings in the hope of generating conversation. I guess most folks would prefer to use a blog, but I'm hooked on the hyper-linked format of the wiki.

This is also a wiki for posting your comments and thoughts about my comments and thoughts. Please join this wiki and comment away. I'm always up for a good debate, welcome derision and will even tolerate the occasional bit of praise if you are so moved.

August 2005 Editorial

williamlweaver — Thu, 10 Jan 2008 21:50:57 CST

Contributed editorial appearing in
Scientific Computing & Instrumentation 22:9, August 2005, pg. 8.

Look Who's Talking
Spectral learning algorithm for speech separation

By Bill Weaver, Ph.D.

I do not know the first thing about teaching science to the blind. Even my earliest course in problem solving begins with Step 1: Read the Problem; Step 2: Draw a Picture of the Situation — both initial steps inextricably associated with sight. I've exhausted cases of dry-erase markers and meters of chalk creating elaborate diagrams on the board and lean heavily on classroom projectors for animations and presentations. Entire courses in biology require students to memorize images seen through microscopes and sketch intricate cartoons of membrane transport mechanisms. It's no wonder that computer scientists and electronics engineers are fervently developing algorithms and hardware for machine vision. Even if the ultimate goal of intelligent, sighted machines may be far off, tools capable of sprinting through hours of security video, cabinets of medical images or mountains of scanned documents can make an intelligent researcher very efficient.

One important area of research into machine vision is image segmentation, wherein image pixels are divided into groups or regions that represent objects or parts of objects. This is a bit different than finding a specific object in an image by comparing a known model of the object with the image through correlation or other similar operation. In segmentation, image characteristics including texture, brightness and color are used to associate pixels into statistically similar clusters while additional characteristics such as contrast, intensity gradients and edge detection help to discern the interface between different objects. In this fashion, the segmentation algorithm is "blind" to the identity of the individual objects and simply points out their position, shape and number.

Professor Michael I. Jordan and his research group at the University of California, Berkeley, have recently developed a blind algorithm for finding the optimum values of a "similarity matrix," used to partition image pixels or points into disjoint clusters with points in the same cluster having high similarity, and points in different clusters having low similarity, using a process termed "spectral clustering." Instead of starting with the entire image and partitioning the points into exactly two groups of belonging or not belonging to the current cluster (as in the creation of a binary search tree), the algorithm considers the clustering of the image into any number k-subsets simultaneously. The optimal number of clusters, k, and to which cluster each point belongs, is treated as an error minimization problem through comparison of the algorithmic results with the "correct" answer - a segmented image created by a human operator. Like other supervised learning algorithms, once it is trained to produce the correct results it can be applied to new, but similar input.

The learning spectral clustering algorithm was presented at the 2004 Neural Information Processing Systems (NIPS) conference this past December where Jordan's group also described its utility to separating speech from multiple speakers that has been recorded by a single microphone. When an image of multiple objects is recorded by a single digital camera, it is actually being captured by millions of individual pixels and displayed as a rectangular matrix of intensity and color values. A single microphone records pressure values as a function of time, which can be plotted as a rectangular matrix having time along the horizontal dimension and pressure intensity along the vertical. As sound is most often associated with frequency analysis, a windowed, short-time Fourier transform can be applied to the pressure intensity values to obtain a rectangular matrix of frequency versus time, known as a spectrogram. Whereas multiple simultaneous objects are separated by position in an image, simultaneous speech from multiple speakers differs by frequency (pitch) and resonant features known as timbre.

Traditional voice recognition algorithms must know the identity of an individual speaker before their speech can be separated from other speakers or background noise, much like the known models used in object recognition. Since the learning spectral algorithm is only concerned with clustering similar items, it can separate the speakers blindly, without knowledge of their identity. In addition to clustering similar harmonic features of pitch and timbre, the algorithm tracks non-harmonic cues such as continuity, where two time-frequency points are close in time or frequency, and "common fate cues" — elements that exhibit similar time variation such as identical start and stop time and frequency co-modulation resulting from the "speech psychophysics" of punctuation, inflection and emphasis. The sound bites used for algorithm training are easily created by merging the speech files of individual speakers and the original separate files are used to determine the optimal results. With no knowledge of the language spoken, the identity of the speakers, or the mechanics used to create the sounds, the algorithm can be used as a tunable filter to extract a desired conversion. While not required by the algorithm, its utility to image and speech analysis and informatics in general reflects the far-reaching vision of its developers.

