Page Not Found
Page not found. Your pixels are in another canvas.
Sitemap
A list of all the posts and pages found on the site. For you robots out there is an XML version available for digesting as well.
Pages
About Me:
About me
Page not in menu
This is a page not in th emain menu
Posts
Late Trains and Laughter: The Hackathon Journey Chronicles
Published:
Aditya, Anupam, Ayushi and I had to travel to Madras for Smart India Hackathon. We didn’t have direct train but had to switch trains at Nagpur. We boarded our train to Nagpur at the night slept thinking that next day by early morning we would be there and would then board our train to Madras. But when we woke up the next day we got the shock our life! Our train had apparently derailed during the night and was delayed a lot. Surprisingly I slept like a baby and did not feel a thing, but Aditya wasn’t so lucky, felt a huge jerk. With the current delay we would not be able to catch our train to Madras! We started looking for alternatives and stupidly enough tried to look for uber cabs from the next station😂. Being in a very rural area at that time there was no chance to get a cab. Finally we approached the TC of our train and explained him our situation. He told us get down this train at the next station and board the train that was coming behind it. He assured that the other train would get more priority on the track and had more chance to reach Nagpur in time for us to catch the next train. We took the TC’s advice and got into the other train. After an hour of nervous waiting we were finally at the Nagpur station and luckily enough our train to Madras was just entering the station on the adjacent platform. Thanks to the TC’s quick thinking, we made it to Nagpur just in time to catch our train to Madras, turning a potential disaster into a memorable adventure!
The Great Hostel Bread-Pizza Experiment
Published:
Having recently graduated, I’ve decided to mix things up by sharing some entertaining anecdotes from my college days every now and then. These stories will bring a refreshing change from my usual posts on tech and research papers. So here is first one….
Three of us, Utkarsh, Aditya, and I wanted to cook something in the hostel. We wanted to cook something that wouldn’t make us feel like MasterChef contestants or instant noodle experts! So we settled for bread pizza. More than making the pizza, the difficult part was convincing our other friends to help us and assure them that it would taste good. Since the plan was finalized at the last moment(as with any plans😂), there was no time to order stuff, so the three of us borrowed some cycles and set out to buy them from the market outside our campus. Since the then campus(temporary campus) was in Sejbahar(rural outskirts of Raipur), we were a bit skeptical that we would get all the ingredients we needed. After cycling and foraging through some shops, we finally gathered all the required ingredients. It was already late, and we were starting to get calls from our hungry friends, so we rushed back to our beloved hostel(Ena). After reaching, we cut up all the veggies with the help of our friends. With our engineering minds at work, we devised an assembly line to transform bread slices into pizza masterpieces, and soon, our pizza factory was in full swing! And just like that, our makeshift pizza factory turned a hungry hostel night into a deliciously unforgettable memory!
Future Blog Post
Published:
Will post once I have something interesting!
Vision GNN: An Image is Worth Graph of Nodes
Published:
The paper proposes to represent the image as a graph structure and introduce a new Vision GNN (ViG) architecture to extract graph-level feature for visual tasks.
Beyond Real-world Benchmark Datasets: An Empirical Study of Node Classification with GNNs
Published:
The paper discusses the results of conducting extensive experiments with a synthetic graph generator that can generate graphs having controlled characteristics for fine-grained analysis for node classification tasks.
Feature Overcorrelation in Deep Graph Neural Networks: A New Perspective
Published:
The paper is a research paper that proposes a new perspective to look at the performance degradation of deep graph neural networks (GNNs), which is feature overcorrelation.
FedGNN: Federated Graph Neural Network for Privacy-Preserving Recommendation
Published:
The paper proposes a federated framework for privacy-preserving GNN-based recommendation which can train GNNs in a decentralized manner on user data.
DistDGL: Distributed Graph Neural Network Training for Billion-Scale Graphs
Published:
The paper discusses design and development of a distributed graph neural network training framework based on existing Deep Graph Library.
Spelling Corrector using n-gram language model
Published:
There have been a lot of advancements in Natural Language Processing with advent of Large Language Models. Today powerful tools like ChatGPT can do complicated tasks like writing an email or a poem. It might be difficult to make such models all on your own but here is a small project in which you can make your own small language model and do a simple task spelling correction using the beautiful concept of probabilities.
Bayesian Networks in Python
Published:
Bayesian Networks are a really interesting tool when it comes to conditional probabilities. When you have set of dependent events you can form a network/graph of the dependencies. Now you only need the absolute probabilities for some root events and the conditional probabilities for directly linked events and you can start computing the probabilities for certain events in the network with complex dependencies. Here is a tutorial showing you how can you construct a bayesian network in Python on a library called pomegranate and start computing of different linked events.
Question-Answering using Basic NLP
Published:
The chatbots these days are very advanced and can answer complex queries but here I have some basic NLP concepts and made simple Question Answering System which answer questions like who, where, what type question. The idea is that typically for such questions the answer is a single word and is a noun. So first take our text then using coreference we replace all prnouns with appropriate nouns. Now in the transformed text we look for the sentence which is most similar to our query. To do this we do part of speech tagging on the sentence and extract all the nouns and verbs from the sentence. Now the relevant sentence will have most nouns and verbs in common with query. We are focusing on nouns and verbs because other parts of speech just make the sentence more descriptive to humans but the main content is added by nouns and verbs. Once we obtain the relevant sentence we just output the nouns/proper nouns in the sentence as our answers. Here is a tutorial to make such a system in Python.
MNIST scaled up dataset
Published:
An year back I was working on a project of making a memory efficient CNN training framework. We needed a dataset to train on large convolutional network. I scaled the standard MNIST digit dataset from 28x28 images to 224x224 and stored them in the raw binary format. I later thought someone might find it useful so I posted it on Kaggle. Now, today when I checked people had downloaded the dataset over the year. Finding that your work is useful to someone is really satisfying and I felt really excited. I look forward to make more such contributions in the future. Here is the link to the dataset in case any one needs it.
paper-reviews
DistDGL: Distributed Graph Neural Network Training for Billion-Scale Graphs
Published:
The paper discusses design and development of a distributed graph neural network training framework based on existing Deep Graph Library.
FedGNN: Federated Graph Neural Network for Privacy-Preserving Recommendation
Published:
The paper proposes a federated framework for privacy-preserving GNN-based recommendation which can train GNNs in a decentralized manner on user data.
Feature Overcorrelation in Deep Graph Neural Networks: A New Perspective
Published:
The paper is a research paper that proposes a new perspective to look at the performance degradation of deep graph neural networks (GNNs), which is feature overcorrelation.
Beyond Real-world Benchmark Datasets: An Empirical Study of Node Classification with GNNs
Published:
The paper discusses the results of conducting extensive experiments with a synthetic graph generator that can generate graphs having controlled characteristics for fine-grained analysis for node classification tasks.
Vision GNN: An Image is Worth Graph of Nodes
Published:
The paper proposes to represent the image as a graph structure and introduce a new Vision GNN (ViG) architecture to extract graph-level feature for visual tasks.
research-projects
Memory Efficient GPU based CNN training framework
October 2021-January 2022 Mentor: Dr. Vishwesh Jatala
Designing a memory efficient CNN training framework to train large CNNs on a single GPU
Design Space Exploration for URL Shortener
January 2022-May 2022 Mentor: Dr. Gagan Raj Gupta
Analyzing and comparing the scalability and efficiency of different URL shortener designs
Exploring the LED Cipher
October 2022-December 2022 Mentor: Dr. Dhiman Saha
Understanding and cryptanalysing the LED cipher
Distributed Graph Neural Networks
January 2023-July 2023 Mentor: Dr. Gagan Raj Gupta and Dr. Vishwesh Jatala
Scalable and distributed efficient training of GNNs
Recommended citation: D. Deshmukh, G. R. Gupta, M. Chawla, V. Jatala and A. Haldar, "Entropy Aware Training for Fast and Accurate Distributed GNN," 2023 IEEE International Conference on Data Mining (ICDM), Shanghai, China, 2023, pp. 986-991 https://arxiv.org/abs/2311.02399
talks
Talk 1 on Relevant Topic in Your Field
Published:
This is a description of your talk, which is a markdown files that can be all markdown-ified like any other post. Yay markdown!
Conference Proceeding talk 3 on Relevant Topic in Your Field
Published:
This is a description of your conference proceedings talk, note the different field in type. You can put anything in this field.
teaching
Associate Mentor
Feelance tutoring, IB Hubs, 2021
Clarified 200+ students queries related to Python programming and Web Development
Teaching Assistant in Discrete Maths
Undergraduate course, IIT Bhilai, Department of Electrical Engineering and Computer Science, 2023
Conducted tutorial sessions to clear students doubts. Designed assignments to aid students learning. Conducted and graded quizzes.