arXiv research

RVFL NNs perform well without direct links and output bias for regression.

SGD benefits from a directional bias in kernel regression models.

Stochastic Gradient Descent shows directional bias with moderate learning rates, impacting optimization outcomes.

Study reveals how neural network architectures bias their learning based on feature directions.

Study compares ML algorithms for predicting stock market directional bias.

Extends Popularity Bias Memorization theorem to new conditions.

Proposes a neural network model to improve predictions in biased datasets.

Study proposes a new method to estimate bias-correction term for ATE estimation.

Deep ReLU networks escape from the origin via saddle points with a low-rank bias.

New algorithms estimate Hessians using random directions for faster stochastic optimization.

GD iterates for non-homogeneous deep nets increase margin and converge in direction.

We study the implicit bias of AdaGrad on separable linear classification problems. We show that AdaGrad converges to a direction that can be characterized as the solution of a quadratic optimization problem with the same feasible set as the hard SVM problem. We also give a discussion about how different choices of the …

GD at EoS edge minimizes logistic loss without monotonic convergence.

Unified framework for mean testing under truncation bias.

Improved ridge regression with Frequent Directions for large-scale tasks.

An intriguing phenomenon observed during training neural networks is the spectral bias, which states that neural networks are biased towards learning less complex functions. The priority of learning functions with low complexity might be at the core of explaining generalization ability of neural network, and certain ef…

A bias classifier is introduced to resist adversarial attacks.

Gradient descent implicitly follows regularization for general losses.

Bias correction improves language model training performance.

Ensembles improve classifier performance by reducing bias, not variance.

Decomposes bias in linear models under demographic parity constraints.

Geometric framework explains and controls implicit bias in machine learning.

Bayesian method corrects bias in imbalanced datasets.

Choosing appropriate step sizes is critical for reducing the computational cost of training large-scale neural network models. Mini-batch sub-sampling (MBSS) is often employed for computational tractability. However, MBSS introduces a sampling error, that can manifest as a bias or variance in a line search. This is bec…

We correct for sampling bias in training models to improve real-world performance.

We study the implicit bias of gradient descent methods in solving a binary classification problem over a linearly separable dataset. The classifier is described by a nonlinear ReLU model and the objective function adopts the exponential loss function. We first characterize the landscape of the loss function and show th…

Contemporary debates on filter bubbles and polarization in public and social media raise the question to what extent news media of the past exhibited biases. This paper specifically examines bias related to gender in six Dutch national newspapers between 1950 and 1990. We measure bias related to gender by comparing loc…

Paper debiases multiple word embedding biases simultaneously.

We rebias estimates to improve interval calibration and prediction accuracy.

New model reduces matrix factorization bias, yielding truly low-rank solutions.

Geometric framework analyzes bias in variational inference for posterior functionals.

New algorithm identifies causal relationships from graphs, even with selection bias.

New method quantifies inductive bias for machine learning tasks.

Paper distinguishes causal structures under latent confounding and selection bias.

This paper examines fairness and arbitrariness in bias mitigation methods.

New methods correct bias in LLM-as-a-Judge evaluations, but reliability depends on judge quality and model calibration.

CPCR mitigates bias in PCR for overparameterized models.

Study improves confidence measures in medical imaging pipelines by addressing bias.

Parametric images provide insight into the spatial distribution of physiological parameters, but they are often extremely noisy, due to low SNR of tomographic data. Direct estimation from projections allows accurate noise modeling, improving the results of post-reconstruction fitting. We propose a method, which we name…

Proposes bounds on bias from low-dimensional representations in CATE estimation.

New algorithm achieves optimal regret in average reward MDPs without prior bias information.

Two-layer networks favor simple features, especially in complex datasets.

Unified framework for stability and generalization of Push-Sum in decentralized learning over directed graphs.

Gradient descent stagnates in low-precision, but unbiased rounding schemes improve convergence.

New insights into bias-variance tradeoff for data-driven optimization under local misspecification.

Convolutional GANs favor low spatial frequencies, affecting fine detail generation.

Machine Learning (ML) is increasingly applied in real-life scenarios, raising concerns about bias in automatic decision making. We focus on bias as a notion of opinion exclusion, that stems from the direct application of traditional ML pipelines to infer subjective properties. We argue that such ML systems should be ev…

Improved GSPGS estimators reduce bias in noisy function measurements.