BrainLab is a graphical user interface that lets users upload datasets to train a neural network with. Users can choose to create an account, save their trained models, and test new inputs against a model to predict outputs. They can also test inputs on the spot without saving the model.
A neural network is trained based on layers of inputs, outputs, as well as hidden layers, each containing a specific number of neurons which can be adjusted according to the complexity of the dataset, for increased accuracy. BrainLab gives its users a general overview of what their network looks like, without getting into too much detail, or requiring any prior knowledge from users about NNs.
Once the network is finished training, the user will see a graph that shows how the accuracy of the network progressed throughout the training. There is also data visualization for how the network classified the dataset, and whether a given sample was accurately classified (predicted vs actual). All visuals were created with d3.
Technologies: Python, Keras (a Python library), SQL, Node.js, Angular.js, Angular Material, and d3.
- Annika Sundberg, Ayako Mikami, Yoo Jung Dan, Winnie Chen