Work History Part 7: RIT Electrical Engineering Dept.

Employer: RIT Electrical Engineering Dept.

Location: Rochester, NY

Position: Software / Firmware Engineer

Period: June 2010 - December 2010

Ever since I began my thesis work I've had a growing interest in applying my skills to the field of medicine and medical research. I find devices that interact with the underlying biomechanical processes of the human body fascinating. I enjoy learning about how the human body functions and technological advances ranging from neural modification to gene manipulation and artificial limb development.

Through the recommendation of a certain kindhearted software engineering professor, I was put in contact with the up-and-coming RIT Biomedical Engineering Dept. Within a matter of days I was hired by the RIT Electrical Engineering Dept. (which forms the backbone of the new biomedical engineering dept.) to begin working on the body area sensor network (BASN) project.

The BASN consists of multiple wireless nodes strapped to a user's body. Each node is about the size of a watch and the band is similar to a watchband. Each node collects biometric information, such as body temperature and limb orientation and acceleration, based on where it is positioned on the user's body. Each collector node wirelessly sends data back to a central access point. In turn the access point aggregates all of the data and wirelessly sends it to a USB dongle attached to a host machine where it is processed by a host-side application.

Each node consists of custom-designed PCB architecture and firmware API. A complementary API allows host-side applications to communicate with the nodes using the USB dongle. The centerpiece of each node is the Texas Instruments CC430 microcontroller. Wireless communication leverages the Texas Instruments SimpliciTI protocol.

The goal of the project is to implement a development platform that will allow future developers to create custom applications such as:

• Seizure / muscle tremor detection and patient caretaker alert system.
• Local positioning system using dead reckoning from a known starting location. Can be used to track firefighters as they enter a burning building.
• Wireless electrocardiograph (ECG). Current ECG models use 10 wired electrodes and are cumbersome.
• Soldier injury detection system. In a battle scenario, the command center can track the physical condition and stress levels of soldiers in the field.

Responsibilities and accomplishments:

• Developed 3D proof-of-concept applications demonstrating the ability to capture and process biometric data.
• Leveraged existing firmware to set up a low-power RF communication network between wireless nodes and a host-side USB dongle.
• Implemented a device driver in Linux to provide an API for interacting with the RF network through the host-side USB dongle.

Masters Thesis Part 3: Algorithm Overview

I represent each RF using a multivariate Gaussian distribution with a mean vector (position), covariance matrix (alignment and size), and gain factor (response magnitude). I determine the value of each of these parameters for each RF using a competitive algorithm where the result of a custom greedy algorithm competes against the result of a genetic algorithm. Gaussians are trained one at a time in a staged approach so that the addition of each new Gaussian to the model takes into consideration the effects of all of the previously trained Gaussians.

Actually, the genetic algorithm applies a little more intelligence by dividing a population of Gaussians into species based on similar characteristics and mating (i.e. crossing-over) the best performing Gaussian members within the same species during each generation of evolutionary training. The addition of each new Gaussian to the model attempts to address some residual error in the model resulting from recorded neuron firing rate response values that have not yet been fully explained by the addition of previous Gaussian distributions. The theory is that the best performing Gaussian of each species is associated with a different local minima in the residual error search space, thereby increasing the chances that one of those Gaussians is associated with the global minimum. The Gaussian which results in the greatest drop in residual error among all of the species is selected and added to the model.

The Gaussian model (also known as a Scaled Gaussian Mixture Model) is then used to extract features that are fed into a linear regression (LR) technique. Why is LR necessary on top of the Gaussian model? The reason is because not all RFs are the same. A great deal of research indicates the existence of hotspot segments in the visual field which have a major impact on the response values generated by RFs in non-hotspot segments. Specifically, the directional preference of RFs associated with hotspot segments plays a major role. Thus, the response value of RFs associated with hotspot segment needs to be weighted differently than those for non-hotspot segments and the response values each segment needs to be divided into directional contribution vectors that are each weighted differently.

Overall my algorithmic approach trains models that are capable of predicting neuron firing rate response values with moderate to high accuracy. As such my research offers support for a number of existing concepts in the field of visual neuroscience, such as the gradient hypothesis and opponent organization.

Read the full report (PDF) for a complete explanation.

