Over the next couple of weeks, we'll look at neural nets & deep learning : discussion , readings, examples.

To understand this completely means wading through a lot of math.

Some of the tutorials will instead focus on how to use code libraries and tools - what the inputs are, how to call their API, and what you get. You'll need some intuition to do that, even without mastering the math.

- guru99 : deep learning for beginners
- Fizz Buzz in Tensor Flow (from "scratch" book author)
- neural nets and deep learning - online book with nice examples of handwriting analysis
- How to build your own Neural Network from scratch in Python - another reasonable example of the basic ideas
- wikipedia: artificial neural nets
- wikipedia: deep learning
- stanford deep learning tutorial
- tensorflow playground - tinker with a neural net in your browser
- keras tutorial - keras and tensorflow are deep learning python libraries

... and our "scratch" textbook, chapters 18 & 19.

My plan is to talk through the guru99 notes and the chap 18 textbook material on Friday, and then ask you to look at this yourself over the weekend. I suggest answering the following questions, running his code, and making some plots to see what is converging to what.

Here's what you should try to understand :

In the simple linear regression setup, the model was y=a*x+b. The target (goal) was to predict y, the inputs (features) was x, and the model parameters were (a,b). During "learning", we found a way to start at some (a0,b0) guess, and then change (a,b) until the least squares error was a minimum.

So, now we have a new model : neural nets. Pick an example from the readings.

- What are its features and target? How much data is this?
- What specifically are the model parameters? How many variables are there?
- What is the error function?
- What is the idea behind the algorithm to update the model parameters?

- Data Science: Reality Doesn't Meet Expectations | HN discussion
- Swift: Google’s bet on differentiable programming | HN discussion
- Covid vs. US Daily Average Cause of Death visualization

https://cs.marlboro.college /cours /spring2020 /data /notes /apr10

last modified Tue October 15 2024 6:33 am

last modified Tue October 15 2024 6:33 am