Neural Network

I was finding that the number of points to train the Neural Network on the Fuchs Data was not that large.

For example, it seems that after only around 2,000 points, we find the percent error on the noiseless data drops to approximately 0. I will run the SVR on the same training splits and compare this.

Things to Do

• Learn how to implement an Invertible Neural Network
• Read up more about Fusion Physics