https://github.com/mkashirin/scratches

Scratches is a project, which provides a comprehensive guide to creating deep learning models from scratch using Python and NumPy.
https://github.com/mkashirin/scratches

artificial-intelligence artificial-neural-networks deep-learning deep-neural-networks from-scratch machine-learning machine-learning-algorithms numpy python

Last synced: about 2 months ago
JSON representation

Scratches is a project, which provides a comprehensive guide to creating deep learning models from scratch using Python and NumPy.

• Host: GitHub
• URL: https://github.com/mkashirin/scratches
• Owner: mkashirin
• License: mit
• Created: 2024-02-19T16:15:14.000Z (11 months ago)
• Default Branch: main
• Last Pushed: 2024-06-07T17:45:01.000Z (7 months ago)
• Last Synced: 2024-06-07T19:04:03.507Z (7 months ago)
• Topics: artificial-intelligence, artificial-neural-networks, deep-learning, deep-neural-networks, from-scratch, machine-learning, machine-learning-algorithms, numpy, python
• Language: Python
• Homepage:
• Size: 15.3 MB
• Stars: 2
• Watchers: 2
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
Scratches




Scratches is a project that is inspired by the "Deep Learning from Scratch"

book by Seth Weidman, which provides a comprehensive guide to creating deep

learning models from scratch using Python and NumPy. The project aims to provide

pure Python and NumPy implementations of classic machine learning algorithms

such as k-nearest neighbors, linear and multiple regressions, and elementary and

convolutional neural networks.



## Requirements

The only system requirement for this application is that you use Conda or

Miniconda to manage your Python packages.

## Installation and usage

Use the Git command-line interface (CLI) to clone this repository into your

working directory using the following command:

```bash

git clone https://github.com/mkashirin/scratches

```

To create a virtual environment, please follow the lines below:

```bash

conda init

conda env create --file="environment.yml" --name="scratches"

conda activate scratches

```

Although NumPy is a crucial dependency for the functioning of the algorithms,

Jupiter, Matplotlib, and Pandas are also present in the environment in order to

provide a seamless experience.

If you wish to change the default path for your environment, you can edit the

"prefix" value in the "environment.yml" file (the default location is

"~/anaconda3/envs/scratches").

After that You can just run the Jupyter sever to access the notebooks from the

**examples** directory by executing the following command:

```bash

jupyter lab

```

And that's it. You are all set!

## Suggestions

The only specific suggestion is to not use it outside the educational context.

If you are still unsure, do not worry. The documentation in the source code can

be considered sufficient. The code has been written in a clear and concise

manner, focusing on readability rather than efficiency.

So, feel free to experiment with machine learning models! Combine various

structures to create your own neural networks. Explore the code to gain a deeper

understanding of fundamental ML and AI principles.

## Licencing

This project is distributed under the MIT open source licence.