Anindyadeep Sannigrahi

MVGAEs

Creating new potential state of the architecture for link prediction on citation networks. Here I created a new architecture that is surpassing the present state of the art results in link predictions in benchmark citation graph dataset like Cora, Citeseer, Pubmed. The model is computationally more efficient and also giving an AUC scores of 0.99, 0.98 and 0.99 for the respective benchmark datasets. (paper COMING SOON)

Efficiently balanced unbalanced dataset using collagination and augmentation methods. Created simple Conv Nets and compared different Convolutional Neural Network architecture in terms of, performance of the model and overall computational efficiency for classification of skin cancer in skin lesion images.

YogaPosGNN

Classifying Real time yoga positions using Graph Neural Networks. It is using PyTorch Geometric and PyTorch in the backend. Using GNNs, it is a graph classification problem, and this hugely decreased the training time, as well as has very low latency in delivering the evaluations in real time.

• PyTorch
• PyTorch Geometric
• Mediapipe
• OpenCV

SignLangGNN

Another real time graph classification based task, where I tried to classify the sign language from the hand signs using the Hand graphs, generated from Mediapipe.

• PyTorch
• PyTorch Geometric
• Mediapipe
• OpenCV
• NumPy

BetaLactamase

Predicting the p-chembl value from the chemical structures using RdKit and PyTorch Geometric. I framed this problem as a graph regression problem. I used Graph attention based models to predict the p-chembl values a given chemical structure which is a very important factor of drug engineering.

• PyTorch
• PyTorch Geometric
• RdKit
• Scipy, NumPy

WhatsappInspect

This is a library in which users can use the exportable whatsapp chats data and can utitlize them with this library to get more tabular and graphical insights. For more checkout Github and PyPI

• NumPy
• Pandas
• Plotly
• Emoji, Demoji
• WordCloud

MLfromScratch

This project tries to create most of the used machine learning algorithms by only using numpy. The aim of this project is to understand ML algorithms in its depth. A workable library is coming soon.

• NumPy
• MatPlotlib
• Pandas

AbstractTorch

A small library in making, which will provide a feel of easy model making like keras, in pytorch. Also this library will make life easier using some cool functions handelling data handelling and easy model making.

• PyTorch
• NumPy
• Sklearn