Love the course! Just thought I’d create a topic to discuss suggestions on making this course even easier/less frustrating for beginners based on your stated goals to make this accessible to all. Some suggestions I’d like to put out there (up for discussion, add your own as well:)
-I installed anaconda/keras/theano on my machine and ran it using CPU and the Pima Indians diabetes dataset (quick). This avoids AWS issues/complexity), waiting for approval, cost, etc. Downside: not as fun/interest as dogs/cats perhaps.
-Avoid using image prediction (convolution networks) as this has added complexity (in understanding how the model works) vs simpler models.
-Make Lesson 1 be an end to end solution ie load, fit, then predict and save the predictions to file (something like the redux notebook)
-Start with a deep learning neural network or regression spreadsheet instead of code/AWS, etc.
-A ‘run in browser’ coding environment similar to Dataquest.io (avoids any setup issues).
Open to any suggestions, discussion here or let me know if there’s another spot for these sorts of discussions
Thanks again for creating this course!