CMPU 395 -- Assignments

The assignments will require programming in Python, but the programs themselves are not that important. Focus will be on the insights gained from using various machine learning algorithms.

The assignments should preferably be made in groups of two. A short (about two pages) lab report with diagrams and comments on insights/reflections should be handed in before the deadline for each assignment.

• A few tips and tricks on using Python are assembled here.

The task is to simulate the activity in an actively firing neuron using the Hodgkin-Huxley equations.

• Instructions
• Template code: odesolver.py
• Report due: Friday, Sep 18

The task is to analyze how a Multi-Layer Perceptron classifies data by studying the decision boundaries.

• Instructions
• The mlp package from the book: mlp.py
• Report due: Friday, Oct 2

The task is to implement a Support-Vector Machine using a quadratic programming optimizer.

• Instructions
• Report due: Friday, Oct 16

The task is to build Self-Organizing Maps using topology preserving competitive learning.

• Instructions
• animals.py, cities.py, house.py,votes.py
• Report due: Friday, Nov 6

The task is to perform pattern completion and other associations using a recurrent attractor network.

• Instructions
• utils.py, figs.py
• Report due: Friday, Nov 20

The task is to use Q-Learning for finding an optimal behavioral strategy by exploring a given environment.

• Instructions
• Template for code showing the robot as a comic strip
• Report due: Friday, Dec 4