Exam/Homework/Reading Assignments:

• HW 01 Designing McCulloch-Pitts neuron (posted 01.17.2008)
  • Solution pdf
• HW 02 Perceptron, hard/soft activation functions, delta rule (posted 01.28.08)
• Solution to problem 2.4 covered in class (see the lecture slides)
• Selected submitted solutions, not necessearly awarded with full credit 1, 2, 3, 4, 5
• HW 03 Feedforward bipolar neural network design, NNOTT (posted 02.20.08)
• Solution pdf
• Selected submitted solutions, not necessearly awarded with full credit 1, 2, 3, 4)
• Exam #1 Preparatory examples (posted 02.20.07)
  • Solution is attached to your graded assignment
• HW 04 AD converter, fuzzy logic (posted 04.09.08)
• Solution pdf
Exam #2Preparatory examples (posted 04.23.08)

Before you send me an email, please check the FAQs (your answer might already be there).

Paper:

• Paper topic: Neural networks are known by their inherent ability to deal with wide range of problems, such as classification, prediction, approximation, even when the outcome is not known. ECE students may be interested in circuit implementation of certain neural architecture. Ecohydraulics students may be interested in specific modeling, classification, prediction (watershed, etc.). You may want to do a preliminary literature search first.

• Type of paper: Required paper can have more of a research (theoretical) or more of a project (applicative) flavor.

• Format: Formatting should be according to an IEEE conference guidelines. It should be composed of title, abstract, introduction, problem definition, proposed solution, experimental/testing, conclusion, and future directions.
  Some examples are:
  • IEEE Computer Soc. Publishing Forms page (see Formatting Instructions);
  • IEEE Author Tools page with Template for IEEE Transactions with template in MS Word or pdf format and WIN and MAC Bibilography file.

• Submission: Please follow these rules when submitting your paper:
  • Please compile all your results into a single file!
  • In email, use the proper subject line and signature as stated in course web page.
  • Include the same signature on every page of your assignment.
  • Use the following convention for naming the file: " HWnn_Family_name.xxx ", nn being the homework number.

• General tip: After you choose a problem, give a problem definition on a general level first. Then decompose it to feasible components and tackle one after another. In other words, put some boundaries on a problem and keep it reasonably difficult for a given deadline.

• Teaming up: If you are a member of a team, first select a project manager for your team. He or she will be submitting paper drafts and questions for all of you. He or she will also be cc-ing to all of the members of your team. Hence, when I reply I will be replying to all. Paper title page will contain all the team members names, as well as the detailed description of division of work in team (who is doing what).

Outreach students: Outreach students are encouraged to team up with other outreach students or students taking the course live/cv. Though EO students can assume deadlines 10 days longer than the ones posted below, I would encourage to use the same final submission date. This way, your teammates can present your papers during the last session of the course.

 

• First draft due: February 12, 2008.
  • Deliverable: brief outline, up to one page long. It should include:
    • working title,
    • author(s) contact info
    • short abstract, and
    • problem description (problem statement). It is very important to provide as precise as possible problem statement. I would encourage providing both narrative and mathematical problem description. 
  • Formatting: at this point, your deliverable should look like an abstract you would submit to an IEEE conference (please refer to format specifications from above).
  • Content: Initial proposal should be limited to a problem description. At this stage, a problem statement is a main task. You may describe a set of related problems instead of a single problem. At this point in the course, you should have a good feel for problems where neural networks will be a superior solution. However, an elaborate discussion on possible neural algorithms should be left for the second draft. I will also be assisting you in identifying neural network based solution(s) that would be a good fit for your problem.
  • Topics: Any application related to learning, adaptive control, classification, pattern recognition, and many others would be a good choice (see some ideas listed below). I am hesitant to force you into a list of topics – I would rather see you coming up with something along the line of your current project problems, thesis or dissertation, or something you would be simply interested in working on. I will be happy to discuss topics with you - feel free to contact me anytime.

• Second draft due: March 03, 2008.
  • Deliverable: in addition to the content from the first draft, detailed literature (background) review should be performed for this phase.
  • Formatting: you are planning on publishing your paper at IEEE conference, so the formatting applies accordingly.
  • Based on literature review that you have done, you should elaborate on other comparative approaches to the same problem. This is important because: 1) it documents that you understand the current state of the art of your particular problem; 2) helps you identify deficiencies with current solution approaches. These deficiencies you will be attempting to alleviate.
  • This draft should contain updated problem description. The survey of existing solutions needs to be referenced. You may start elaborating on steps you are planning on taking with regards to your problem solution.
  • This will be a more elaborate paper, but not more than 2-4 pages long. At this point, you will have the following components of your paper:
    • title (revised, if necessary)
    • author(s) contact info
    • abstract
    • I. Introduction and background (literature review)
    • II. Problem statement (updated)
    • V. References
  • At this point a solution proposal does not need to be presented.
  • Division of work in team: please clearly state the division of work within a team (the work performed by each team member).

