New Models

New models in this release ! No need to create a model with TFJS, now a wrapper named NeuralNetwork can do the job for you, you just need to give the parameters you want. Also, added convolutional neural networks and layers (dropout, flatten, maxpooling2d, conv2d).

Features

• Added new classes for creating Networks (NeuralNetwork, ConvolutionalNeuralNetwork)
• Added new interfaces for creating layers configurations
• Added convolutional layers
• Added Dropout layers
• Added Flatten layers
• Added MaxPooling2D layers
• Possibility to import from TFJS array of layers (Model.FromNetwork)
• Possibility to export to a TFJS array of layers (network.createLayers())
• Possibility to import from a Keras/TF model

... And more to come !

0.0.1 - Launching ReImproveJS

ReImproveJS v0.0.1

This library aims to create an entire world of reinforcement learning with Javascript using TensorFlowJS.

Current features

• Create an academy to host a place for agents to evolve
• Create Teachers to teach to agents, and configure them to change the way agents will learn with a teacher
• Assign agents to teachers and start learning
• Send data to teachers and let them do the job !
• Rest and gaze at your evolving agents.