Dynamically define, train, and run models during gameplay using a state-of-the-art tensor library.
AILIVE pairs the power of Unity's robust realtime rendering and UI pipelines with DyNet's dynamic neural network training to enable you to build flexible visualizion tools to help you monitor and debug traning instances.
Training a model at runtime means that you can build user-driven reward systems, adapt AI behaviors to new information on the fly, and develop responsive player interactions only possible with machine learning.
DyNet can change the way it learns at runtime, reconfiguring to suit variable data and dynamic environmental factors. This is particularly useful for natural language processing, where text might be very different in length.