Building Neuron networks from scratch to better understand the ready made libraries and frameworks.
In this repo, a model is created using only NumPy, i.e forward and backward propagation, loss function, and the mechanics of the layers with weights are implemented from scratch. Starting from P1 (basic neuron) to P6 and NN (the coplete class/ish)
NN.py contains the implemented classes and is used in the NN from scratch - mnist digits2.ipynb notebook in which a model for digits classification is built and trained.
Then some models are made using tensoreflow in Tensoreflow folder.