William L. Weaver is an Associate Professor in the Department of Integrated Science, Business and Technology (ISBT) at La Salle University in Philadelphia, Pennsylvania, USA. As a Contributing Editor for Scientific Computing magazine, his latest column can be found at www.ScientificComputing.com.

July 2005 Editorial

williamlweaver — Thu, 10 Jan 2008 21:50:44 CST

Contributed editorial appearing in
Scientific Computing & Instrumentation 22:8, July 2005, pg. 8.

When Push Comes to Shove
Quantum imaging with magnetic resonance force microscopy

By Bill Weaver, Ph.D.

The magnetic resonance force microscope (MRFM) uses an ultrathin silicon cantilever (yellow) with a nanometer size magnetic tip (blue) to detect the magnetic signal from an individual electron buried below the surface of the sample. Because the electron has a quantum mechanical property called 'spin,' it acts like a tiny bar magnet and can either attract or repel the magnetic tip. The interaction between the spin and the tip is localized to the bowl-shaped region in the sample called the 'resonant slice,' which moves as the cantilever vibrates. With the aid of a high-frequency magnetic field generated by a coil (right, background), the orientation of the electron (green arrow) flips as the resonant slice passes through. The magnetic force between the electron and magnetic tip alternates between attraction and repulsion every time the electron flips its orientation, causing the cantilever frequency to change slightly. A laser beam (left) is used to measure precisely the variations in cantilever vibration frequency. Image reproduced by permission of IBM Research, Almaden Research Center. Unauthorized use not permitted.

The amount of force required to press the keys of a computer keyboard is around 1 newton (N). That's about the force provided by a pad of Post-it notes as it mistakenly rests on your spacebar, filling the buffer until your computer emits a barrage of clicks for help. On the high side of things, the world-record holder GE90-115B turbofan jet engine provides over 560,000N of thrust (equivalent to a 64-ton anvil resting on your spacebar). Since they are extremely large on a human scale, the effects of massive forces on hefty objects can be easily observed. It's a bit more difficult to visualize tiny forces in our mind's eye since we don't often see their effects.

Let's consider the weight of an individual Post-it note from our pad and we have 0.01N. Take some really sharp scissors and cut it into a million equal pieces.We are down to 10-8N or 10 nanonewtons (nN). In the mid-1980s, researchers at Stanford University and IBM developed the atomic force microscope (AFM), an instrument capable of measuring this level of force between the sharp tip of an AFM probe and a surface. This non-destructive nudge permits the AFM to sense the distance between the tip and surface-layer atoms while it is scanned across the sample. Modern AFMs capable of creating 3-D images of surfaces having features 0.1nm – 8µm in height determine the force optically by measuring the deflection of a laser beam that is bounced off a short cantilever connecting the tip to the probe body. To achieve high sensitivity, the cantilever is vibrated at its resonance frequency and the surface force affects this frequency resulting in a detectable modulation of the reflected optical beam.

Although the atomic-scale AFM images are breathtaking, it is a surface technique while the Grail of nanotechnology is the ability to image an entire molecule in 3-D. Not wishing to disturb a sample by poking it with a sharp tip, Professors John Sidles and Joseph Garbini at the University of Washington, in collaboration with researchers from Cornell University, the University of Michigan, Army Research Lab and IBM are extending the capabilities of Magnetic Resonance Imaging (MRI) down to the atomic level. Their technique, known as Magnetic Resonance Force Microscopy (MRFM), combines the non-invasive 3-D nature of MRI with the atomic-scale scanning ability of the AFM. In MRFM, the AFM tip is replaced by a tiny permanent magnet and the cantilever is oscillated over the sample in a perpendicular configuration, much like the pendulum of a grandfather clock. A strong static magnetic field causes atomic nuclei to spin parallel or anti-parallel to its field lines as in the MRI technique. However, a second weaker oscillating magnetic field is used to flip the nuclear spins between parallel and antiparallel orientations instead of the radio-frequency light used by MRI. As the cantilever swings above the sample at its resonance frequency, the magnetic field of its tip experiences a force on the order of 10-18N or 1 attonewton (aN) from the flipping nuclei. To visualize this force, we need to slice one of our petite nN Post-its into 10 billion smaller pieces.

