<< Fast exact multiplication by the hessian. While influence estimates align well with leave-one-out. The datasets for the experiments can also be found at the Codalab link. With the rapid adoption of machine learning systems in sensitive applications, there is an increasing need to make black-box models explainable. 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. Are you sure you want to create this branch? Wei, B., Hu, Y., and Fung, W. Generalized leverage and its applications. kept in RAM than calculating them on-the-fly. Some of the ideas have been established decades ago (and perhaps forgotten by much of the community), and others are just beginning to be understood today. 7 1 . On Second-Order Group Influence Functions for Black-Box Predictions Some JAX code examples for algorithms covered in this course will be available here. An empirical model of large-batch training. Gradient-based Hyperparameter Optimization through Reversible Learning. On the origin of implicit regularization in stochastic gradient descent. Jianxin Ma, Peng Cui, Kun Kuang, Xin Wang, and Wenwu Zhu. Understanding black-box predictions via influence functions. Differentiable Games (Lecture by Guodong Zhang) [Slides]. Understanding Black-box Predictions via Influence Functions Proceedings of the 34th International Conference on Machine Learning . Understanding Black-box Predictions via Influence Functions (2017) Negative momentum for improved game dynamics. A spherical analysis of Adam with batch normalization. In. Lectures will be delivered synchronously via Zoom, and recorded for asynchronous viewing by enrolled students. More details can be found in the project handout. For the final project, you will carry out a small research project relating to the course content. Huang, L., Joseph, A. D., Nelson, B., Rubinstein, B. I., and Tygar, J. Adversarial machine learning. Requirements Installation Usage Background and Documentation config Misc parameters How can we explain the predictions of a black-box model? Understanding Black-box Predictions via Influence Functions A tag already exists with the provided branch name. >> Uses cases Roadmap 2 Reviving an "old technique" from Robust statistics: Influence function We use cookies to ensure that we give you the best experience on our website. For toy functions and simple architectures (e.g. Datta, A., Sen, S., and Zick, Y. Algorithmic transparency via quantitative input influence: Theory and experiments with learning systems. Szegedy, C., Vanhoucke, V., Ioffe, S., Shlens, J., and Wojna, Z. and even creating visually-indistinguishable training-set attacks. Li, B., Wang, Y., Singh, A., and Vorobeychik, Y. outcome. Subsequently, One would have expected this success to require overcoming significant obstacles that had been theorized to exist. on the final predictions is straight forward. Insights from a noisy quadratic model. 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. How can we explain the predictions of a black-box model? Fast convergence of natural gradient descent for overparameterized neural networks. Or we might just train a flexible architecture on lots of data and find that it has surprising reasoning abilities, as happened with GPT3. A. the training dataset were the most helpful, whereas the Harmful images were the In this paper, we use influence functions --- a classic technique from robust statistics --- to trace a model's prediction through the learning algorithm and back to its training data, thereby identifying training points most responsible for a given prediction.

Chi E La Moglie Di Max Mariola, Park Models For Sale At La Hacienda Rv Resort, Articles U