Topic: Group 2 and group 26

Summary

Solving a binary classification problem by an artificial neural network as described by group 26:

1. Input: An m x n matrix.
2. Output: A parameterization of the neural network and a vector of probability values generated by that network.

Steps formulated by group 2:

1. Implement a perceptron that can be parameterized/trained.
2. Connect perceptrons in a dense feed-forward net and implement back-propagation.
3. Consider concurrency/parallelization.

Default recommendations

The recommendation is to use the ANNs as a compressed representation of multiple black-and-white images.

You can deviate from the default recommendations; follow them if you do not see any good reason not to.

Benchmark scenario

The benchmark scenario for the ANN topic is the same one as for the image compression topic.

Return to list of topics