AILIVE - Runtime Neural Networks

We just updated AILIVE to include saving & loading trained algorithms and updated to use DyNet 2.0, but since I haven't posted in this forum before, I'll give the whole overview!

​

AILIVE enables you to create machine learning algorithms at runtime, which means the player can interact and influence the learning process- enabling a whole new set of gameplay verbs! We have a ton of information on our website, including information that can help you think about how to effectively use machine learning in your project and documentation for our demo.

AILIVE is a great way to start experimenting with machine learning and see results interactively!

Our Neural Net Starter Demo
- Code is well commented
- Website includes documentation for our demo project
- Versatile neural net can be adapted to other projects
No Pre-Training or Big Data Necessary
- Although you can import data, AILIVE is built to learn fast during play
Power of C++, Code in C#
- AILIVE is built on highly efficient compiled C++ code
- Fast computation of neural networks reduces impact on framerate/loading times

Machine Learning veterans will find AILIVE a very powerful tool as well!

Integration for DyNet
- Dynamic Neural Network library, similar to PyTorch
- Well documented & familiar
- Extremely flexible. Write the neural net YOU want!
Leverage Unity for Visualization
- Running live, you can use Unity’s UI tools to visualize the learning process.
- Find unexpected behaviors and track the algorithm’s effectiveness at a glance
- Showcase your results to people who are less savvy at dissecting traditional debug information

The best way to see it in action is to try it out and see the results of teaching on the fly, but here is a taste of our sample project, Lab Rat Race.

​

And this chart is a quick guide on how the possibilities of using Machine Learning as part of a live game loop! There is more explanation on our website.

​

For any new updates, follow our Twitter!
For more information, check out our website!
And of course, take a look at our asset store page!