arXiv research

Introduces CHL, a new loss function for continuous similarity learning.

New insights into continual learning with task similarity.

Study shows how task similarity affects forgetting in teacher-student setup.

Optimal task order improves continual learning performance.

Unique continuation result for expanding Ricci solitons.

CAM-GAN improves GANs for continual learning with efficient feature map transformations.

This paper investigates methods for quantifying similarity between audio signals, specifically for the task of of cover song detection. We consider an information-theoretic approach, where we compute pairwise measures of predictability between time series. We compare discrete-valued approaches operating on quantised au…

Improved similarity search in embeddings using InfoNCE loss.

Study on curvature blow-up and convergence of continuity method on Hirzebruch surface.

This paper formalizes state similarity metrics for reinforcement learning.

In 1997, J. Jost [27] and F. H. Lin [39], independently proved that every energy minimizing harmonic map from an Alexandrov space with curvature bounded from below to an Alexandrov space with non-positive curvature is locally Hölder continuous. In [39], F. H. Lin proposed a challenge problem: Can the Hölder continuity …

We propose a family of near-metrics based on local graph diffusion to capture similarity for a wide class of data sets. These quasi-metametrics, as their names suggest, dispense with one or two standard axioms of metric spaces, specifically distinguishability and symmetry, so that similarity between data points of arbi…

Continuous curves inscribe isosceles trapezoids in complex plane.

Study analyzes catastrophic forgetting in continual learning using teacher-student networks.

We study fractional stochastic volatility models in which the volatility process is a positive continuous function $σ$ of a continuous Gaussian process $\widehat{B}$. Forde and Zhang established a large deviation principle for the log-price process in such a model under the assumptions that the function $σ$ is globally…

Continuous epimorphisms between certain mapping class groups are induced by homeomorphisms.

We present the Variational Adaptive Newton (VAN) method which is a black-box optimization method especially suitable for explorative-learning tasks such as active learning and reinforcement learning. Similar to Bayesian methods, VAN estimates a distribution that can be used for exploration, but requires computations th…

Graph-based rehearsal improves continual learning performance.

The study explores homeomorphism groups of self-similar 2-manifolds, including the 2-sphere and Cantor set.

PEAKS selects key training examples incrementally based on prediction error and kernel similarity.

Efficient method for learning continuous exponential families beyond Gaussian.

OGD proves robustness to Catastrophic Forgetting in Continual Learning.

SX-GeoTree improves spatially coherent explanations in geospatial regression trees.

New method uses continuous OT for fairness, outperforming discrete OT.

Solves complex Monge-Ampère equations with Hölder continuous solutions in Kähler manifolds.

Study on neural networks' performance in sequential task learning.

Continuized Nesterov acceleration accelerates stochastic gradient descent and gossip algorithms.

New similarity measure for covariate shift improves nonparametric regression rates.

Excessive reuse of test data has become commonplace in today's machine learning workflows. Popular benchmarks, competitions, industrial scale tuning, among other applications, all involve test data reuse beyond guidance by statistical confidence bounds. Nonetheless, recent replication studies give evidence that popular…

LSH methods extend to function spaces for efficient similarity search.

Study on self-similar surfaces and their mapping class groups generated by involutions.

Continuous time random walks impose a random waiting time before each particle jump. Scaling limits of heavy tailed continuous time random walks are governed by fractional evolution equations. Space-fractional derivatives describe heavy tailed jumps, and the time-fractional version codes heavy tailed waiting times. Thi…

CLOPS improves deep learning for continuous physiological data.

New method detects symmetries beyond affine transformations.

Designs a framework to transfer causal models between similar environments.

Unified understanding of three continual learning regularisation methods.

Paper studies continuous prediction with experts' advice using differential equations.

New method uses dynamic programming for meta continual learning.

Logarithmic regret achieved in continuous-time linear-quadratic reinforcement learning.

Distributed representations of sentences have become ubiquitous in natural language processing tasks. In this paper, we consider a continual learning scenario for sentence representations: Given a sequence of corpora, we aim to optimize the sentence encoder with respect to the new corpus while maintaining its accuracy …

New framework discovers non-affine continuous symmetries in neural networks.

Entangled bisimulation improves policy learning from visual input.

Similarity graphs are an active research direction for the nearest neighbor search (NNS) problem. New algorithms for similarity graph construction are continuously being proposed and analyzed by both theoreticians and practitioners. However, existing construction algorithms are mostly based on heuristics and do not exp…

We define a new type of metric comparison similar to the comparison of Alexandrov. We show that it has strong connections to continuity of optimal transport between regular measures on a Riemannian manifold, in particular to the so called MTW condition introduced by Xi-Nan Ma, Neil Trudinger and Xu-Jia Wang.

A new method for continual learning in GANs learns new modes with limited data.

In the field of Natural Language Processing (NLP), we revisit the well-known word embedding algorithm word2vec. Word embeddings identify words by vectors such that the words' distributional similarity is captured. Unexpectedly, besides semantic similarity even relational similarity has been shown to be captured in word…

Introduces a continuous version of LWE problem.

New method balances multivariate model fitting for mixed likelihoods.