MNIST classification with a low parameter convolutional neural network based on inception modules.
The 'kernels' folder contains Kernels for Kaggle.
- MNIST Data Augmentation with Elastic Distortion: Overview/visualization of elastic distortion and select Keras data augmentation methods on MNIST.
mnist-inception-9970-test-acc-kernel.ipynb: Method for training a 76,264 parameter inception model that scored 99.7% test accuracy in the Kaggle Digit Recognizer competition.
mnist-inception.ipynb: Used to explore methods and generate results, which can be appended to a csv file.mnist-inception-results.ipynb: Used to explore results (which are loaded in from the csv file) and other related ideas.