Handwriting Recognition in R with Keras/Tensorflow
Now we'll see how we can build neural networks programmatically in R.
If you don't feel like engaging with code at all, THAT'S OK! You can just watch on Zoom if you prefer. Really, only people who are interested need to click this link and get access to the code.
Here's the link to the resource:
https://rstudio.cloud/spaces/75500/join?access_code=BP9yh3wYnISb99JMRfiy4LR7g8elUAlfZRp4VXcj
Note: The resource will remain active for just three days. Unfortunately, after that time I will need to disable the resource due to the amount of CPU and RAM (and MONEY) it takes to run.
It's OK (in fact, it's expected) if you don't understand much of the code. Just try to get the big picture of what is going on. If you want to see how to engage with this code, you should take a course in Neural Networks -- I'm teaching one in Spring 2022 at NVU Johnson!
When you click the link, first you'll need to set up an account. After that you'll get to this screen. Click "Projects" along the top.
After that, click on the project "Artificial Neural Networks in R" from the project list (it's the only one).
It will take a few minutes for the project to open and get all the software installed on a cloud based computer for you.
After it's open, click on "01_Script.R" as shown below.
This will open a script file that has all the code needed to train a neural network.
THE ONLY THING you need to do is click the run button as circled below. I'll talk you through it as we click our way through the code one line at the time. I also wrote pretty detailed comments that might help you understand!
I'm always amazed how just a few lines of code can be used to build a mathematical function to recognized hand writing! :D