In this repo I have uploaded the code and datasets for non MNIST data and nonMNIST data.
I hace tried to implemented the problem with different ways
- implementing neural networks using low level API’s in tensor flow.
- implementing the same neural network using tf.estimator
- this gives you more modularity as you can specify the basic things like number of each type of layers and their specifics, evaluation metrics with much more ease and this helps to understand the core concept.
- also used sklearn to implement the same problem