DeepMathematics

Inspiration

DeepMath is a Machine Learning based probability problem generator made for people who are new into the field of probability, providing them with unique moderate level problems. This program is made to educate and demystify the difficulty of introductory probability.

What it does

DeepMath takes a math dataset containing certain probability problems and generates new problems based on the data. It uses tensorflow and Natural Language Processing to create the new problems, the Jaccard Distance metric was used to create new problems. The solutions is created by another model that evalutes the newly generated problems using probabilistic neural networks

How we built it

Sympy and Numpy handled the mathematical evaluations, making the process of generating new answers for our dataset a lot simpler. Tensorflow did the NLP part, using our previous dataset to create a feed-forward network that utilized the Jaccard Distance Metric to generate new problems. The entire code base is written in Python. It utilizes some of the fundamentals of probabilistic problem solving

Challenges we ran into

Just creating and testing the models overall was a huge time sink, it took us so long to get the model working, which is why our UI is so basic

Accomplishments that we're proud of

We are really proud of the overall model itself, it took us so long to have it running and working, overheating our laptops trying to optimize the models, the overall math that went into figuring it all out

What we learned

We learned that creating probabilistic problems is a lot harder than solving them. Creating problems actually requires the fundamental understanding of what makes probability so fascinating and yet so challenging