Sarcasm Detection for social media comments
Tools : Python, PyTorch, Hugging Face, Jupyter
Tags : NLP, Deep Learning
Sarcasm detection is a difficult language problem for a simple reason - through definition, it is ambiguous. To make social media a safer space, it is imperative that true meaning is derived from sentences.
With this as motivation, I build a sarcasm detection model that is trained on data including comments and their replies on the Reddit Forum. I make use of a compact, quick and lightweight transformer model from Hugging Face - DistilBERT.
Check out the github for a tutorial style implementation of the same!
Spatio-temporal modeling of COVID-19 disease spread
Tools : R, Python, Pandas, ggplot2, PyTorch
Tags : Scripting, Deep Learning, Data Science
Modeling localized disease patterns using multiple data sources is a challenging problem for epidemiologists.
Further, capturing spatial relations and temporal dynamics using individual-level mobility data is also an important open problem.
We propose to tackle this problem by combining SafeGraph dataset for mobility and exposure as well as the New York City Health department’s report of weekly cases at the level of zip code to model zip code level COVID-19infection spread.
Finally, we aim to identify local hotspots for infections by directly accounting exposure of populations to POIs in our network-based metapopulation model.
DL for Scene Recognition
Tools : Python, PyTorch, Jupyter
Tags : Scripting, Deep Learning, Computer Vision
This project uses Deep CNNs for scene recognition (Ex : What does this image represent? - An image/office/kitchen?)
In essence, the project aims to solve the problem of effective scene recognition with the help of various techniques including data augmentation and operation optimization.
First, scene classification is done with the help of three models built with PyTorch. Later, multi-attribute recognition is done within scenes (Ex : Does this scene contain Nature/Humans/Clouds/Water bodies?)
DL for Semantic Segmentation
Tools : Python, PyTorch, Jupyter
Tags : Scripting, Deep Learning, Computer Vision
This project attempts to solve the problem of semantic segmentation (the process of labeling each object in an image) using Deep CNNs. Later, I observe the efficacy of transfer learning from the created model onto another data set.
To summarize - initially, I recreate this paper. The idea is to use 'dilation' (which increases the receptive power of a composite CNN) and then 'aggregate context' over different portions of the image using the 'Pyramid Pooling Module' (PPM). Keep an eye out for the blog, where I will be getting into the specifics!
Finally, I use the previous model titled 'PSPNet' and fine-tune pretrained weights on another semantic segmentation data set titled 'KITTI'.
Traveling Salesman Problem
Tools : Python, Bash
Tags : Algorithms, Scripting
"Given a list of cities and the distances between each pair of cities, what is the shortest possible route that visits each city exactly once and returns to the origin city?" - The Traveling Salesman Problem
This project focuses on a detailed analysis of algorithms used to solve the Traveling Salesman Problem as optimally as possible. Following are the techniques discussed and implemented :
Genetic Algorithm
Simulated Annealing
Approximation Algorithm using Minimum Spanning Tree heuristic
Branch and Bound
Pattern Recognition and Machine Learning
Tools : Python, cvxopt, cvxpy, Pandas, sklearn
Tags : Machine Learning
A repository of commonly used Machine Learning Algorithms developed from scratch are available here.
They include the following :
Recommender Systems vectorized and coded from scratch
SVMs using cvxpy and cvxopt
Clustering using Expectation Maximization algorithm
Trend analyses of S&P500 using a Hidden Markov Model
Perceptron for Classification coded from scratch with NumPy
Logistic Regression/Gradient descent coded from scratch with NumPy