https://github.com/eslamdyab21/nn-from-scratch

Building Neuron networks from scratch to better understand the ready made libraries and frameworks.
https://github.com/eslamdyab21/nn-from-scratch

Last synced: 16 days ago
JSON representation

Building Neuron networks from scratch to better understand the ready made libraries and frameworks.

• Host: GitHub
• URL: https://github.com/eslamdyab21/nn-from-scratch
• Owner: eslamdyab21
• Created: 2022-11-07T14:08:29.000Z (about 2 years ago)
• Default Branch: main
• Last Pushed: 2022-12-05T01:23:52.000Z (about 2 years ago)
• Last Synced: 2023-03-04T18:59:03.706Z (almost 2 years ago)
• Language: Jupyter Notebook
• Size: 3.91 MB
• Stars: 2
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# NN-from-scratch

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.