In order to maximize the effect of the aN atomic spin force on the delicate cantilever and differentiate it from other molecular and electrostatic forces, the frequency of the oscillating magnetic field is adjusted to match the resonance frequency of the tip, providing a small push on the tip while it is directly overhead, much like a child pushing their friend on a swing set from the middle as they swing by. After this frequency is found, the oscillating magnetic field is turned off, causing the cantilever to miss its push, and turned back on when the cantilever is at the extreme point of its swing, causing the atomic spin force to push in the opposite direction on the next pass. If you have ever experienced the misfortune of pushing against a friend on a swing at the midpoint of their cycle, one or both of you have had the chance to consider your actions as you lay sprawled on the ground.

This detection technique, known as interrupted oscillating cantilever-driven adiabatic reversals (iOSCAR), has been used successfully by the IBM group led by Dr. Dan Rugar to detect the spin of a single unpaired electron in silicon dioxide. This initial success was obtained by averaging the iOSCAR signal over a 13-hour period to overcome the effects of other forces in the sample. Sidles' group believes it is possible to measure individual nuclear spins in as little as 10 seconds with appropriate advances in the technique over the next five years. If they are successful, MRFM will easily push aside other contenders for the title of the world's most useful microscope.

William L. Weaver is an Associate Professor in the Department of Integrated Science, Business and Technology (ISBT) at La Salle University in Philadelphia, Pennsylvania, USA. As a Contributing Editor for Scientific Computing magazine, his latest column can be found at www.ScientificComputing.com.

June 2005 Editorial

williamlweaver — Tue, 08 Jan 2008 08:42:42 CST

Contributed editorial appearing in
Scientific Computing & Instrumentation 22:7, June 2005, pg. 10.

Can You Read My Mind?
Acquiring signals for neuroprosthetics

By Bill Weaver, Ph.D.

Now that summer is in full swing, the onslaught of blockbuster sci-fi movies and books has arrived. Much to the chagrin of pedigreed critics, alien technology coupled with special effects is a common formula for popular acclaim. Daring explorers discover a hidden trove of alien artifacts and, recognizing the portent to improved quality-of-life, the good guys race to understand its function while the bad guys battle to harness it for power and dominion over others. As the aliens are often from a similar universe, common elements such as electrical potential, current, resistance and frequency are easily recognized by our scientists. But, without a fundamental understanding of their linguistics, we have no idea how the logic of the circuitry converts input into output or what the output is meant to control. We can dissect the whole into its parts, but what is needed is a working system that permits output to be observed in response to specific input — a systematic, rather than an analytical approach.

This formula for good science fiction also works for great science fact. We have an alien artifact right between our ears — the mammalian brain. There are entire disciplines dedicated to the study of how humans develop, learn, interact and degenerate. However, only recently do we have the tools and techniques required to investigate the physical operation of the brain. One such research group, lead by Professor John Donoghue at Brown University in Rhode Island, is investigating the interpretation of neural activity in the relatively new field of neurotechnology. Previous studies have discovered specific regions of the brain that are responsible for voluntary behavior, including the primary motor cortex (MI) that generates signals for the brainstem and spinal systems that deliver commands to the arms, legs and face. If these signals can be measured and deciphered using modern technology, nerve networks damaged by injury or disease may be replaced by an electronic counterpart, known as a Neuro Motor Prostheses (NMP) that would re-route the messages to the muscles or replacement limbs.

The first task is measuring output from the MI neurons. When a single neuron fires, or "spikes," it produces an electrical potential as a result of the firing of other upstream neurons in its neural network. Much like frequency modulation (FM) communication, the number of spikes as a function of time in multiple neurons of the MI encodes information for direction, speed, position and forces on the limbs. Non-invasive sensors have been developed to measure neuron potential. However, the spikes become spatially and temporally averaged as they cross the skull and do not provide the high-frequency information from individual neurons necessary to interpret voluntary intent. Several groups are developing invasive micro array sensors made of biocompatible and biostable material that extend 1 to 2 mm into the cortical tissue and are able to detect spikes at 10s of kHz. These sensors have been embedded into the MIs of monkeys and their output recorded with simultaneous measurements of actual limb movement.

