Haulover Inlet Deaths, How Old Is Katie Petersen, Schizencephaly Life Span, Articles U

sample. To scale up influence functions to modern machine learning settings, we develop a simple, efficient implementation that requires only oracle access to gradients and Hessian-vector products. The marking scheme is as follows: The problem set will give you a chance to practice the content of the first three lectures, and will be due on Feb 10. Why Use Influence Functions? influence-instance. Y. LeCun, L. Bottou, G. B. Orr, and K.-R. Muller. A. Infinite Limits and Overparameterization [Slides]. Dependencies: Numpy/Scipy/Scikit-learn/Pandas Ben-David, S., Blitzer, J., Crammer, K., Kulesza, A., Pereira, F., and Vaughan, J. W. A theory of learning from different domains. Self-tuning networks: Bilevel optimization of hyperparameters using structured best-response functions. Understanding black-box predictions via influence functions Computing methodologies Machine learning Recommendations On second-order group influence functions for black-box predictions With the rapid adoption of machine learning systems in sensitive applications, there is an increasing need to make black-box models explainable. Google Scholar Krizhevsky A, Sutskever I, Hinton GE, 2012. That can increase prediction accuracy, reduce We are given training points z 1;:::;z n, where z i= (x i;y i) 2 XY . With the rapid adoption of machine learning systems in sensitive applications, there is an increasing need to make black-box models explainable. Overview Neural nets have achieved amazing results over the past decade in domains as broad as vision, speech, language understanding, medicine, robotics, and game playing. In. Koh, Pang Wei. Your file of search results citations is now ready. 2019. as long as you have a supervised learning problem. Understanding Black-box Predictions via Influence Functions - PMLR P. Nakkiran, B. Neyshabur, and H. Sedghi. The precision of the output can be adjusted by using more iterations and/or Training test 7, Training 1, test 7 . Debruyne, M., Hubert, M., and Suykens, J. the prediction outcomes of an entire dataset or even >1000 test samples. To scale up influence functions to modern machine learning settings, we develop a simple, efficient implementation that requires only oracle access to gradients and Hessian-vector products. . Understanding Black-box Predictions via Influence Functions Proceedings of the 34th International Conference on Machine Learning . The more recent Neural Tangent Kernel gives an elegant way to understand gradient descent dynamics in function space. training time, and reduce memory requirements. The algorithm moves then The second mode is called calc_all_grad_then_test and In. 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. ICML 2017 best paperStanfordPang Wei KohCourseraStanfordNIPS 2019influence functionPercy Liang11Michael Jordan, , \hat{\theta}_{\epsilon, z} \stackrel{\text { def }}{=} \arg \min _{\theta \in \Theta} \frac{1}{n} \sum_{i=1}^{n} L\left(z_{i}, \theta\right)+\epsilon L(z, \theta), \left.\mathcal{I}_{\text {up, params }}(z) \stackrel{\text { def }}{=} \frac{d \hat{\theta}_{\epsilon, z}}{d \epsilon}\right|_{\epsilon=0}=-H_{\tilde{\theta}}^{-1} \nabla_{\theta} L(z, \hat{\theta}), , loss, \begin{aligned} \mathcal{I}_{\text {up, loss }}\left(z, z_{\text {test }}\right) &\left.\stackrel{\text { def }}{=} \frac{d L\left(z_{\text {test }}, \hat{\theta}_{\epsilon, z}\right)}{d \epsilon}\right|_{\epsilon=0} \\ &=\left.\nabla_{\theta} L\left(z_{\text {test }}, \hat{\theta}\right)^{\top} \frac{d \hat{\theta}_{\epsilon, z}}{d \epsilon}\right|_{\epsilon=0} \\ &=-\nabla_{\theta} L\left(z_{\text {test }}, \hat{\theta}\right)^{\top} H_{\hat{\theta}}^{-1} \nabla_{\theta} L(z, \hat{\theta}) \end{aligned}, \varepsilon=-1/n , z=(x,y) \\ z_{\delta} \stackrel{\text { def }}{=}(x+\delta, y), \hat{\theta}_{\epsilon, z_{\delta},-z} \stackrel{\text { def }}{=}\arg \min _{\theta \in \Theta} \frac{1}{n} \sum_{i=1}^{n} L\left(z_{i}, \theta\right)+\epsilon L\left(z_{\delta}, \theta\right)-\epsilon L(z, \theta), \begin{aligned}\left.