Docsity
Docsity
Prepare for your exams
Prepare for your exams

Study with the several resources on Docsity

Earn points to download
Earn points to download

Earn points by helping other students or get them with a premium plan

Guidelines and tips
Guidelines and tips
Sell on Docsity
Sell on Docsity
Docsity AI
Docsity AI
Log inSign up

Prepare for your exams

Study with the several resources on Docsity

Find documents

Prepare for your exams with the study notes shared by other students like you on Docsity

Search for your university

Find the specific documents for your university's exams

Docsity AINEW

Summarize your documents, ask them questions, convert them into quizzes and concept maps

Explore questions

Clear up your doubts by reading the answers to questions asked by your fellow students

Earn points to download

Earn points by helping other students or get them with a premium plan

Share documents
download point icon20 Points

For each uploaded document

Answer questions
download point icon5 Points

For each given answer (max 1 per day)

All the ways to get free points
Get points immediately

Choose a premium plan with all the points you need

Study Opportunities

Choose your next study program

Get in touch with the best universities in the world. Search through thousands of universities and official partners

Community

Ask the community

Ask the community for help and clear up your study doubts

Free resources

Our save-the-student-ebooks!

Download our free guides on studying techniques, anxiety management strategies, and thesis advice from Docsity tutors

The Neural Model - Advanced Digital Projects Lab | ECE 395, Study Guides, Projects, Research of Electrical and Electronics Engineering

University of Illinois - Urbana-ChampaignElectrical and Electronics Engineering
Prof. Lippold Haken

Material Type: Project; Professor: Haken; Class: Advanced Digital Projects Lab; Subject: Electrical and Computer Engr; University: University of Illinois - Urbana-Champaign; Term: Fall 2008;

Typology: Study Guides, Projects, Research

Pre 2010

Uploaded on 03/11/2009

koofers-user-xv6-1
koofers-user-xv6-1 🇺🇸

5

(2)

9 documents

1 / 4

Toggle sidebar

This page cannot be seen from the preview

Don't miss anything!

bg1
Skot Wiedmann
December 17, 2008
Advanced Digital Systems Lab
Lippold Haken
Introduction
The neural model for this project is based on a historic electrical
engineering PHD thesis. The model is accurately followed in spirit
rather than in exact components, and therefore the circuitry is
created with modern parts and redesigned to perform the desired
functions. There is almost no record of the original unit in actual use,
so a type of archeological reenactment will be the main goal.
This model attempts to accurately reproduce the function of brain
tissue, and therefore differs significantly from modern software neural
networks in several ways. Software models distinguish between
training and processing, where this model is simultaneously learning
and processing data. There are also mechanisms in this model that, in
the absence of significant activity, create data with no input, which is
consistent with unused areas in the human brain. This model is
capable of operating in a continuous time domain where software
models must obey the discrete time constraints of the computer
platform. All of these factors may have a significant impact on the
behavior of the neural model. This is an attempt to both return to a
forgotten approach in hopes of discovering something lost, and to
achieve something new through this unusual type of model.
The physical result will be a large console with a few hundred jacks
that will allow a user to patch together any neural network structure.
Pulses will be presented to the device and its response will be
monitored. Experimentation will be significant in exploring the
configurations and signals that will be possible. Ultimately, very
complex and parallel signals will be produced that correspond to the
brains activities. The unknown nature of these signals is the exciting
prospect of this project.
Circuits
The model divides each neuron into four different circuit elements:
Facilitator, Energy Transducer (increasing threshold), Energy
Transducer (decreasing threshold), and Autonomous Component.
This separation is useful because it allows less circuitry to create the
functional equivalent of more neurons by omitting the unused parts of
pf3
pf4

Partial preview of the text

Download The Neural Model - Advanced Digital Projects Lab | ECE 395 and more Study Guides, Projects, Research Electrical and Electronics Engineering in PDF only on Docsity!

