1 code implementation • 25 Feb 2024 • Ali Ebrahimpour Boroojeny, Matus Telgarsky, Hari Sundaram
We show the effectiveness of automatic differentiation in efficiently and correctly computing and controlling the spectrum of implicitly linear operators, a rich family of layer types including all standard convolutional and dense layers.
no code implementations • 24 Feb 2024 • Jingfeng Wu, Peter L. Bartlett, Matus Telgarsky, Bin Yu
We consider gradient descent (GD) with a constant stepsize applied to logistic regression with linearly separable data, where the constant stepsize $\eta$ is so large that the loss initially oscillates.
1 code implementation • 14 Feb 2024 • Clayton Sanford, Daniel Hsu, Matus Telgarsky
We show that a constant number of self-attention layers can efficiently simulate, and be simulated by, a constant number of communication rounds of Massively Parallel Computation.
no code implementations • 13 Jun 2023 • Justin D. Li, Matus Telgarsky
We first elucidate various fundamental properties of optimal adversarial predictors: the structure of optimal adversarial convex predictors in terms of optimal adversarial zero-one predictors, bounds relating the adversarial convex loss to the adversarial zero-one loss, and the fact that continuous predictors can get arbitrarily close to the optimal adversarial error for both convex and zero-one losses.
no code implementations • 4 Aug 2022 • Matus Telgarsky
This work establishes low test error of gradient flow (GF) and stochastic gradient descent (SGD) on two-layer ReLU networks with standard initialization, in three regimes where key sets of weights rotate little (either naturally due to GF and SGD, or due to an artificial constraint), and making use of margins as the core analytic technique.
no code implementations • 6 May 2022 • Miroslav Dudík, Robert E. Schapire, Matus Telgarsky
Not all convex functions on $\mathbb{R}^n$ have finite minimizers; some can only be minimized by a sequence as it heads to infinity.
no code implementations • 14 Feb 2022 • Matus Telgarsky
This work provides test error bounds for iterative fixed point methods on linear predictors -- specifically, stochastic and batch mirror descent (MD), and stochastic temporal difference learning (TD) -- with two core contributions: (a) a single proof technique which gives high probability guarantees despite the absence of projections, regularization, or any equivalents, even when optima have large or infinite norm, for quadratically-bounded losses (e. g., providing unified treatment of squared and logistic losses); (b) locally-adapted rates which depend not on global problem structure (such as condition numbers and maximum margins), but rather on properties of low norm predictors which may suffer some small excess test error.
no code implementations • ICLR 2022 • Yuzheng Hu, Ziwei Ji, Matus Telgarsky
We show that the simplest actor-critic method -- a linear softmax policy updated with TD through interaction with a linear MDP, but featuring no explicit regularization or exploration -- does not merely find an optimal policy, but moreover prefers high entropy optimal policies.
no code implementations • 1 Jul 2021 • Ziwei Ji, Nathan Srebro, Matus Telgarsky
We present and analyze a momentum-based gradient method for training linear classifiers with an exponentially-tailed loss (e. g., the exponential or logistic loss), which maximizes the classification margin on separable data at a rate of $\widetilde{\mathcal{O}}(1/t^2)$.
no code implementations • NeurIPS 2021 • Ziwei Ji, Justin D. Li, Matus Telgarsky
This work studies the behavior of shallow ReLU networks trained with the logistic loss via gradient descent on binary classification data where the underlying data distribution is general, and the (optimal) Bayes risk is not necessarily zero.
no code implementations • ICLR 2021 • Daniel Hsu, Ziwei Ji, Matus Telgarsky, Lan Wang
This paper theoretically investigates the following empirical phenomenon: given a high-complexity network with poor generalization bounds, one can distill it into a network with nearly identical predictions but low complexity and vastly smaller generalization bounds.
no code implementations • 19 Jun 2020 • Ziwei Ji, Miroslav Dudík, Robert E. Schapire, Matus Telgarsky
Recent work across many machine learning disciplines has highlighted that standard descent methods, even without explicit regularization, do not merely minimize the training error, but also exhibit an implicit bias.
no code implementations • NeurIPS 2020 • Ziwei Ji, Matus Telgarsky
In this paper, we show that although the minimizers of cross-entropy and related classification losses are off at infinity, network weights learned by gradient flow converge in direction, with an immediate corollary that network predictions, training errors, and the margin distribution also converge.
no code implementations • ICLR 2020 • Ziwei Ji, Matus Telgarsky, Ruicheng Xian
This paper establishes rates of universal approximation for the shallow neural tangent kernel (NTK): network weights are only allowed microscopic changes from random initialization, which entails that activations are mostly unchanged, and the network is nearly equivalent to its linearization.
no code implementations • ICLR 2020 • Ziwei Ji, Matus Telgarsky
Recent theoretical work has guaranteed that overparameterized networks trained by gradient descent achieve arbitrarily low training error, and sometimes even low test error.
no code implementations • 18 Jun 2019 • Bolton Bailey, Ziwei Ji, Matus Telgarsky, Ruicheng Xian
This paper investigates the approximation power of three types of random neural networks: (a) infinite width networks, with weights following an arbitrary distribution; (b) finite width networks obtained by subsampling the preceding infinite width networks; (c) finite width networks obtained by starting with standard Gaussian initialization, and then adding a vanishingly small correction to the weights.
