EZML

Inspiration

Our inspiration for EZML is to help lower the barrier between individuals of different backgrounds interested in machine learning.

What it does

EZML at its current complexity can allow users who may not be experienced with programming using machine learning toolkits to easily build, train, and evaluate some simple, generic models that live on the backend. At its fullest potential, we imagine EZML allowing users to construct decent machine learning models that the application could transcend being merely a teaching tool, i.e. allowing a small business owner to make some basic predictions from sales data without any coding background.

How I built it

EZML is built with a python3 backend using Scikit-Learn and Tensorflow/Keras and a Java frontend using JavaFX.

Challenges I ran into

One of our big challenges has been getting our front and back end to communicate properly. So many file paths...

Accomplishments that I'm proud of

We are proud of the application we have built so far. As I am writing this we are still working to get our front and back end to work together properly, but both ends work well on their own.

What I learned

Some JavaFX, a little bit of ML, a lot about designing a decent sized project with components in two different languages that aren't necessarily made to work together.

What's next for EZML

There is a lot of functionality we would like to add to EZML. Firstly, allow the user to tweak more factors about the learning methods, which currently run very generically. Secondly, adding more methods, like a CNN. Third, allowing the user to visualize the results of their models, a key part of machine learning research. We plan to have some instructional text in the application by the time we submit this, as we imagine EZML could be used as a learning tool, but we haven't quite gotten there yet.

Built With

• java
• python
• scenebuilder
• scikit-learn
• tensorflow/keras