Masters Thesis Part 2: Receptive Field Theory

So how did I go about creating a model capable of predicting individual neuron responses of primates to various patterns of moving white dots? The key was to use receptive field (RF) theory. In general, a RF is a physiological construct associated with an area of the visual field. Let's say you're staring straight ahead and perceive some motion across the upper right of your visual field. You may have a RF associated with that region such that as you perceive motion across it your nervous system reacts by speeding up (excitation) or slowing down (inhibition) the number of electric impulses between neurons in your visual cortex within a certain period of time.

Each RF is very simple in structure. Imagine a doughnut. The outer ring and missing center form 2 separate areas across which the RF responds to motion. There are 2 flavors of RFs. The first is on-center off-surround where motion across the center causes an excitatory response and motion across the ring causes an inhibitory response. The second is off-center on-surround, which behaves in the opposite way. Multiple RFs can be aligned in various ways to respond to various directions of motion. If you consider your visual field to be 2-D with an x-axis (left/right) and y-axis (up/down), you can place multiple RFs within that plane to respond to all kinds of motion.

An extension of the RF theory is that each RF has a directional preference. This makes sense if you think about how multiple simple on-center off-surround RFs and off-center on-surround RFs can be combined to form more complex RFs. With the addition of a directional preference, the 2-D visual field space becomes a 3-D RF space. The task then becomes a matter of placing RFs within the RF space, aligning them properly, sizing them properly, and understanding the relationships between them.

Read the full report (PDF) for a complete explanation.

Masters Thesis Part 1: Primate Experiments

My computer science thesis work (PDF) at the Rochester Institute of Technology involved analyzing the firing rate response patterns of neurons in the Medial Superior Temporal (MST) region of Rhesus monkeys, which resides above the temple on both sides of the skull. The primate experiments were performed by Dr. Charles Duffy and Dr. William Page at the University of Rochester Neuroscience Department.

Each primate was trained to sit in a special chair and stare directly at a red dot in the center of a rear projection screen covering 90 deg. x 90 deg. of his visual field. The rest of the screen was black. The primate was rewarded with sweet juice if he remained still and stared at the dot for 5 minutes. After successfully completing this task various patterns of moving white dots were shown in the screen, like snow in the wind (except with uniform density and constant velocity). The screen was divided into 9 segments (like the side of a traditional Rubik's cube). Each dot would appear at one edge of the screen and travel in a certain pattern before exiting at another edge of the screen.

The simplest patterns consisted of linear motion where all of the dots moved in the same vertical or horizontal direction across one of the 9 screen segments. More complex patterns consisted of linear motion in 2 of the 9 screen segments where the dots could be moving in different directions within each segment. The most complex patterns were designed to simulation self-motion and occupied all 9 of the screen segments. For example, imagine the old Windows starfield screensaver where the dots travel from the center of the screen outwards, which makes you feel like you're moving into the screen. A cylindrical piece of the primate's skull cap was removed and an electrode was positioned to record individual neuron firing rate responses as the primate perceived each moving pattern on screen.

My task was to design a mathematical model capable of predicting how the primate responded to self-motion patterns given his response to simpler patterns in regards to individual MST neurons. Why? To further our understanding the mammalian brain. Such a biologically-inspired model can potentially be used to program the AI for more realistic robots, or potentially be used to program the firmware for a cybernetic implant to help treat the sensory degradation symptoms of Alzheimer's disease patients.

Note that firing rate signals were captured in a region of the brain before any higher level cognition was performed. In other words, the response values were not affected by what the primate was thinking about when viewing the patterns on screen (i.e. bananas). Well, in theory that's what we'd like to believe.

Work History Part 6: RIT Industrial Engineering Dept.

Employer: RIT Industrial Engineering Dept.

Location: Rochester, NY

Position: LabVIEW Specialist

Period: December 2009 - May 2010

The Electrophotographic Development and Transfer Station (EDTS) fixture was a generous donation by Kodak to the Print Research and Image Systems Modeling Laboratory in the Center for Imaging Science at RIT. I briefly mentioned this job in a previous post.

The ultimate goal is to use the fixture as a tool to teach students about the fundamentals of xerography. Additionally, the fixture exposes the internal components of a xerographic printing press, which makes it a desirable candidate for running experiments that require modifying individual components, spaces between components, carriage speeds over specific stations, etc. Xerox expressed an interest in using the fixture to analyze the characteristics of small-particle developer (i.e. toner) and examining its impact on print quality.