Even after the signals are acquired, the task of interpreting the information is formidable. Donoghue's group has developed a Switching Kalman Filter (SKF) to analyze the sequences of noisy spikes detected by the micro array sensors. A traditional discrete Kalman Filter (DKF) can be used to minimize the mean-square error between noisy measured data and a model parameter. Whereas linear regression can retrieve best-fit model values from a group of previously collected measurements, the DKF does so in real time by adjusting the model value in response to differences in predicted and measured subsequent values. Unfortunately, the DKF assumes the true signal is a single, steady-state value that varies slowly in time. NMP sensors have shown neurons communicate using a rapid sequence of pulse frequencies to coordinate the position, velocity and acceleration of the limb throughout the full range of motion. The SKF utilizes a Markov chain analysis to discover the hidden patterns of pulse sequences sent by the MI to the muscles. This is akin to using this hybrid Hidden Markov Model (HMM) - SKF to interpret the sequences of alphanumeric characters sent by an alien GPIB or RS-232 communication system. Once the HMMs are discovered, the NMP signals can be interpreted and used to stimulate muscle tissue, control human-computer interfaces (HMIs) or actuate robotic limbs.

The Christopher Reeve Paralysis Foundation has increased awareness of spinal cord injury and the debate over stem cell research. John Donoghue and his colleagues have recently commercialized their NMP research as Cyberkinetics Neurotechnology Systems and are transitioning their technology to human patients in clinical trials to investigate treatment of spinal cord injury, brain stem trauma, amyotrophic lateral sclerosis, muscular dystrophy and stroke. Perhaps solutions using our own developing technology are closer than discovering aliens?

William L. Weaver is an Associate Professor in the Department of Integrated Science, Business and Technology (ISBT) at La Salle University in Philadelphia, Pennsylvania, USA. As a Contributing Editor for Scientific Computing magazine, his latest column can be found at www.ScientificComputing.com.

May 2005 Editorial

williamlweaver — Tue, 08 Jan 2008 07:51:14 CST

Contributed editorial appearing in
Scientific Computing & Instrumentation 22:6, May 2005, pg. 12.

One Giant Leap
Measuring the small with precision

By Bill Weaver, Ph.D.

The great philosopher, Glinda, the Good Witch of the North, who is quoted in the 1900 text, The Wonderful Wizard of Oz, offered the truism "It's always best to start at the beginning." Consequently, on their journey to a four-year undergraduate degree, the first discussion I have with our new freshman students is the question of "How Many?" We explore the width of corpulent king's thumbs to create the "inch" and the standardization of other monarchal body parts to get digits, nails, palms, hands, shaftments, spans, feet, cubits, yards and fathoms. The standardization of precise measurement tools enabled the concept of distributed manufacturing, wherein individual components of massive and complex creations could be fabricated in parallel anywhere on the globe and assembled at a speed that was inconceivable prior to the industrial revolution.

Some 200 years later, we find ourselves in the opening acts of revolutions in information and biotechnology. Rather than focusing on the industrial "big" of colossal factories and buildings, these technologies have their foundations in the "small." Tiny, nanometer-sized electronic components form complex integrated circuits while nanometer-sized proteins compose living systems and medicines. The micrometer precision of high-quality manufacturing is now 1000 times larger than the nanoscale precision required by these nascent technologies and is recognized by The National Nanotechnology Initiative (NNI), located at www.nano.gov, as one of the "Grand Challenges" that must be addressed. Much like the molded plastic rulers we used in grade school that have ridges representing 1/16 of an inch on one side and 1 mm on the other, standardized rulers are invaluable to the latest technologies. However, these ridges need to be a million times closer.

At these tiny separations, light is not simply reflected by the ridges. Rather, the ridge spacing is similar to the wavelength of visible light resulting in destructive interference that causes the photons to diffract from the surface. The result is that different colors of light appear to reflect from the surface at different angles in a pattern of dark and light regions and has been the basis of optical diffraction gratings since the early 1800s. Instead of creating reflective ridges, it is easier to create non-reflective grooves by scratching the reflective surface with a suitably hard material, such as diamond.

Problems arise when the machine tool is created to manufacture the ruled grating since grating quality degrades if the grooves are not perfectly parallel and the spacing is not exact over the entire surface area. In 1880, Albert Michelson invented the optical "interferometer" that permitted two single beams of light to interfere when they were overlapped with a precision on the order of the wavelength, a few hundred nanometers. Using the interferometer to measure position accurately is the basis for a device known as a displacement measuring interferometer (DMI). MIT Professor George Harrison equipped his grating machine with a DMI in 1955 to provide feedback on the position of the diamond tool, dramatically increasing the precision of ruled gratings and this has been the state-of-the-art for the past 50 years.

