Jay Desai's repositories
EmotionRecognition
Recognizing human emotions based on the audio, video of the subject.
incubator-mxnet
Lightweight, Portable, Flexible Distributed/Mobile Deep Learning with Dynamic, Mutation-aware Dataflow Dep Scheduler; for Python, R, Julia, Scala, Go, Javascript and more
SBOD
Developed a system that makes use of ultrasonic sensor, camera and smartphone for detection, recognition and processing of objects that hinders the path of visually impaired person. Ultrasonic sensors detects and measures the distance of obstacle while image captured from camera is used for object recognition. The output is in the form of audio signals. Environment : Android Studio, Arduino ide, Java, xml, sketch.
AmazonReviews
Sentiment Analysis on Amazon Reviews
AprioriPyspark
Implementing Apriori using pyspark
bert_score_with_cache
BERT score for text generation bert_score_with_cache
BitcoinSentiment
Data source : coindesk, Twitter - twitter api for searching latest tweets on btc - positive or negative tweets vs btc price - CNN, LSTM used - Too many spam tweets made bad data and hence worst results Results : Bad data -> Bad model, needs many more features instead of just one.
bitsandbytes
support for 12.2
coinbasepro-python
The unofficial Python client for the Coinbase Pro API
contextualized-topic-models
A python package to run contextualized topic modeling. CTMs combine contextualized embeddings (e.g., BERT) with topic models to get coherent topics. Published at EACL and ACL 2021.
contrastive-predictive-coding-images
Keras implementation of Representation Learning with Contrastive Predictive Coding for images
CreditCardFraud
The datasets contained transactions made by credit cards in September 2013 by European cardholders. The dataset presented transactions that occurred in two days, where there were 492 frauds out of 284,807 transactions. The dataset was highly unbalanced, the positive class (frauds) account for 0.172% of all transactions. I used Logistic Regression and Random Forest to predict the class after performing oversampling of the minority class using SMOTE (Synthetic Minority Oversampling Technique), which resulted in the equal distribution of both the classes (0 and 1). Used the following modules in Python-Pandas, NumPy, Scikit-Learn, matplotlib, Imbalanced-learn Logistic Regression gave AUC=0.95, but the precision was very low (0.08). So, I tried Random Forests and it gave AUC=0.94 and precision=0.85 (with 'gini' criterion) and AUC=0.92 & precision=0.83 (with 'entropy' criterion). Measured performance metrics-accuracy, precision, recall, AUC and plotted ROC curve Used the following modules in Python-Pandas, NumPy, Scikit-Learn, matplotlib, Imbalanced-learn
detoxify
Trained models & code to predict toxic comments on all 3 Jigsaw Toxic Comment Challenges. Built using ⚡ Pytorch Lightning and 🤗 Transformers. For access to our API, please email us at contact@unitary.ai.
FashionMNIST
A CNN classifier for classifying Fashion MNIST database achieving over 98% accuracy.
lorentz-embeddings
Embed arbitrary graphs in Hyperbolic space
transformers
🤗 Transformers: State-of-the-art Machine Learning for Pytorch, TensorFlow, and JAX.