• Third draft due: March 31, 2008.
  • Deliverable: in addition to previous draft, it should contain a technique (or selection of techniques), you are planning on using to solve the selected problem (III. Solution proposal section). The solution(s) will entail specific learning algorithms, architectures, discussion on advantages and disadvantages of each, etc.
  • Formatting: IEEE conference formatting.
  • This will be a more elaborate paper, but not more than 5-6 pages long. At this point, you will have the following components of your paper:
    • title (revised, if necessary)
    • author(s) contact info
    • abstract
    • I. Introduction and background (literature review)
    • II. Problem statement (refined, in necessary)
    • III. Solution proposal
    • V. References
  • At this point a final solution and experiments do not need to be present.
  • Division of work in team: please clearly state the division of work within a team (the work performed by each team member).

• Final submission: April 22, 2008.
  • Deliverables:
  1. Final, complete paper, up to ten pages. It should contain detailed solution and experimental work. You should include test examples (learning & test patterns), architecture description, discussion on robustness of your solution. You should also include future work directions, problems that you may want to tackle in next paper.
  2. At this point, you will have the complete paper:
     • title (revised, if necessary)
     • author(s) contact info
     • abstract
     • I. Introduction and background (literature review)
     • II. Problem statement (refined, in necessary)
     • III. Solution (description)
     • IV. Test results
     • V. References
  3. Please submit copies of papers/books used in your research.
  4. Please name your file like this: FamilyName_Paper_xxxx.xxx and FamilyName_Presentation_xxxx.xxx

• Presentations: last week of the course (April 24, April 29, 2008).

• Each of you will present your project and we will all discuss it. You should use this discussion to finalize your project paper and hopefully publish it at some conference.
• Presentation: please limit your presentation to 10min+5min for discussion. Along with final version you should email me your slides so I can post them on our course web page. This way others will be able to read about your project and prepare questions. Please print out all project papers & presentations before coming to the final session.

• Paper topic - some ideas
  • These are some of the topics of papers from previous years. These will give you some examples of what applications can be targeted.
    
  • Pattern recognition
  1. Pattern Recognition and Conversion of Japanese Hiragana to Roman Characters
  2. Using Neural Networks to Identify Disordered Regions for Protein Alignments
  3. An Experiment To See If Hopfield Networks Can Be Used in Handwritten Digit OCR
  4. Artificial Neural Network For Automated Prediction of Popularity of Digitized Images
  5. Pattern Classification Of Numericals By Neural Networks
  6. Pattern recognition using Hopfield associative memories
  7. Feedforward ANN for bipolar pattern recognition
     
  • Clustering
  1. Good Vibrations: Investigating Neural Network Applications in
     Prognostics of Jet Engine Ball Bearings
  2. Single Neuron Classification of Non-Linearly Separable Data
     
  • GA
  1. Evolving Neural Networks Using Particle Swarms
  2. CBAC Optimization for Artificial Neural Network (Cluster Based Averaged Crossover to converge in Particle swarm time)
     
  • ECE problems
  1. Implementing the 2-bit A/D Converter using the Counterpropagation Networks
  2. A Study of Advancements Made in Circuit Efficiency Using Threshold Logic
  3. An Artificial Neural Network Approach for Detecting High-Impedance Faults on Three Phase Four-Wire Distribution Circuits
  4. Neural protective relay for a circuit breaker.
  5. HDL modeling of neural networks
     
  • IDS
  1. Artificial Neural Networks for Intrusion Detection
     
  • CE/ME problems
  1. Bed Load Sediment Transport Estimation Using Artificial Neural Networks
  2. Use of Artificial Neural Networks (ANN) in a Stream flow Prediction
  3. Implementing a Neural Network System to Solve the Inverse Kinematics of a Biologically Inspired Robotic Cat Leg
  4. Neural Network Application In Engineering Principle
  5. Biodiesel Blend Level Sensing from the Ultraviolet Absorption Spectra with Application of Neural Network
     
  • Intelligent controllers
  1. Road recognition for an autonomous vehicle
  2. Control of Underwater Autonomous Vehicles Using Neural Networks
     

   

Final Information:

• Course web sites will be closed:
  • Husky class site will be closed after the semester is officially over. I will be granting temporary permissions upon the request.