What Inspired us?

The machine learning and its related area are developing very fast at present. And the industry needs more and more young talents to join and dedicate. However, there is a blank in the machine learning education which specifically targets young children. Therefore, we design and develop this game to inspire young children and draw their attention and interests to machine learning.

What is it?

This is a gam in which children can design his/her own neural network, adjusting parameters and structure. Then the child will use the AI designed to compete against other AIs. The children is expected to learn the overview and reasoning of neural network from this game.

What

Using custom-made neural network library on NEAT, we add interface to allow editing the parameters and structures. This was then forwarded to front-end Unity application which rendered the game and UI. All the work was completed within 24 hours. Certain Interfaces can be used with other machine learning algorithms to allow for modularity and ease of implementation.

Challenges we ran into

Due to the target customer being children, it is important to keep the interface simple to use while retaining the required complexity of the algorithm. The art style designed to be cartoonish, to create a more comfortable user experience. This was implemented using low-poly models and bright colors, in which none of our group members have experience working with.

Accomplishments that we're proud of

An efficient c# NEAT library written in only 24 hours. Custom-made low-poly 3D models. A simple and easy to use interface implemented using dragging the required elements.

What we learned

Each member of our group experienced and learnt difference things about the development process. This includes implementations of the NEAT algorithm developed by Ken Stanley in C#. But most importantly, we learnt a lot about project management and operating under stress. We wasted a lot of time due to miscommunication about the interface, and this could be prevented by better management and communication between our group members. The allocated time of 24 hours was very tight considering the complexity of the NEAT library. This was further constrained by the downtime of the group members due to mismanagement.

What's next for NNKids

As the interfaces for the algorithms are usable, different machine learning algorithms can be implemented easily and quickly. This allows further education of the target audience. We can also implement the algorithm using different games, which allows a greater user retention given the short attention span of the target audience.

Share this project:
×

Updates