math-behind-ai / scratchai Goto Github PK

Name
K-Means

Packages used
pandas - 1.3.4
numpy - 1.21.4

Brief explanation
Unsupervised learning algorithm to group data points into k different clusters.

Input

Set of data points

Working

1. Initialize cluster centers
2. Until the cluster centers converge:
   Assign data points to nearest cluster center
   Update cluster centers

Output

Number of clusters and cluster label for each data point.

Name
Simple Linear Regression

Packages used
Pandas : 1.2.4
Numpy : 1.19.5

Brief explanation
Creating a Simple Linear Regression model, that could have the best fit line among the given dataset and thus make predictions

Some of the most common DL Algorithms are listed below. Feel free to suggest other algorithms, not on the list. and we'll update it.

Name

• Artificial Neural Network
• Adaline Neural Network
• Multilayer Perceptron
• Convolutional neural networks
• Cross Entropy Loss
• Stochastic gradient descent
• Adam Optimizer
• RMS Prop
• Recurrent Neural Networks
• Vision Transformers
• Transformers
• Word embeddings
• LSTM

Packages used
List all the packages and versions used.

Brief explanation
Provide a neat README file in the directory. There are two existing algorithms in the repository, please follow a similar folder structure and document your code with proper markdown syntax/comments.

Is your feature request related to a problem? Please describe.
Using libraries, defeats the main goal of this repository, that is creating everything from scratch. Using sklearn scoring model, defeats the purpose of this repo

Describe the solution you'd like
I'd like to create a function for calculating the R2score as well as accuracy, in Decision Tree

Additional context
Please assign me this issue under hacktoberfest

Is your feature request related to a problem? Please describe.
It is not actually a problem but just a suggestion. Currently, we use the common functions from the sklearn library.

Describe the solution you'd like
What if we create a utils directory and store all the common loss functions, metrics, etc so that by calling the desired function we can use them?

Additional context
For example in the utils directory, it can have a file named loss functions and it can have functions such as cross-entropy loss, etc. And in future examples, instead of using the loss function from another library, we can directly use it which has already been implemented by the community.

Some of the most common ML Algorithms are listed below. Feel free to suggest other algorithms, not on the list. and we'll update it.

Name

• Multiple Linear Regression
• Lasso Regression
• Ridge Regression
• Elastic-Net Regression
• Polynomial Regression
• CatBoost Regression
• Naive Bayes
• K- nearest neighbors
• Support vector machines
• K-means
• K-medoids
• Decision Trees
• Linear Discriminant Analysis
• Principal Component Analysis

Brief explanation

• Provide a neat README file in the directory. There are two existing algorithms in the repository, please follow a similar folder structure and document your code with proper markdown syntax/comments.

Name
Word Embeddings (Word2Vec)

Packages used
tensorflow==2.5.0
numpy ==1.19.5
pandas==1.2.4

Brief explanation
Word embedding is used to extract semantic relations between different words by plotting them in a vector space.