Deep Learning

• Extension of the Class classification model to Nonlinear:

  • Repeatedly applying Nonlinear transformations to gradually represent complex functions.
  • = Deep Learning
  • Deep learning is seen as one of the means of nonlinear transformation.

• When I first learned about this, I was creeped out by how it could perform such complex tasks.

• However, after relearning it in Tokyo University 1S Information Alpha, I started to feel like “yeah, it makes sense”.

  • I feel like there has been a change in my understanding, where I now have the sense that “if a model has a lot of parameters, it can represent complex things”.
    • The feeling that with a model (such as a Neural Network) that has a huge number of parameters, you can do anything.

• The discussion in Tokyo University 1S Information Alpha about separating data, model, and parameter learning is also useful when explaining to others.

  • It also applies to the relationship between Linear Regression and deep learning.