Two Sigma’s Post

Our Founding Director sat down with Colin McEnroe at @ctpublic to discuss how algorithms shape our world for better and for worse. We hear about algorithms all the time, but what is an algorithm exactly? This hour, learn about how algorithms work, a bit about their history, and how they're impacting our culture. Watch the whole segment here: https://lnkd.in/gB5J-8CK

If you are attending NeurIPS 2023 in New Orleans this week, please check out our work: 1. "Every Parameter Matters: Ensuring the Convergence of Federated Learning with Dynamic Heterogeneous Models Reduction". https://lnkd.in/gisWzG6y 2. "A Unified Algorithm Framework for Unsupervised Discovery of Skills based on Determinantal Point Process". https://lnkd.in/gtZfyD_U

NeurIPS is one of the world's top machine learning conferences, and undergraduate acceptances are rare. But powered by colleagues, mentors, and their own drive to discover, Federico Cassano, Noah Shinn, and Neel Sortur made it anyway. Read more: https://lnkd.in/gbfU3rj3 Federico Cassano Karthik Narasimhan Ashwin Gopinath Neel Sortur Noah Shinn Shunyu Yao Robin Walters Linfeng Zhao NeurIPS

𝐍𝐞𝐰 𝐜𝐚𝐬𝐞 𝐬𝐭𝐮𝐝𝐲 𝐚𝐛𝐨𝐮𝐭 "𝐀𝐧𝐨𝐦𝐚𝐥𝐲 𝐃𝐞𝐭𝐞𝐜𝐭𝐢𝐨𝐧 𝐢𝐧 𝐓𝐫𝐚𝐜𝐤 𝐒𝐜𝐞𝐧𝐞𝐬" 𝐩𝐮𝐛𝐥𝐢𝐬𝐡𝐞𝐝! For the sector initiative Digitale Schiene Deutschland, our client Deutsche Bahn asked us to develop a machine learning solution that automatically detects anomalous/hazardous objects on train tracks. We really enjoyed working on this interesting machine learning problem and ended up using MonoViT (a self-supervised learning model based on vision transformer) and the segment anything model. If you are interested in learning more about the problem and our solution, read our full case study here: https://lnkd.in/d3yhs3tE

My new goto model can't wait for the finetunes. before this I was using wizardlm I'm not convinced llama 3 is the best for me especially for processing arxiv papers I might pivot to llama 3 after reducing papers down to necessary components / manageable size, but for now I only rely on models finetuned to handle 32k plus

Amplifying this wonderful new publication from the National Institute of Standards and Technology (NIST)!  Adversarial Machine Learning: A Taxonomy and Terminology of Attacks and Mitigations https://lnkd.in/e-N6qjTg

What is Multimodal RAG? Jacob Marks from Voxel51 explains in this clip from December's Union Fireside chat with Sage Elliott. Listen to the whole episode "How LLMs Are Transforming Computer Vision" to learn more! 🔗 https://lnkd.in/gKhiVTyt

I am happy to share that my paper, titled "Evolution of Compilers: Tracing the History from Early Autoprogramming Systems to Modern Optimizations," has been accepted for presentation at the 4th International Conference on Intelligent Systems and Machine Learning (ICISML - 2024)

Last week I was grateful to be invited to present our current working paper “Causal Inference with Cocycles” at the Munich Center for Machine Learning’s (MCML) 3rd annual workshop on causal machine learning. Causal machine learning goes beyond predicting outcomes based on patterns, and aims to identify actual causes behind those outcomes. 🧐In our work, we look at what most state of the art methods have in common, which is the need to (often) estimate entire distributions/densities from data. 🧠We show that in many cases, one can avoid this by instead just estimating a “cocycle”, which describes how a conditional distribution changes under causal actions (eg interventions). 🔥In various situations, directly targeting the cocycles simplifies the estimation problem and improves robustness to misspecification. 💪🏼This means more reliable, robust answers to causal “what if” questions in practice! Anyone interested can find out more in our paper (see the link in the comments)! Thank you to Stefan Feuerriegel for the invitation and for organising a great selection of speakers covering causal discovery and estimation problems! It’s always inspiring for the more theoretical end of my work to learn about the applied problems researchers in academia and industry are grappling with!

The Heart Disease Prediction Hackathon from Antigranular has finished. The challenge was to to utilise machine learning techniques to analyse the Heart Failure Prediction Dataset, containing synthetic data, and predict heart conditions with a high level of accuracy while maintaining patient privacy (yep using Differential Privacy 🔒). Check out my solution using Opacus, a PyTorch library to train models with Differential Privacy. https://lnkd.in/dujWj5Y6