We will be presenting our poster (No. 1625) on Thursday at 10:45 am.
If you're interested in memorization/privacy, please come by and chat with us!
#NeurIPS2023
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/ @GonBuzaglo@Giladude@galvardi@ozyakir@YNikankin M. Irani)
🧵1/5
Fascinating talk by @GaelVaroquaux, (Godfather of scikit-learn 😅) on model evaluation.
Closed some knowledge gaps I felt I had, especially in evaluating models on imbalanced data
What a great talk by @GaelVaroquaux about model evaluation, with actionable tips & suggestions.
Methods that truly give improvements survive and get used. Being good at evaluation allows you to be the first to know whether your method is a *real* winner.
youtube.com/watch?v=0qhWr2…
