arXiv research

New method reveals why GNNs perform well on certain datasets.

Theory for RLHF generalization under reward shift and clipped KL.

Bayesian framework tackles measurement error in covariates.

Graph neural networks generalize well under certain conditions, explained by learning theory.

Study on SGD dynamics in neural networks, revealing generalisation patterns.

New bounds for shallow neural networks with deterministic parameters.

Quantum reservoirs risk bounds are analyzed using Rademacher complexity.

This work improves generalisation bounds using chaining and information theory.

Deep neural networks achieve stellar generalisation on a variety of problems, despite often being large enough to easily fit all their training data. Here we study the generalisation dynamics of two-layer neural networks in a teacher-student setup, where one network, the student, is trained using stochastic gradient de…

We provide bounds on control learning error in stochastic systems.

Study on infinitely-wide CNNs and their adaptability to function spatial scales.

Study reveals phase transition in neural networks near interpolation.

Study on generalisation in random feature learning and hidden manifold models.

OTS error shows small training error doesn't guarantee small test error.

Improved prediction of soil parameters using Multi-target Stacked Generalisation on EDXRF spectra.

Estimates and convergence of neural network approximations without structural assumptions.

The paper improves transformer generalization bounds using rank-dependent covering number bounds.

This research quantifies neural networks using magnitude, a topological invariant.

Kernel methods benefit from enforcing invariance, reducing generalization error.

The study analyzes multi-class teacher-student perceptron performance and generalization errors.

Gradient descent performs well on weakly convex losses, offering generalization guarantees.

New approach to compute generalization performance using known risk distribution.

PAC-Bayes bound for stable RNNs in time-series data.

ARC algorithm optimizes dynamic pricing with correlated observations.

SGD-trained deep nets often generalize well due to a strong inductive bias towards low-error, low-complexity functions.

SLT reveals how grokking occurs via basin selection in training.

Paper connects neural networks to Gaussian processes for understanding double-descent.

Improved similarity search in embeddings using InfoNCE loss.

PAC-Bayesian bounds improve understanding of K-NN classifier performance.

Double descent observed in tree-based models for genomic prediction.

A new kernel-based CI test improves on existing methods.

For information retrieval and binary classification, we show that precision at the top (or precision at k) and recall at the top (or recall at k) are maximised by thresholding the posterior probability of the positive class. This finding is a consequence of a result on constrained minimisation of the cost-sensitive exp…

The paper generalizes Cartan Geometry using Polacek and Siegel's approach.

New PAC-Bayes bound controls multiple error types simultaneously.

Constructs a unique Levi-Civita connection for generalised metrics.

The paper analyzes fluctuations in ensemble models in high-dimensional settings.

The paper applies generalised geometry to semi-Riemannian immersions and hypersurfaces.

Spectral embedding is a procedure which can be used to obtain vector representations of the nodes of a graph. This paper proposes a generalisation of the latent position network model known as the random dot product graph, to allow interpretation of those vector representations as latent position estimates. The general…

We address the problem of speech enhancement generalisation to unseen environments by performing two manipulations. First, we embed an additional recording from the environment alone, and use this embedding to alter activations in the main enhancement subnetwork. Second, we scale the number of noise environments presen…

No real-world reward function is perfect. Sensory errors and software bugs may result in RL agents observing higher (or lower) rewards than they should. For example, a reinforcement learning agent may prefer states where a sensory error gives it the maximum reward, but where the true reward is actually small. We formal…

New approach to T-duality using Courant algebroids.

Researchers explore statistical perspectives to understand GNN generalization.

Stochastic RNNs classify biological neural network paths with robust error bounds.

Paper tackles spooky effect in OSPA estimation, showing GOSPA solves it.

Defines T-duality and generalised Ricci flow relations using Courant algebroid relations.

We define $(p,q)$ hermitian geometry as the target space geometry of the two dimensional $(p,q)$ supersymmetric sigma model. This includes generalised Kähler geometry for $(2,2)$, generalised hyperkähler geometry for $(4,2)$, strong Kähler with torsion geometry for $(2,1)$ and strong hyperkähler with torsion geometry f…

We present and analyse three online algorithms for learning in discrete Hidden Markov Models (HMMs) and compare them with the Baldi-Chauvin Algorithm. Using the Kullback-Leibler divergence as a measure of generalisation error we draw learning curves in simplified situations. The performance for learning drifting concep…

New method models asymmetric data with improved tail dependence.