\frac{d \hat{\theta}_{\epsilon, z_{\delta},-z}}{d \epsilon}\right|_{\epsilon=0} &=\mathcal{I}_{\text {up params }}\left(z_{\delta}\right)-\mathcal{I}_{\text {up, params }}(z) \\ &=-H_{\hat{\theta}}^{-1}\left(\nabla_{\theta} L(z_{\delta}, \hat{\theta})-\nabla_{\theta} L(z, \hat{\theta})\right) \end{aligned}, \varepsilon \delta \deltaloss, \left.\frac{d \hat{\theta}_{\epsilon, z_{\delta},-z}}{d \epsilon}\right|_{\epsilon=0} \approx-H_{\hat{\theta}}^{-1}\left[\nabla_{x} \nabla_{\theta} L(z, \hat{\theta})\right] \delta, \hat{\theta}_{z_{i},-z}-\hat{\theta} \approx-\frac{1}{n} H_{\hat{\theta}}^{-1}\left[\nabla_{x} \nabla_{\theta} L(z, \hat{\theta})\right] \delta, \begin{aligned} \mathcal{I}_{\text {pert,loss }}\left(z, z_{\text {test }}\right)^{\top} &\left.\stackrel{\text { def }}{=} \nabla_{\delta} L\left(z_{\text {test }}, \hat{\theta}_{z_{\delta},-z}\right)^{\top}\right|_{\delta=0} \\ &=-\nabla_{\theta} L\left(z_{\text {test }}, \hat{\theta}\right)^{\top} H_{\hat{\theta}}^{-1} \nabla_{x} \nabla_{\theta} L(z, \hat{\theta}) \end{aligned}, train lossH \mathcal{I}_{\text {up, loss }}\left(z, z_{\text {test }}\right) , -y_{\text {test }} y \cdot \sigma\left(-y_{\text {test }} \theta^{\top} x_{\text {test }}\right) \cdot \sigma\left(-y \theta^{\top} x\right) \cdot x_{\text {test }}^{\top} H_{\hat{\theta}}^{-1} x, influence functiondebug training datatraining point \mathcal{I}_{\text {up, loss }}\left(z, z_{\text {test }}\right) losstraining pointtraining point, Stochastic estimationHHHTFO(np)np, ImageNetdogfish900Inception v3SVM with RBF kernel, poisoning attackinfluence function59157%77%10590/591, attackRelated worktraining set attackadversarial example, influence functionbad case debug, labelinfluence function, \mathcal{I}_{\text {up,loss }}\left(z_{i}, z_{i}\right) , 10%labelinfluence functiontrain lossrandom, \mathcal{I}_{\text {up, loss }}\left(z, z_{\text {test }}\right), \mathcal{I}_{\text {up,loss }}\left(z_{i}, z_{i}\right), \mathcal{I}_{\text {pert,loss }}\left(z, z_{\text {test }}\right)^{\top}, H_{\hat{\theta}}^{-1} \nabla_{x} \nabla_{\theta} L(z, \hat{\theta}), Less Is Better: Unweighted Data Subsampling via Influence Function, influence functionleave-one-out retraining, 0.86H, SVMhinge loss0.95, straightforwardbest paper, influence functionloss. Measuring the effects of data parallelism on neural network training. Class will be held synchronously online every week, including lectures and occasionally tutorials. most harmful. Some JAX code examples for algorithms covered in this course will be available here. Understanding Black-box Predictions via Influence Functions ordered by helpfulness. Understanding Black-box Predictions via Inuence Functions 2. In Artificial Intelligence and Statistics (AISTATS), pages 3382-3390, 2019. Rethinking the Inception architecture for computer vision. Many Git commands accept both tag and branch names, so creating this branch may cause unexpected behavior. Model selection in kernel based regression using the influence function. We'll consider the two most common techniques for bilevel optimization: implicit differentiation, and unrolling. The reference implementation can be found here: link. D. Maclaurin, D. Duvenaud, and R. P. Adams. Szegedy, C., Vanhoucke, V., Ioffe, S., Shlens, J., and Wojna, Z. Optimizing neural networks with Kronecker-factored approximate curvature. When can we take advantage of parallelism to train neural nets? Noisy natural gradient as variational inference. 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. on the final predictions is straight forward. , . Interpreting black box predictions using Fisher kernels. Negative momentum for improved game dynamics. when calculating the influence of that single image. For these Here, we plot I up,loss against variants that are missing these terms and show that they are necessary for picking up the truly inuential training points. In Proceedings of the international conference on machine learning (ICML). Limitations of the empirical Fisher approximation for natural gradient descent. However, as stated Springenberg, J. T., Dosovitskiy, A., Brox, T., and Riedmiller, M. Striving for simplicity: The all convolutional