Approximate Newton-based statistical inference using only stochastic gradients
arXiv:1805.08920
Abstract
We present a novel statistical inference framework for convex empirical risk minimization, using approximate stochastic Newton steps. The proposed algorithm is based on the notion of finite differences and allows the approximation of a Hessian-vector product from first-order information. In theory, our method efficiently computes the statistical error covariance in -estimation, both for unregularized convex learning problems and high-dimensional LASSO regression, without using exact second order information, or resampling the entire data set. We also present a stochastic gradient sampling scheme for statistical inference in non-i.i.d. time series analysis, where we sample contiguous blocks of indices. In practice, we demonstrate the effectiveness of our framework on large-scale machine learning problems, that go even beyond convexity: as a highlight, our work can be used to detect certain adversarial attacks on neural networks.
References in corpus (13)
- Fashion-MNIST: a Novel Image Dataset for Benchmarking Machine Learning Algorithms
- Confidence Intervals and Hypothesis Testing for High-Dimensional Regression
- A tutorial on conformal prediction
- Online Batch Selection for Faster Training of Neural Networks
- Empirical Analysis of the Hessian of Over-Parametrized Neural Networks
- Stop Wasting My Gradients: Practical SVRG
- Bridging the Gap between Constant Step Size Stochastic Gradient Descent and Markov Chains
- Statistical Inference for Model Parameters in Stochastic Gradient Descent
- HiGrad: Uncertainty Quantification for Online Learning and Stochastic Approximation
- Optimal non-asymptotic bound of the Ruppert-Polyak averaging without strong convexity
- Efficient Representation of Low-Dimensional Manifolds using Deep Networks
- Deep nets for local manifold learning
- On Scalable Inference with Stochastic Gradient Descent