The mechanical process is very slow and the diamond tool must be replaced after carving 15 km of grooves. A modern 12-in diameter, 400-nm grating requires a groove distance of over 230 km and would take months to complete while requiring the minimization of environmental noise in the machine and the laser-based DMI. To address these problems, MIT Research Scientist Mark Schattenburg has led his team to explore the utility of interference lithography (IL). In this process, two coherent laser beams interfere to create an interference pattern of alternating constructive and destructive fringes that are used to expose a photosensitive material. The process contains no moving parts and the entire grating is exposed simultaneously. Unfortunately, the ultimate size of the grating is limited as the interference pattern is circular and only a small portion can be used to expose nearly-linear grooves.

The Schattenburg group has invented a hybrid mechanical/IL machine called the "Nanoruler." A small, but linear IL region containing hundreds of fringes is scanned over the surface of a 12-in diameter photosensitive substrate to create the same 400-nm grating in 20-minutes, greatly reducing the time window over which noise must be minimized. This scanning beam interference lithography (SBIL) device must operate in a vibration-free environment that is temperature-controlled to 5 millidegrees C, but is capable of producing large 300-mm diameter gratings down to 150-nm grove spacing with a precision on the order of 1 nm in less than a half an hour. Modern all-optical DMIs utilize bulky, high-finesse laser systems complete with their own noise environment. Linear optical encoders that read ridges like our plastic mm rulers are less noisy and currently have a precision of a couple micrometers. A DMI based on the precise gratings made by the Nanoruler would be inexpensive and equivalent to having a linear optical encoder with a precision on the order of 1 nanometer. As SBIL and the Nanoruler continue to mature, one of the Grand Challenges has been met and developments can begin in several new industries.

William L. Weaver is an Associate Professor in the Department of Integrated Science, Business and Technology (ISBT) at La Salle University in Philadelphia, Pennsylvania, USA. As a Contributing Editor for Scientific Computing magazine, his latest column can be found at www.ScientificComputing.com.

April 2005 Editorial

williamlweaver — Tue, 08 Jan 2008 07:43:11 CST

Contributed editorial appearing in
Scientific Computing & Instrumentation 22:5, April 2005, pg. 12.

Aim High
The HIAPER Project for environmental research

By Bill Weaver, Ph.D.

HIAPER will allow scientists to conduct experiments on the atmosphere at 50,000 feet while traveling Mach 0.8.

The other day, I was reading an e-mail from the comfort of home on my wireless PDA. The topic was the appropriateness of a high-tech focus in an elective course for business majors. The authors did not think it was a great fit for business students as they would probably not work in high-tech fields. Thankfully, I was sitting down or I may have hit my head on the bathroom sink as I fainted.

The topic of "high-technology" is also being discussed by a school committee tasked with incorporating it into the classroom. One of our early impediments was defining high-tech in the first place. Too often, folks think it is using expensive gadgets for the sake of using expensive gadgets. High-technology professionals identify and use it for what it is — a tool that increases task convenience. If a high-tech solution makes your task more difficult, then it is not the correct solution. High-difficulty tasks are performed and made possible through the utilization of high-technology tools. The ability to communicate with my team through the air of my home is definitely a convenience, even though it does require the use of some modern gadgetry that was managed, marketed and sold by business majors.

Thanks to the availability of high technology, the National Science Foundation's National Center for Atmospheric Research (NSF/NCAR) can conceive of conducting experiments on the atmosphere at 50,000 feet while traveling Mach 0.8. This new flying convenience is disguised as a modified Gulfstream V (GV) aircraft known as the High-performance Instrumented Airborne Platform for Environmental Research (HIAPER). The structural modifications include the affixation of aperture pads and plates, fuselage mounts, upper and lower optical view ports, wing hard points, pylons, and pods, window blanks, and an external upper tail deck. All additions and modifications were made to accommodate a large number of proposed instruments while insuring the flight worthiness of the aircraft.