net. Neural tangent kernel: Convergence and generalization in neural networks. Lectures will be delivered synchronously via Zoom, and recorded for asynchronous viewing by enrolled students. The answers boil down to an observation that neural net training seems to have two distinct phases: a small-batch, noise-dominated phase, and a large-batch, curvature-dominated one. We'll start off the class by analyzing a simple model for which the gradient descent dynamics can be determined exactly: linear regression. Automatically creates outdir folder to prevent runtime error, Merge branch 'expectopatronum-update-readme', Understanding Black-box Predictions via Influence Functions, import it as a package after it's in your, Combined, the original paper suggests that. Understanding Black-box Predictions via Influence Functions International Conference on Machine Learning (ICML), 2017. The To run the tests, further requirements are: You can either install this package directly through pip: Calculating the influence of the individual samples of your training dataset , loss , input space . to trace a model's prediction through the learning algorithm and back to its training data, prediction outcome of the processed test samples. ordered by harmfulness. Riemannian metrics for neural networks I: Feed-forward networks. The datasets for the experiments can also be found at the Codalab link. config is a dict which contains the parameters used to calculate the We'll consider two models of stochastic optimization which make vastly different predictions about convergence behavior: the noisy quadratic model, and the interpolation regime. The ACM Digital Library is published by the Association for Computing Machinery. Bilevel optimization refers to optimization problems where the cost function is defined in terms of the optimal solution to another optimization problem. While this class draws upon ideas from optimization, it's not an optimization class. The degree of influence of a single training sample z on all model parameters is calculated as: Where is the weight of sample z relative to other training samples. For the final project, you will carry out a small research project relating to the course content. understanding model behavior, debugging models, detecting dataset errors, Model-agnostic meta-learning for fast adaptation of deep networks. Students are encouraged to attend synchronous lectures to ask questions, but may also attend office hours or use Piazza. 10.5 Influential Instances | Interpretable Machine Learning - GitHub Pages vector to calculate the influence. How can we explain the predictions of a black-box model? 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 on to the next image. For modern neural nets, the analysis is more often descriptive: taking the procedures practitioners are already using, and figuring out why they (seem to) work. International conference on machine learning, 1885-1894, 2017. For one thing, the study of optimizaton is often prescriptive, starting with information about the optimization problem and a well-defined goal such as fast convergence in a particular norm, and figuring out a plan that's guaranteed to achieve it. On the limited memory BFGS method for large scale optimization. With the rapid adoption of machine learning systems in sensitive applications, there is an increasing need to make black-box models explainable. Understanding Black-box Predictions via Influence Functions Pang Wei Koh & Perry Liang Presented by -Theo, Aditya, Patrick 1 1.Influence functions: definitions and theory 2.Efficiently calculating influence functions 3. I. Sutskever, J. Martens, G. Dahl, and G. Hinton. non-convex non-differentialble . Interacting with predictions: Visual inspection of black-box machine learning models. Things get more complicated when there are multiple networks being trained simultaneously to different cost functions. Russakovsky, O., Deng, J., Su, H., Krause, J., Satheesh, S., Ma, S., Huang, Z., Karpathy, A., Khosla, A., Bernstein, M., et al. Li, J., Monroe, W., and Jurafsky, D. Understanding neural networks through representation erasure.