This is my first time training a neural net on a classification task and speech recognition inspired me to do so, should be a great learning experience.

Run the following code in terminal. This will take a couple hours to train fully. If you train it from scratch, if you clone it to your computer and run it as is it will load the saved model of the weights and bias and spit out the accuracy score.

python run_model.py

After 3000 epochs the model has about 93% accuracy. It might get better performance if I keep training it but 3000 was around 3 hours of training time and I was happy with the performance.

Credit to Siraj. The speech_data.py file is all his doing.