The basic GV itself was delivered with a profusion of high-tech equipment necessary to insure safe operation. It is outfitted with a flight management system (FMS) that integrates displays, navigation, performance, guidance and flight sensor systems including dual cockpit monitors, avionics and guidance computers, three micro air data computer (MADC) sensors and triple inertial reference systems (IRS). Two global position sensors (GPS) communicate with the NAVSTAR GPS satellite constellation to determine precise three-dimensional aircraft position in coordination with two redundant distance measuring equipment (DME) systems that provide distance, time to station, ground speed and station identification.

The dual, solid state air traffic control (ATC) systems reply to the automated interrogators of the Air Traffic Control Radar Beacon System (ATCRBS) for tracking, identification and altitude reporting while the on-board Traffic Collision Avoidance System (TCAS) communicates with other transponder-equipped traffic in the vicinity that may present a collision hazard. The color Primus X-band digital weather radar system scans for adverse weather and turbulence while the Ground Proximity Warning System (GPWS) provides terrain-ahead awareness and windshear detection.

Once the HAIPER GV pulls away from the gate, it must provide for the power, gas and data needs of the experimental instruments as well as the comfort of the experimenters. The galley, lavatory and gas cylinder storage areas are located at the rear of the plane. Integrated drive generators (IDGs) in each of the two Rolls-Royce turbofan jet engines provide regulated 60 Hz 115 VAC single-phase, 400 Hz 115/208 three phase, and 28 VDC power through secondary power distribution boxes (SPDB) located in the main cabin, the nose, the baggage compartment and at each of the six wing hard points.

The GV Aircraft Data System (ADS) provides distributed data sampling modules (DSM) through 16-bit ISA and 32-bit PCI PC-104 architecture. Analog signals are acquired at per-channel software selectable sample rates ranging from 10 Hz to 10 kHz at 14 to 16 effective bits of resolution. Digital inputs are provided in serial, parallel and event counting formats using RS-232/422, ARINC-429 and HDLC protocols. Networking services are provided by an onboard gigabit Ethernet over Cat-6 twisted pair wire and optical fiber cables. System time-of-day is acquired through the GPS and broadcast over the Ethernet in IRIG-B or network time protocol (NTP) in addition to a 1-PPS start-of-second signal for synchronization. Autonomous instruments may also be controlled from remote ground stations through a 128 K bit/sec satellite communication system.

The instruments under development for flight on the HAIPER GV include small ice detector probes, radiation and ozone measurement, trace organic gas analyzers, laser spectrometers, high-resolution lidar, mass spectrometers, temperature profilers and cloud particle imagers. Even before it takes to the skies on its initial flight this June, the HAIPER GV will be one of the highest-technology gadgets around.

William L. Weaver is an Associate Professor in the Department of Integrated Science, Business and Technology (ISBT) at La Salle University in Philadelphia, Pennsylvania, USA. As a Contributing Editor for Scientific Computing magazine, his latest column can be found at www.ScientificComputing.com.

March 2005 Editorial

williamlweaver — Tue, 08 Jan 2008 07:38:13 CST

Contributed editorial appearing in
Scientific Computing & Instrumentation 22:4, March 2005, pg. 12.

Creeping Crawlers
Autonomous adaptive data mining agents

By Bill Weaver, Ph.D.

Measurement scientists deal constantly with the problem of too much data. Faced with a large pile of ore, it is legitimate to ask how much of the desired material it contains. A reasoned sampling strategy coupled with the appropriate statistics can yield an average value that describes the entire collection without having to analyze every bit of it. After the ore is processed, the actual value can be compared with the average prediction. This procedure works well when the entire collection will be processed, but a more difficult question arises when it comes time to decide where in the ground to obtain the next load of ore. In this larger picture, each pile of processed ore represents a small sample of the entire region. Ore rich in the desired material frequently is not distributed randomly, but is located in neighborhoods of quality ore. Rather than merely commenting on the average content of the region, the successful mining operation can excavate along veins of superior ore without having to process entire mountains.

This same problem arises in the realm of information. Data mining shares a strong analogy with material mining, as the valuable objects are distributed throughout a large collection of undesirable material and must be hunted, located and retrieved. Much like the large pile of ore that is processed entirely, search engines such as Google attempt to analyze and catalog every page of the World Wide Web and permit users to search their archived index for the information they are seeking. However, unlike traditional mining, the Web is a dynamic place whose content and size changes rapidly. An index-based search engine must continually expand its storage capacity and revisit pages often to catalog newly added and updated information.

