arXiv research

This research tackles uncertainty estimation in autoregressive structured prediction tasks.

The paper addresses probability calibration for incomplete sequences.

We discuss algorithms for estimating the Shannon entropy h of finite symbol sequences with long range correlations. In particular, we consider algorithms which estimate h from the code lengths produced by some compression algorithm. Our interest is in describing their convergence with sequence length, assuming no limit…

We give a new and complete proof of Hamilton's injectivity radius estimate for sequences with bounded and almost nonnegative curvature operators, unbounded diameters, and bump-like origins. Such sequences arise in particular from dilations about a singularity of the Ricci flow on a 3-manifold.

Non-parametric estimators improve quickest changepoint detection under irregular sequence lengths.

This paper compares log-likelihood and BLEU scores for sequence generation tasks.

The study finds counterexamples to curvature estimates for minimizing surfaces.

Paper optimizes change-point detection using learned distributions from training sequences.

Optimizes sliding window approach for tracking Gaussian densities.

Eye Movement analysis with Hidden Markov Models (EMHMM) is a method for modeling eye fixation sequences using hidden Markov models (HMMs). In this report, we run a simulation study to investigate the estimation error for learning HMMs with variational Bayesian inference, with respect to the number of sequences and the …

Reduces change detection to estimation using confidence sequences.

Estimates missing mass in Markovian sequences with linear runtime and near-optimal risk.

Develops a new point process model for detecting neural spike sequences.

Estimates stationary mass and frequency from non-i.i.d. data.

Transformers can simulate MLE for Bayesian network sequences.

In this thesis we describe how to estimate the distance spanned in the pants graph by a train track splitting sequence on a surface, up to multiplicative and additive constants. If some moderate assumptions on a splitting sequence are satisfied, each vertex set of a train track in it will represent a vertex of a graph …

Paper shows how online betting algorithms' regret can be used to create tight confidence sequences.

The paper tackles video prediction by estimating conditional densities implicitly.

A new method optimizes neural sequence models for better task performance.

Establishes 4D regularity for certain metric spaces.

Empirical Bayes method improves Gaussian sequence model inference.

We consider a problem of clustering a sequence of multinomial observations by way of a model selection criterion. We propose a form of a penalty term for the model selection procedure. Our approach subsumes both the conventional AIC and BIC criteria but also extends the conventional criteria in a way that it can be app…

We consider a binary sequence generated by thresholding a hidden continuous sequence. The hidden variables are assumed to have a compound symmetry covariance structure with a single parameter characterizing the common correlation. We study the parameter estimation problem under such one-parameter models. We demonstrate…

CRN model estimates treatment effects over time using adversarial balancing.

Bayesian-Deep Learning model predicts Covid-19 evolution in Spain.

LMC improves sampling from complex distributions using quasi-random sequences.

We present a study on reinforcement learning (RL) from human bandit feedback for sequence-to-sequence learning, exemplified by the task of bandit neural machine translation (NMT). We investigate the reliability of human bandit feedback, and analyze the influence of reliability on the learnability of a reward estimator,…

ATS2S model predicts RUL of industrial equipment using attention mechanism.

Sequence-to-Sequence (seq2seq) modeling has rapidly become an important general-purpose NLP tool that has proven effective for many text-generation and sequence-labeling tasks. Seq2seq builds on deep neural language modeling and inherits its remarkable accuracy in estimating local, next-word distributions. In this work…

Deep generative models have been successfully used to learn representations for high-dimensional discrete spaces by representing discrete objects as sequences and employing powerful sequence-based deep models. Unfortunately, these sequence-based models often produce invalid sequences: sequences which do not represent a…

A new method improves text generation quality and diversity.

Transformers handle infinite dimensional inputs effectively by feature extraction and dynamic feature selection.

The paper estimates variance of random sections on complex manifolds.

New approach speeds up DNA sequence alignment.

Spectral regularization simplifies sequence models by focusing on grammatical simplicity.

A new transformer model uses Monte Carlo methods for sequence prediction.

Efficient inference method for adaptive experiments with tighter confidence sequences.

The paper proposes a method to construct confidence sets using likelihood ratios for sequential decision-making.

In this work, sequence-to-sequence (seq2seq) models, originally developed for language translation, are used to predict the temporal evolution of complex, multi-physics computer simulations. The predictive performance of seq2seq models is compared to state transition models for datasets generated with multi-physics cod…

Unified view of gradient-based algorithms for stochastic convex composite optimization.

We present a theoretical analysis of Maximum a Posteriori (MAP) sequence estimation for binary symmetric hidden Markov processes. We reduce the MAP estimation to the energy minimization of an appropriately defined Ising spin model, and focus on the performance of MAP as characterized by its accuracy and the number of s…

New tools evaluate and optimize conditional sequence models in bioinformatics.

Method approximates Lipschitz domains with smoother shapes.

Generative Adversarial Networks (GANs) have seen steep ascension to the peak of ML research zeitgeist in recent years. Mostly catalyzed by its success in the domain of image generation, the technique has seen wide range of adoption in a variety of other problem domains. Although GANs have had a lot of success in produc…

A new framework predicts links in time-dependent networks using Bernoulli autoregression.

Study LASSO for high-dimensional VAR models with weakly dependent innovations.

Study improves regularity estimates for harmonic maps into ellipsoids.

New method accelerates gradient descent on curved spaces.