🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

Check out our latest work where we reconstruct large portions of the actual training data from trained neural networks. Joint work with gal vardi Niv Haim Project page: giladude1.github.io/reconstruction/ abs: arxiv.org/abs/2206.07758

New preprint with gal vardi, Peter Bartlett, Nati Srebro, and Wei Hu on the implicit bias of gradient descent/flow (GD/GF) in two-layer leaky ReLU networks when trained on high-dimensional data: arxiv.org/abs/2210.07082

Thrilled to share that our paper "𝘙𝘦𝘤𝘰𝘯𝘴𝘵𝘳𝘶𝘤𝘵𝘪𝘯𝘨 𝘛𝘳𝘢𝘪𝘯𝘪𝘯𝘨 𝘋𝘢𝘵𝘢 𝘧𝘳𝘰𝘮 𝘛𝘳𝘢𝘪𝘯𝘦𝘥 𝘕𝘦𝘶𝘳𝘢𝘭 𝘕𝘦𝘵𝘸𝘰𝘳𝘬𝘴" was accepted to #NeurIPS 2022 as an 𝗼𝗿𝗮𝗹 𝗽𝗿𝗲𝘀𝗲𝗻𝘁𝗮𝘁𝗶𝗼𝗻! w/ Gilad Yehudai gal vardi Webpage: giladude1.github.io/reconstruction 🧵1/n

Glad to share our new work: "Deconstructing Data Reconstruction: Multiclass, Weight Decay and General Losses" arxiv.org/abs/2307.01827 About trainset reconstruction in multiclass & regression tasks + perks (w/ Gon Buzaglo Gilad Yehudai gal vardi Yakir Oz Yaniv Nikankin M. Irani) 🧵1/5

An entire family murdered in cold blood. Kedem family: Father Jonathan, mother Tamar, 6-year-old girls Shachar and Arbel, and 4-year-old boy Omer. Look at their happy faces. Their love. All of them murdered by Palestinian terrorists at Nir-oz kibbutz. Just because they’re

Hamas terrorists massacred hundreds of young Israelis at a rave. Listen to a survivor of the massacre>>

Tonight is the 4th night our babies will spend in Gaza. We do not know if they were harmed, injured, or even if they are still alive. This is the time to support Israel. Do everything you can to save our children, elders, men, and women from Hamas, a terrorist ISIS-like

New preprint with gal vardi: arxiv.org/abs/2410.01774 We prove task- & sample-complexity bounds for in-context learning in transformers for classification tasks, focusing on a 1-layer linear attention architecture. We find they can exhibit benign overfitting in-context!

Optimization induces implicit bias. We study general steepest descent in homogeneous nets & show (generalized) convergence to a (generalized) KKT pt. Adam presents a curious case between l2 & l1: arxiv.org/abs/2410.22069 With Nikos Tsilivis & gal vardi NYU Center for Data Science AI at Meta

Happy to share our work on: "A Theory of Learning with Autoregressive Chain of Thought" arxiv.org/abs/2503.07932 Joint work with great collaborators: gal vardi, Adam Block, Surbhi Goel, Zhiyuan Li , Theodor Misiakiewicz, Nati Srebro