Concise Notes for Learning Keras as Quickly as Possible

View these in the order provided; don’t try to memorize everything:

https://keras.io/#installation (You may choose to install TensorFlow now if you haven’t already.)

https://keras.io/getting-started/functional-api-guide/ (Stop reading after the model diagram.)

https://keras.io/getting-started/sequential-model-guide/ (Stop reading once you get to “Examples”.)

About callbacks
https://machinelearningmastery.com/how-to-stop-training-deep-neural-networks-at-the-right-time-using-early-stopping/

You can use the functional API for unique architechures
https://machinelearningmastery.com/keras-functional-api-deep-learning/

At this point you should have a good understanding. Google things and refer to the documentation as needed.

Keras in R

I prefer using Keras in Python, but you can also run it from R.
The R Keras installation can be a little unfriendly, so here are some tips:

To save time, avoid issues by updating these packages first.
install.packages(c("ps", "Rcpp", "digest", "processx", "devtools"))

Install TensorFlow.
devtools::install_github("rstudio/tensorflow")
tensorflow::install_tensorflow() Include the argument gpu=TRUE if you want GPU processing.

Verify TF installation.
library(tensorflow)
tensorflow::tf_config()

Install Keras.
devtools::install_github("rstudio/keras")
If you update other packages when prompted and one of them fails, perform install.packages('package_name') separately, then run devtools::install_github("rstudio/keras") again.

Allow installation of Miniconda unless you insist otherwise.

If you are familiar with R, the Keras usage will be easy to understand: https://keras.rstudio.com/