no code implementations • 11 Jun 2019 • Ziwei Ji, Matus Telgarsky
On the other hand, with a properly chosen but aggressive step size schedule, we prove $O(1/t)$ rates for both $\ell_2$ margin maximization and implicit bias, whereas prior work (including all first-order methods for the general hard-margin linear SVM problem) proved $\widetilde{O}(1/\sqrt{t})$ margin rates, or $O(1/t)$ margin rates to a suboptimal margin, with an implied (slower) bias rate.
no code implementations • 8 Jun 2019 • Yu-cheng Chen, Matus Telgarsky, Chao Zhang, Bolton Bailey, Daniel Hsu, Jian Peng
This paper provides a simple procedure to fit generative networks to target distributions, with the goal of a small Wasserstein distance (or other optimal transport costs).
no code implementations • NeurIPS 2018 • Bolton Bailey, Matus Telgarsky
This paper investigates the ability of generative networks to convert their input noise distributions into other distributions.
no code implementations • ICLR 2019 • Ziwei Ji, Matus Telgarsky
This paper establishes risk convergence and asymptotic weight matrix alignment --- a form of implicit regularization --- of gradient flow and gradient descent when applied to deep linear networks on linearly separable data.
no code implementations • 20 Mar 2018 • Ziwei Ji, Matus Telgarsky
Gradient descent, when applied to the task of logistic regression, outputs iterates which are biased to follow a unique ray defined by the data.
no code implementations • 6 Nov 2017 • Ziwei Ji, Ruta Mehta, Matus Telgarsky
Consider the seller's problem of finding optimal prices for her $n$ (divisible) goods when faced with a set of $m$ consumers, given that she can only observe their purchased bundles at posted prices, i. e., revealed preferences.
1 code implementation • NeurIPS 2017 • Peter Bartlett, Dylan J. Foster, Matus Telgarsky
This paper presents a margin-based multiclass generalization bound for neural networks that scales with their margin-normalized "spectral complexity": their Lipschitz constant, meaning the product of the spectral norms of the weight matrices, times a certain correction factor.
no code implementations • ICML 2017 • Matus Telgarsky
Neural networks and rational functions efficiently approximate each other.
no code implementations • 13 Feb 2017 • Maxim Raginsky, Alexander Rakhlin, Matus Telgarsky
Stochastic Gradient Langevin Dynamics (SGLD) is a popular variant of Stochastic Gradient Descent, where properly scaled isotropic Gaussian noise is added to an unbiased estimate of the gradient at each iteration.
no code implementations • 21 Jul 2016 • Daniel Hsu, Matus Telgarsky
This paper investigates the following natural greedy procedure for clustering in the bi-criterion setting: iteratively grow a set of centers, in each round adding the center from a candidate set that maximally decreases clustering cost.
no code implementations • 14 Feb 2016 • Matus Telgarsky
For any positive integer $k$, there exist neural networks with $\Theta(k^3)$ layers, $\Theta(1)$ nodes per layer, and $\Theta(1)$ distinct parameters which can not be approximated by networks with $\mathcal{O}(k)$ layers unless they are exponentially large --- they must possess $\Omega(2^k)$ nodes.
no code implementations • 27 Sep 2015 • Matus Telgarsky
This note provides a family of classification problems, indexed by a positive integer $k$, where all shallow networks with fewer than exponentially (in $k$) many nodes exhibit error at least $1/6$, whereas a deep network with 2 nodes in each of $2k$ layers achieves zero error, as does a recurrent network with 3 distinct nodes iterated $k$ times.
no code implementations • 15 Jun 2015 • Matus Telgarsky, Miroslav Dudík, Robert Schapire
This paper proves, in very general settings, that convex risk minimization is a procedure to select a unique conditional probability model determined by the classification problem.
no code implementations • 2 Oct 2014 • Alekh Agarwal, Alina Beygelzimer, Daniel Hsu, John Langford, Matus Telgarsky
Can we effectively learn a nonlinear representation in time comparable to linear learning?
1 code implementation • 8 Nov 2013 • Matus Telgarsky, Sanjoy Dasgupta
Suppose $k$ centers are fit to $m$ points by heuristically minimizing the $k$-means cost; what is the corresponding fit over the source distribution?
no code implementations • 13 May 2013 • Matus Telgarsky
This manuscript provides optimization guarantees, generalization bounds, and statistical consistency results for AdaBoost variants which replace the exponential loss with the logistic and similar losses (specifically, twice differentiable convex losses which are Lipschitz and tend to zero on one side).
no code implementations • 29 Oct 2012 • Anima Anandkumar, Rong Ge, Daniel Hsu, Sham M. Kakade, Matus Telgarsky
This work considers a computationally and statistically efficient parameter estimation method for a wide class of latent variable models---including Gaussian mixture models, hidden Markov models, and latent Dirichlet allocation---which exploits a certain tensor structure in their low-order observable moments (typically, of second- and third-order).