Skip to content

Bishwasz/Neural_Network_From_Scratch

Repository files navigation

Neural_Network_From_Scratch

Used numpy to create Neural network to identify MNIST handwritted digits.

Here are some trials:

(Put it on light Mode to see the labels)

Configuration for the Below Trial trials:
Hidden Layers:1
Images Trained: 5000
Nodes in Hidden Layer: 100
Learn Rate: 0.01
Activation Function: Relu+SoftMax
Loss function: Cross Entropy
Batch Size:50
Alt text

Configuration for the Below Trial trials:
Hidden Layers:1
Images Trained:5000
Nodes in Hidden Layer: 100
Learn Rate: 0.04
Activation Function: Relu+SoftMax
Loss function: Cross Entropy
Batch Size:50
Alt text

Here are some resources that helped me.
https://www.youtube.com/watch?v=sIX_9n-1UbM
https://www.youtube.com/watch?v=hfMk-kjRv4c&t=629s

About

creating NN from scratch to identify MNIST

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages