Posts
-
NeurIPS 2024: Forgetting/Unlearning is hard!
Recently, I attended the NeurIPS 2024 conference and encountered many fascinating papers and talks. Among these discussions, a recurring theme emerged: machine unlearning, particularly in the context of defining and quantifying unlearning.
-
Hidden State Visualization of Llama and SmolLM Models
I was playing around with hidden states of the language model as I was trying to implement an unlearning/manipulation method. So I thought it would be nice to visualize the hidden states of the model for two samples to show how they learn to predict the classification (the next token). I used the Llama and SmolLM models for this experiment.
-
Introducing H3xplorer - A Hexagonal Map Visualization Tool
I am excited to launch H3xplorer, an online tool for creating and exploring hexagonal maps using Uber’s H3 system. Designed for data scientists, urban planners, and geospatial enthusiasts, H3xplorer transforms raw geographic data into hexagon grids, making it easier to visualize patterns, density, and flows.
-
Machine Unlearning: AI and the Art of Forgetting
I am writing on an open problem in artificial intelligence (AI) and machine learning (ML) called machine unlearning. The problem is about how to remove the effect of the data from the model and put it in a state as if the data never has been seen by the model. This is just a definition not a solution.
-
Teaching Programming Beyond Syntax: A Holistic Approach
In the bustling corridors of a university, I recently encountered a sobering truth that left me both saddened and reflective. A student, having just passed a programming course, was unable to articulate or effectively use the most basic programming concepts. This moment was a stark reminder of a significant gap in our educational approach to programming—a subject that should be as vibrant and dynamic as the technology it aims to master.
-
Right To Be Forgotten, Is It Possible?
I’m pursuing a master’s in machine learning. The problem at hand is that we want to remove facts that have been learned by machines. We don’t want to necessarily degrade the accuracy.
-
PI Day
Are you ready to cut into some $\pi$ (Pi) facts that are both tasty and educational? Let us dive into the world of Pi, which has no end.
-
Training data poisoning to get what you want in LLMs, A Question
There are numerous methods for jailbreaking Large Language Models (LLMs), such as ChatGPT, that one can find across the internet. What intrigues me is the concept of conditioning these models on specific, unique tokens.
-
Hey Julia
Hi, Julia is definitely worth considering if you’re looking for a new programming language to learn in 2022. It’s designed for high performance and has been gaining popularity lately as the future of machine learning.
-
Corona product in the networkx python library.
Hi, I just want to announce that my contribution will be released in networkX 3.0, which adds corona product. NetworkX is a Python package for the creation, manipulation, and study of the structure, dynamics, and functions of complex networks.
-
Paper-free semester
Hi, I am going to write about how to have a paper-free semester.
-
Bloom's Taxonomy for Mathematics
Hi, In this post, we are going to review a short but interesting description of Bloom’s Taxonomy in Mathematics.
-
Corona Product in Graph Theory
Hi, In this post, we are going to know about Corona Product in graph theory.
subscribe via RSS