April 14

First : discuss end of semester projects. What are your plans?

Second : questions about the material we discussed last week, the neural nets chapter in "From Scratch".

Next : continuing down the "deep learning" rabbit hole ...

I have installed tensorflow and keras on jupyter.marlboro.college ... these look like like the best options for exploring this material in code , and is what the text auther uses in the next chapter.

My plan is to continue on in "from scratch" to the deep learning chapter ... which is more abstract and setting the stage for how libraries like tensorflow model this sort of problem.

In class :

• talk through neural nets and deep learning online book. (handwriting)
• describe what a "tensor" is and why we would want them.
• look at basic classification with keras/tensorflow

vocabulary :

• tensor : N-dimensional matrix ; for N=3 a list of list of lists.
• layers : neural net notion of collection of neurons ... generalized.
• softmax (mathy)
• cross-entropy (more mathy)
• loss function : error between model prediction and training data
• optimizer : back propagate changes to model
• metrics : print or plot training convergence
• CPU vs GPU : how fast will all this run ?
• MNIST handwritten digits database