Delivery

Delivery of the complete project for evaluation

Exploited matrices + regularization implemented

• Unit class not present anymore, computations are more "parallelized" exploiting matrix multiplications.
• Implemented L1 and L2 regularization

Cross validation implemented

2.1

Fixed bug with tanh

Backprop implemented

2.0

Updated requirements.txt

1st network demo (only inference)

The 1st demo of our network is working! 😃

Characteristics:

• Training and backprop are not implemented yet, only inference works.
• Weights initialization is random between 0 and 1
• Net's topology can be decided on net's creation
• Available activation functions: ReLU, Sigmoid