Companion code for Forward Thinking: Building and Training Neural Networks One Layer at a Time and Forward Thinking: Building Deep Random Forests submitted to NIPS 2017.
Authors:
- Chris Hettinger
- Tanner Christensen
- Ben Ehlert
- Jeffrey Humpherys
- Tyler Jarvis
- David Kartchner
- Kevin Miller
- Sean Wade
- numpy==1.11.3
- tensorflow-gpu==1.0.0
- keras==2.0.4
- matplotlib==2.0.0
We used a single desktop computer with:
- Intel i5-7400 processor
- Nvidia GeForce GTX 1060 3GB GPU
- 8GB DDR4 RAM
With our configuration, it took approximately 2 hours to run the
run_mnist_cnn.shscript.
TODO: Ask Sean to make this pip installable. The ForwardThinking module is pip installable. Download and install this package by running
pip install forward_thinking
Once the package has been installed, you may run the included run_mnist_cnn.sh script.
This will run the forward thinking neural network (achieved 99.72% in our tests), the backpropagation equivalent of our model (achieved 99.63% in our tests), and saves and displays a plot comparing the test and train accuracies.