Background

Machine learning has had a meteoric rise recently. As a result, more and more people are implementing deep learning models. This recent spike in interest is great, however far too many beginners find them selves simply copying and pasting entire models of deep learning. Copying and pasting deep learning models is easy but dangerous as it prevents a true understanding of what's going on behind the scenes.

What is it?

Thus, we built a graphical editor for deep learning networks, for beginners to import their models and view what these models actually look like, and even interact with it. This environment allows a user to load in a Keras model.json, and easily view the structure of their model and even edit it.

How?

We implemented this with a sample multi layer perceptron as a base model. However, this program is primarily more useful for more complex networks such as recurrent and convolutional neural networks. Our prototype demonstrates this visual editing for deep learning is possible and useful.

What's next?

This is a prototype using a simple neural network model. A multilayer perceptron, what we implemented, is fairly easy to conceptualize. Thus, this is a demonstration that paves the way for more complex models, such as convolutional neural networks and recurrent networks.

Built With

Share this project:
×

Updates