To assist index-based search engines, several researchers are developing "topical-crawlers" — algorithms that surf the actual Web (not an index) looking for links to pages containing information relevant not only to a short query, but also similar in context to topics and user profiles. Different strategies are being pursued and each has its advantages. The first such crawler was developed in the mid-1990s, known simply as WebCrawler. It uses a Breadth-First algorithm that keeps a list of hyperlinks found on each Web page that it follows sequentially in search of query matches. Soon after, the Best-First algorithm incorporated an evaluation of the hyperlink list. Links determined to be the most promising are promoted to the top of the list and irrelevant links are removed to make room for new candidates.

The PageRank algorithm is used by Google to rate the importance of pages returned by a query of its index. The number of Web pages that link to a particular page increases its rank and promotes it toward the top of the query results. Topical-crawlers have successfully employed PageRank as a condition for the Best-First evaluation algorithm, but the rank of each page must be updated dynamically, requiring a large amount of recursive calculation and the creation of a cataloged history of visited pages. The Fish-Search algorithm is a Best-First type crawler that assumes relevant information is found in similar neighborhoods of the Web, much like following a vein of rich ore in a material mine. The crawlers search more exhaustively in areas of the Web where pertinent links are found and abandon searches that do not produce relevant pages.

With the goal of infusing computational intelligence into the Best-First search, Professor Filippo Menczer of Indiana University and his colleagues have developed the InfoSpiders algorithm. An adaptive population of crawlers is initialized with pages retrieved from a traditional search engine. The seed pages are retrieved and analyzed for the number of topic keywords appearing in the vicinity of each hyperlink contained on the page. The frequency scores are used as input to a neural network inside each crawler whose single output is the one link predicted to be the most valuable. Each InfoSpider performs a jump to its chosen link, and this new page is scored according to the number of topic keywords contained on it. A high score increases the crawler's "energy level" while a low score decreases it. In this way the InfoSpider receives direct feedback of its performance and uses a back-propagation learning strategy to update the weights of its neural net. Agents whose energy falls below a set threshold are declared dead and removed from the population, while those agents with high energy reproduce according to a specific inheritance and mutation strategy intended to spawn offspring with enhanced abilities to find high-energy pages.

Professor Menczer's research group has coded the InfoSpiders algorithm into an experimental prototype multi-threaded Java applet called MySpiders that is available for use at myspiders.informatics.indiana.edu. With advancing research into autonomous adaptive data mining agents, perhaps the material mining industry can benefit from the technology and the announcement of a robotic exploration system called MyGophers may be on the horizon.

William L. Weaver is an Associate Professor in the Department of Integrated Science, Business and Technology (ISBT) at La Salle University in Philadelphia, Pennsylvania, USA. As a Contributing Editor for Scientific Computing magazine, his latest column can be found at www.ScientificComputing.com.

February 2005 Editorial

williamlweaver — Tue, 08 Jan 2008 07:29:48 CST

Contributed editorial appearing in
Scientific Computing & Instrumentation 22:3, February 2005, pg. 8.

Turning up the Heat
Melanophila acuminate and biomimetic sensors

By Bill Weaver, Ph.D.

The Holroyd Science Center here at La Salle was designed in the 1950s according to the adage "Form Follows Function." The venerable structure has educated generations of successful scientists, doctors and teachers in the natural sciences, but it is slowly losing its efficacy. Like myriad other university science buildings, the disciplines of geology, physics, biology and chemistry are separated by floor, each having its own department office, equipment room, lecture halls, laboratories and student lounge. Mathematics and computer science are housed in a separate building altogether. This form reflects the analytical thinking prevalent in the last century at a time when the word "interdisciplinary" was scarce, and it perpetuates the soft bigotry between the disciplines. Our integrated science, business and technology (ISBT) degree program does not acknowledge any differentiation among the sciences and is currently skirmishing to dissolve the barricades between the schools of Arts & Science, Business, and Engineering.

Medical research benefits widely from amalgamated programs such as bioinformatics and computational chemistry. Informatics allows researchers to scour vaults of experimental data for broad trends and correlations while predictive computation of intermolecular forces and energy configurations permit the rapid screening of innumerable candidate drug compounds in silico. The symbiosis among disciplines also advantages computer science. Several optimization methods, including genetic algorithms (GA) and particle swarm optimization (PSO), are patterned after biological processes. Even though integrated electronic circuits can perform calculations a million times faster than their physiochemical counterparts found in biology, nature has developed elegant solutions permitting the tiniest organisms to perform tasks far beyond the capabilities of our best technology.

