-
-
Nifty logo.
Inspiration
I've realized many times while building projects with machine learning models: even though they're so easy to visualize, they're also very hard to put into code. I felt obligated to make building neural networks as easy as visualizing them.
What it does
ezml-dense is a web application that offers a graphical user interface which allows for the building of simple neural networks that can be copied and pasted into your program. Here’s how it works.
How I built it
The web application is entirely frontend, built in React.js. Its simple yet powerful UI handles all that you would need for building simple neural networks through Javascript.
Challenges I ran into
This is a frontend-intensive project, and I'm not a frontend person. On top of that, I used React.js, which I haven't used in any of my previous projects. Thus, I had reasonable amounts of pain as I tore my hair out at 3am wondering why CSS isn't working as I expect it to.
Accomplishments that I'm proud of
I'm proud of my beautiful reactive UI. I don't think I've created such a frontend masterpiece before. I guess that's just indicative of my learning.
What I learned
I learned a heaping ton of frontend development, and this time, using a Javascript framework instead of plain HTML/CSS/JS.
What's next for ezml-dense
There are infinite possibilities with the direction that ezml-dense can take. I can implement other types of machine learning models, so that I won't have to call it ezml-dense anymore to specify that this is for dense feedforward models. I can go in depth with the features in this version as well. Perhaps I can allow users to tamper with settings for each node, or allow users to train and test the model in the browser.
