loredart_nn 2.0.1

Small package for creating, training and saving neural network, written in pure Dart. Supports MLPs and CNNs.

2.0.1

• Small fixes in README and comments

2.0.0

• Fully rewritten logic based on new Tensors and new backpropagation logic (implicit linear computational graph)
• Extended supported optimizers (with weight decay), layers, activation functions etc
• This version is not compatible with previous ones

Some points that still needs to be worked upon:

• progress bar for model training
• model saving (now only Json is supported)
• support for parallel computations, etc.

0.0.2+1

• Assets update

0.0.2

• Add implementation of Minibatch Gradient Descent
• Change and update DNN logs and behavior

0.0.1+2

• Mini update of code and format.

0.0.1

• Initial version. Support DNN with Dense layers, SGD optimizer, variety of activations, and losses.