arXiv research

NAPP-ERM improves ERM with differential privacy guarantees by iteratively achieving target regularization and delivering strong convexity.

HEBAE improves VAEs by adaptively balancing reconstruction and regularization.

Over-parameterized models, such as DeepNets and ConvNets, form a class of models that are routinely adopted in a wide variety of applications, and for which Bayesian inference is desirable but extremely challenging. Variational inference offers the tools to tackle this challenge in a scalable way and with some degree o…

We present here a result of Monomialization of real analytic two-symmetric tensor fields over regular real analytic surfaces. We apply it to the (extension of the pull-back of the) inner metric of a resolved surface of a real analytic surface singularity. Doing so we recover Hsiang & Pati property at each point of the …

Variational inference offers scalable and flexible tools to tackle intractable Bayesian inference of modern statistical models like Bayesian neural networks and Gaussian processes. For largely over-parameterized models, however, the over-regularization property of the variational objective makes the application of vari…

This article studies hypoellipticity on general filtered manifolds. We extend the Rockland criterion to a pseudodifferential calculus on filtered manifolds, construct a parametrix and describe its precise analytic structure. We use this result to study Rockland sequences, a notion generalizing elliptic sequences to fil…

Kähler cones over Sasakian manifolds are flat if projectively induced.

Training certifiable neural networks enables one to obtain models with robustness guarantees against adversarial attacks. In this work, we introduce a framework to bound the adversary-free region in the neighborhood of the input data by a polyhedral envelope, which yields finer-grained certified robustness. We further …

Bayesian neural networks ignore data in infinite units limit.

The study defines a canonical nilpotent structure for certain collapsed manifolds.

Random feature approximation speeds up spectral methods and improves learning rates.

How can we make machine learning provably robust against adversarial examples in a scalable way? Since certified defense methods, which ensure $ε$-robust, consume huge resources, they can only achieve small degree of robustness in practice. Lipschitz margin training (LMT) is a scalable certified defense, but it can als…

We consider a statistical inverse learning problem, where we observe the image of a function $f$ through a linear operator $A$ at i.i.d. random design points $X_i$, superposed with an additive noise. The distribution of the design points is unknown and can be very general. We analyze simultaneously the direct (estimati…

Extends random feature analysis to spectral methods and improves learning rates.

Improved VAE for heavy-tailed data using Student's t-distributions.

Heavy-tailed regularization improves deep neural network performance.

The variational autoencoder (VAE) is a powerful generative model that can estimate the probability of a data point by using latent variables. In the VAE, the posterior of the latent variable given the data point is regularized by the prior of the latent variable using Kullback Leibler (KL) divergence. Although the stan…

Credit scoring models support loan approval decisions in the financial services industry. Lenders train these models on data from previously granted credit applications, where the borrowers' repayment behavior has been observed. This approach creates sample bias. The scoring model (i.e., classifier) is trained on accep…

Optimal self-distillation improves generative models' velocity risk and mode recovery.

This research improves binary classification by balancing overfitting and generalization with a novel Bayesian approach.

DP-BNNs improve accuracy, privacy, and reliability in neural networks.

Study reveals learning curves and benign overfitting in spectral algorithms for large dimensions.

Efficient local Lipschitz bounds improve neural network robustness.

Unsupervised learning with generative adversarial networks (GANs) has proven to be hugely successful. Regular GANs hypothesize the discriminator as a classifier with the sigmoid cross entropy loss function. However, we found that this loss function may lead to the vanishing gradients problem during the learning process…

This study compares machine learning algorithms for predictive performance and interpretability.

In this paper we first identify a basic limitation in gradient descent-based optimization methods when used in conjunctions with smooth kernels. An analysis based on the spectral properties of the kernel demonstrates that only a vanishingly small portion of the function space is reachable after a polynomial number of g…

Proposes QGC to distinguish between lower and upper tail connectivity in financial networks.

Optimal SD improves ridge regression performance strictly and precisely.