I worked directly with the P10503 senior design team and representatives from both Kodak and Xerox. All of the team's documentation can be found on their website. The project poster (PPT) provides a general overview. My primary objective was to modify the existing rudimentary LabVIEW software in order to execute control over all of the various electrical and mechanical components of the fixture. When I began the fixture could not print an image. When I left it could. I call that a success!

Responsibilities and accomplishments:

• Led a 6-person team consisting of mechanical engineers, an industrial engineer, an electrical engineer, and a software engineer.
• Implemented LabVIEW software to control a flat plate xerographic printing fixture through NI DAQ hardware.
• Wired electrical and mechanical components, including stepper motors, DC motors, relays, high voltage power supplies, and electrostatic voltmeters.

NeuroEvolution of Augmenting Topologies

During my second quarter as a computer science grad I took a course on neural networks. As mentioned in a previous post, I had previously taken a course in pattern recognition. Since I already knew quite a bit about neural nets I decided to research the topic of NeuroEvolution of Augmenting Topologies (NEAT) instead of work on the scheduled class project (comparing the performance of feed-forward, recurrent, and radial basis function neural nets).

NEAT is a genetic algorithm for evolving artificial neural networks that protects innovation by dividing network types into species based on similar characteristics. During crossover new networks inherit nodes and node connections from their parent networks to form a new network structure. Node genes and connection genes are labeled with innovation numbers so that as those genes can be tracked as they are passed on to future generations. After crossover new networks undergo mutation where new nodes or new connections are added to the network and connections weights are adjusted by a random process.

Unlike traditional learning where connection weights are adjusted via a backpropagation technique, NEAT selects members from the spawned population, called genomes, for breeding. The evolved members are composed of nodes with sigmoidal activation functions that behave in the same way as traditional multi-layer perceptron networks.

Using jNEAT, a Java implementation of NEAT, I attempted to solve 2 different classification problems. The first problem was to classify 150 iris samples as 1 of 3 breeds based on 4 features. The second problem was to classify 178 wine samples as originating from 1 of 3 batches of wine based on 13 features. After conducting a number of different experiments using various configurations settings, I was able to generate a network with 96% correct classification for the iris test dataset and a network with 85% correct classification for the wine test dataset.

About a year later I went on to apply the concept of speciation to a genetic algorithm in my masters thesis.

Resources:

• Research report (PDF)
• Presentation (PPT)

Finite State Process Autocoder

As a software engineering undergrad who was part of the honors program at RIT, I was required to complete an independent study. At the time I was taking a course in concurrent systems and was making extensive use of the Labelled Transition System Analyzer (LTSA), a concurrent system modeling tool which uses Finite State Process (FSP) algebra notation to represent component behavior.

Once a system is specified in FSP notation, the LTSA can parse and compile the FSP notation to generate a graphical Labeled Transition System (LTS) model. The LTSA also provides a means to “animate” the system by performing discrete or shared process behaviors one at a time and stepping through the states of the system. Under the hood, the LTSA generates and stores an internal state machine representation of the system.

One of the major difficulties encountered by software engineers who use the LTSA is deciding how to translate the FSP system representation into an executable implementation. A common practice is to use the LTSA as a system design tool and to deviate from the FSP during the implementation phase in order to create a custom-tailored system from the ground up. The major disadvantage is that the implemented system may no longer exhibit the same behavior as the FSP and therefore will need to undergo an entirely separate, and more complex, verification process.

Clearly, it is advantageous if the system implementation reflects the original FSP design. At the time when I was conducting my independent study there was no widely accepted procedure for converting the FSP into source code. I began developing the an automatic code generation backend called the FSP Autocoder to use the LTSA to parse FSP notation and perform semantic analysis to produce code stubs for primitive concurrent objects.

Before I could begin working on the compiler aspect of the autocoder, I needed to design a “universal” (i.e. widely applicable) concurrent system architecture that could be traced back to the FSP system representation and vice-versa. I considered many approaches, such as employing a state machine architecture, rendezvousing threads in conjunction with a round-robin thread scheduler, and using a Communicating Sequential Processes (CSP) library.

Resources:

• Technical paper (DOC)
• Presentation (PPT)