1.0.5

Changes

• implemented csnn.ELU activation function
• implemented method Module.apply to apply recursively a function over each layer
• implemented multilayer perceptron in csnn.feedforward.Mlp

Full Changelog: 1.0.4...1.0.5

1.0.4

Changes

• added csnn.LeakyReLU and csnn.SELU
• implemented method csnn.init_parameters for parameter initialization as numpy arrays
• implemented some convex neural networks in the csnn.convex submodule: csnn.convex.PwqNN, csnn.convex.PsdNN, csnn.convex.FicNN

1.0.3

Changes

• implemented Sigmoid, Tanh and GELU activation functions
• implemented RNNCell and RNN modules
• implemented Dropouts
• improved examples

v1.0.2

Changes

Major

• updated to Python 3.9
• implemented SoftPlus activation
• implemented example for Fully connected Input Convex Neural Networks (FICNN)
• implemented example for PieceWise Quadratic networks (PWQ)

Minor

• better project management, dependencies, tests and readme

v1.0.1

Changes

Minor

• minor performance improvements
• better testing
• better dependencies versioning

v1.0.0

First Release