Patterning the design of processes and structures after biological solutions is known as "biomimetics" - literally, "mimicking life." In addition to computer algorithms, engineers are using biomimetics to design next-generation materials and sensors. Entomologists continue to marvel at the ability of the jewel beetle (Melanophila acuminate) to sense forest fires from miles away. The beetles are attracted to the smoldering wood in swarms to mate and lay their eggs in the newly deceased trees. The recently burned wood has no natural defenses against the beetle larvae that feed on the layer of plant tissue just under the protective bark. Zoologist Helmut Schmitz of the University of Bonn has determined the beetle is not attracted by smell or sound, but is able to detect infrared (IR) radiation in the range of 3 - 5 µm produced by a 100,000-square meter forest fire from a distance of 12 km. Distances of up to 80 km have been reported in the literature; eclipsing the capability of currently-available environmental IR detectors.

Instead of exploiting the band gap of a cooled semiconductor or the amplification of an avalanche photodiode, Melanophila uses an array of photomechanical sensors called "sensilla." Each sensillium is comprised of a wax-filled spherical cavity that is connected to the input of a single nerve cell. IR radiation is absorbed by the wax of this biological transducer causing its temperature to increase. This produces an increase in cavity pressure and volume that ultimately triggers an impulse in the nerve cell. With the goal of developing sensitive forest fire detectors, researchers in Dr. Schmitz's lab are measuring the expansion of a layer of simple polyethylene film when exposed to IR radiation with the use of a piezoelectric crystal. Initial prototypes are not as sensitive as commercially available cryogenically-cooled IR detectors. However, they operate at room temperature, making wide deployment feasible.

Researchers at the Materials and Manufacturing Directorate of the Air Force Research Laboratory have used the structure of Melanophila's IR sensilla to create a synthetic IR detector for military applications. They have chosen to base their sensor on bacterial thermoproteins, biological macromolecules that expand when excited by IR radiation, which can be genetically engineered and produced in the laboratory via fermentation. Without a synthetic nerve to sense the thermoprotein expansion, the researchers are investigating the use of circular dichroic (CD) spectroscopy, a technique that measures changes in protein secondary structure. They are also sandwiching a thin film of thermoprotein between an IR-transparent substrate and vapor-deposited thin gold film. Thermoprotein expansion alters the angle of a laser beam reflected by the gold film and this change is used to quantitate the amount of IR radiation.

La Salle will break ground for its new Science & Technology Center this autumn. The layout will reflect current trends in collaborative, interdisciplinary research and include multiple-use laboratories, configurable research space, shared analytical equipment and brainstorming conference rooms. Now, I must find a way to biomimetically justify the incorporation of a Dunkin' Donuts and Starbucks.

William L. Weaver is an Associate Professor in the Department of Integrated Science, Business and Technology (ISBT) at La Salle University in Philadelphia, Pennsylvania, USA. As a Contributing Editor for Scientific Computing magazine, his latest column can be found at www.ScientificComputing.com.

William L. Weaver, Ph.D. - Recently Updated Pages

Affiliations

Research Collaborations

Technical Consulting Services

NoThankYou

1987 - 1988 Graduate School Rejections

1987 - 1988 Graduate School Acceptances

1991 - 1993 Job Search Rejections after Ph.D. Degree

1991 - 1993 Job Offers after Ph.D. Degree

1998 - 1999 Academic Position Search Rejections

1998 - 1999 Academic Position Offers

1999 - Present Publication & Presentation Rejections

March 2006 Editorial

2006 Columns

February 2006 Editorial

2005 Columns

Columns in Scientific Computing

2008 Columns2007 Columns2006 Columns2005 Columns2004 Columns2003 Columns2002 Columns2001 Columns2000 Columns

January 2006 Editorial

November 2005 Editorial

October 2005 Editorial

September 2005 Editorial

Vitae

Publications

Home

Contents

August 2005 Editorial

July 2005 Editorial

June 2005 Editorial

May 2005 Editorial

April 2005 Editorial

March 2005 Editorial

February 2005 Editorial

2008 Columns
2007 Columns
2006 Columns
2005 Columns
2004 Columns
2003 Columns
2002 Columns
2001 Columns
2000 Columns