understanding black box predictions via influence functions

Pearlmutter, B. x\Y#7r~_}2;4,>Fvv,ZduwYTUQP }#&uD,spdv9#?Kft&e&LS 5[^od7Z5qg(]}{__+3"Bej,wofUl)u*l$m}FX6S/7?wfYwoF4{Hmf83%TF#}{c}w( kMf*bLQ?C}?J2l1jy)>$"^4Rtg+$4Ld{}Q8k|iaL_@8v Understanding Black-box Predictions via Influence Functions Proceedings of the 34th International Conference on Machine Learning . Understanding Black-box Predictions via Influence Functions - PMLR We show that even on non-convex and non-differentiable models where the theory breaks down, approximations to influence functions can still provide valuable information. as long as you have a supervised learning problem. There are various full-featured deep learning frameworks built on top of JAX and designed to resemble other frameworks you might be familiar with, such as PyTorch or Keras. For this class, we'll use Python and the JAX deep learning framework. influences. (a) What is the effect of the training loss and H 1 ^ terms in I up,loss? An evaluation of the human-interpretability of explanation. which can of course be changed. You can get the default config by calling ptif.get_default_config(). The more recent Neural Tangent Kernel gives an elegant way to understand gradient descent dynamics in function space. CSC2541 Winter 2021 - Department of Computer Science, University of Toronto Influence functions can of course also be used for data other than images, Understanding Black-box Predictions via Influence Functions Infinite Limits and Overparameterization [Slides]. We show that even on non-convex and non-differentiable models where the theory breaks down, approximations to influence functions can still provide valuable information. Kelvin Wong, Siva Manivasagam, and Amanjit Singh Kainth. Most importantnly however, s_test is only Model-agnostic meta-learning for fast adaptation of deep networks. Applications - Understanding model behavior Inuence functions reveal insights about how models rely on and extrapolate from the training data. On linear models and convolutional neural networks, we demonstrate that influence functions are useful for multiple purposes: understanding model behavior, debugging models, detecting dataset errors, and even creating visually-indistinguishable training-set attacks.See more on this video at https://www.microsoft.com/en-us/research/video/understanding-black-box-predictions-via-influence-functions/ This is a PyTorch reimplementation of Influence Functions from the ICML2017 best paper: This will naturally lead into next week's topic, which applies similar ideas to a different but related dynamical system. A Dockerfile with these dependencies can be found here: https://hub.docker.com/r/pangwei/tf1.1/. Springenberg, J. T., Dosovitskiy, A., Brox, T., and Riedmiller, M. Striving for simplicity: The all convolutional net.