-
Regression Customization Screen
-
Prediction Output Screen
-
Main Screen
Inspiration
We realized that there's a gap between scientists who need to analyze data easily and programmers who can actually use ML/DS libraries, so we wanted to bridge that gap with a graphics interface
What it does
Allows a user to input a dataset, select the independent and dependent variables to put into the algorithm and create a model. There are also options to plot data and predict a data point based on the model.
How I built it
Used Visual Studio for the front end and wrote python scripts that were called by the front end.
Challenges I ran into
A lot of errors occurred with the FastAI library, we had to come up with workarounds for those. There were issues when using datasets not provided by FastAI themselves.
Accomplishments that I'm proud of
Wrote a regression script that allows a user to plot the predictions vs the actual values and predict a datapoint.
What I learned
ML is reeeaaaally tough to do right, so many things have to go well, from the dataset to the library implementation to the outputs themselves.
What's next for MLite
Possibly rewrite to a JS front end, add functionality for PyTorch so that we can have more customization options.
Log in or sign up for Devpost to join the conversation.