Skot Wiedmann

December 17, 2008 Advanced Digital Systems Lab Lippold Haken

Introduction

The neural model for this project is based on a historic electrical engineering PHD thesis. The model is accurately followed in spirit rather than in exact components, and therefore the circuitry is created with modern parts and redesigned to perform the desired functions. There is almost no record of the original unit in actual use, so a type of archeological reenactment will be the main goal. This model attempts to accurately reproduce the function of brain tissue, and therefore differs significantly from modern software neural networks in several ways. Software models distinguish between training and processing, where this model is simultaneously learning and processing data. There are also mechanisms in this model that, in the absence of significant activity, create data with no input, which is consistent with unused areas in the human brain. This model is capable of operating in a continuous time domain where software models must obey the discrete time constraints of the computer platform. All of these factors may have a significant impact on the behavior of the neural model. This is an attempt to both return to a forgotten approach in hopes of discovering something lost, and to achieve something new through this unusual type of model. The physical result will be a large console with a few hundred jacks that will allow a user to patch together any neural network structure. Pulses will be presented to the device and its response will be monitored. Experimentation will be significant in exploring the configurations and signals that will be possible. Ultimately, very complex and parallel signals will be produced that correspond to the brains activities. The unknown nature of these signals is the exciting prospect of this project.

Circuits

The model divides each neuron into four different circuit elements: Facilitator, Energy Transducer (increasing threshold), Energy Transducer (decreasing threshold), and Autonomous Component. This separation is useful because it allows less circuitry to create the functional equivalent of more neurons by omitting the unused parts of

a particular neuron. By far the most work needed to be dedicated to the Facilitator because it is vacuum tube based, and it provides the sensitive memory for each neuron.

Power supply

The original power supplies for the unit were large dry cell batteries providing many different voltage rails. The first thought was to create on board references to eliminate many of the different voltage supplies that were necessary. However, after some initial design attempts, the cost of individual references on a dozens of boards was greater than the cost of adding another power supply. So the decision was made to keep all the original power supply rails, and simply replace the batteries with standard dc power supplies.

Facilitator

The Facilitator accepts pulses at it’s + and – inputs and produces pulses at its output. The pulses at the – input decrease the voltage on the storage cell while the pulses on the + input increase it. Additionally, the pulses at the + input are multiplied by the voltage on the storage cell and appear at the output. Therefore, the pulses at the

  • input are both increasing the gain through the Facilitator and acting as the data which is processed. In this way the system does not distinguish between training and processing data, or in fact, the training data is not distinguished from data to be processed. Since the Facilitator was the most challenging circuit, it was the first priority. After several attempts to analyze the vacuum tube section of the circuitry, it became obvious that the standard equations for triode operation were inadequate. This is because of the unusual way that it is forced to operate outside of its standard region. In doing so, several tasks are accomplished rather elegantly (and cryptically I might add) that required several components in the redesign. First, the triode is used as a very high input impedance buffer. This is the first and obvious use. The voltage on the grid is the memory of the sensitive storage cell. Secondly, the cathode resistor value was missing from the diagrams. Since this resistor is essential for calculating the gain through the triode, this had to be estimated and adjusted to a reasonable operation. The triode is also configured as an inverter, with negative excursions causing positive output. The grid could also be briefly brought into the positive region which caused grid current. This is often considered a bad thing because it drastically changes the input impedance and is outside the operating region of the triode. However, in this case it is used to discharge the

test probe solved this problem, and resistors were added from these points to ground. Realizing that adding a test probe solved the problem took some time and some luck. I would not be entirely surprised if the original circuit had this problem but it was not noticed.

Prototype

A hardware prototype of the Facilitator was created first on breadboard, and then on perf board. Sockets were used for the storage cell, which allowed different values of capacitors to be tested, as well as a lithium battery as on the original. Sockets were also used on the gain setting resistors, as well as the current limiting resistors for charging and discharging the storage cell. Overall the prototype works well and responds to pulses as expected.

Future Work

As of this writing, the other elements are still to be fully tested, however they will be much easier to adapt since they do not use vacuum tubes and have been analyzed and understood. Once they are prototyped, fabrication of the hundreds of boards necessary for this project may begin. Although the amount of circuitry is significant, some careful planning and design will allow construction to be tidy and efficient. Ultimately, patching together new structures and discovering the results will reward the user with an insight into neural activity.