A python utility to measure the time and accuracy of neural networks trained on the MNIST dataset.
User can choose to meausure how various factors affect the training speed and training accuracy of a neural network and visually see the results in a matplotlib graph. These factors include:
- Number of neurons per hidden layer in each network
- Number of hidden layers in each networks
- Number of epochs trained for each network
Dependencies:
- Tensorflow (For GPU support use tensorflow-gpu instead and follow tensorflow's instructions on how to set up tensorflow to use a GPU as it can be quite involved depending on your operating system.)
- Matplotlib
How to use: Just run ANNProfiler.py and follow the prompts
If you didnt find it there already, check it out on github! https://github.com/AlexWorland/ANNProfiler