CSM212: Neural Networks Assignment
Aims and Deliverables
The aim of this assignment is to implement a neural network
capable of learning a function that maps an 8x8 bitmap
to the name of a character.
Final deliverables should include
- Program source code
- Report including documentation and statistics
- Division of labour should be indicated (if applicable)
Deadline
Monday 25th May 12.00 in Marlene's Office
Approach
The assignment splits naturally into two sections. Team efforts
(max. 2 people) could adopt the following division of labour:
- Neural Network Simulator
- Compute output from input
- Update algorithm using back-propagation.
- Parameters (e.g. learning rate; activation function; thresholds).
- User Interface
- Flexible Network Configuration (e.g. 2 layer; 3 layer)
- Representing Training/Test Data
- Performance measurement and results
Training Data
Digits Training Data (in ascii format)
can be downloaded here.
Test Data
You are expected to construct your own test data (e.g. by
adapting the training data).
References
Russel and Norvig chapter 19 (already handed out)
contains just about everything you need to carry this assignment out.
CSAI Home
CS-212 Home
Mike Rosner (mros@cs.um.edu.mt)
Last modified: Fri Apr 17 12:21:02 MET DST