The performance of standard learning procedures has been observed to differ widely across groups. Recent studies usually attribute this loss discrepancy to an information deficiency for one group (e.g., one group has less data). In this work, we point to a more subtle source of loss discrepancy---feature noise. Our mai…
On-device research index
arXiv research
A locally-built, LLM-digested index of recent arXiv papers in quant finance, geometry/topology, and statistical ML — keyword search served straight from SQLite on this machine.
168,742 papers · 148 categories
Trend · papers per month
48 results for “discrepancy loss”
Bayes-consistent disagreement discrepancy loss improves model robustness.
problem Distribution shift in real-world neural network deployment.
method Introducing a novel disagreement loss that is Bayes consistent.
result Proves existing surrogates for disagreement discrepancy are not Bayes consistent.
Study shows the corrected Akaike criterion is inadmissible for estimating Kullback-Leibler discrepancy.
problem Inadmissibility of the corrected Akaike information criterion for estimating Kullback-Leibler discrepancy.
method Loss estimation framework to demonstrate inadmissibility and provide improved estimators.
result Improved estimators of Kullback-Leibler discrepancy are provided and perform well in reduced-rank situations.
Though machine learning algorithms excel at minimizing the average loss over a population, this might lead to large discrepancies between the losses across groups within the population. To capture this inequality, we introduce and study a notion we call maximum weighted loss discrepancy (MWLD), the maximum (weighted) d…
Bayesian inference uses Stein discrepancy for robustness in intractable likelihoods.
problem Intractable likelihoods in Bayesian inference.
method Generalised Bayesian inference with Stein discrepancy as the loss function.
result Robust generalised posteriors with closed form or accessible using MCMC.
A new loss function ED simplifies training energy-based models without scores.
problem Training energy-based models is computationally expensive.
method Energy Discrepancy (ED) loss function that does not rely on scores or MCMC.
result ED effectively interpolates between score matching and negative log-likelihood.
Generative adversarial networks (GANs) generate data based on minimizing a divergence between two distributions. The choice of that divergence is therefore critical. We argue that the divergence must take into account the hypothesis set and the loss function used in a subsequent learning task, where the data generated …
This paper defines the notion of class discrepancy for families of functions. It shows that low discrepancy classes admit small offline and streaming coresets. We provide general techniques for bounding the class discrepancy of machine learning problems. As corollaries of the general technique we bound the discrepancy …
A new method trains discrete EBMs without sampling.
problem Training EBMs on discrete spaces is hard.
method Energy Discrepancy (ED), a contrastive loss.
result ED offers theoretical guarantees for various perturbation types.
Quantum models face barren plateaus, but specific losses can be trainable.
problem Barren plateaus and loss concentration in quantum generative models.
method Investigated explicit and implicit losses, and their interplay.
result Explicit losses lead to new barren plateaus, while implicit losses can be trainable.
Partial model averaging improves Federated Learning performance.
problem Periodic model averaging causes significant model discrepancy in Federated Learning.
method Proposes a partial model averaging framework that encourages local models to stay close to each other.
result Partial averaging achieves up to 2.2% higher validation accuracy than full averaging.
SRRM improves recursive transport surrogates in the small-discrepancy regime.
problem Insufficient understanding of recursive partitioning methods' statistical behavior and resolution in the small-discrepancy regime.
method Introduced Selective Recursive Rank Matching (SRRM) to improve the resolution of Recursive Rank Matching (RRM).
result SRRM yields a higher-fidelity practical surrogate for the Wasserstein distance at moderate additional computational cost.
A new method for domain generalization using unlabeled data.
problem Learning from multiple domains with limited labeled data.
method Combines meta learning and semi-supervised learning with entropy-based pseudo-labeling and discrepancy loss.
result Significantly outperforms state-of-the-art methods on benchmark datasets.
Active learning algorithms propose which unlabeled objects should be queried for their labels to improve a predictive model the most. We study active learners that minimize generalization bounds and uncover relationships between these bounds that lead to an improved approach to active learning. In particular we show th…
Deep learning model approximates stochastic responses.
problem Approximating stochastic responses using neural networks.
method Generative neural network with conditional maximum mean discrepancy (CMMD) loss.
result Excellent performance on benchmark problems.
Generative neural network simulates characteristic functions.
problem Simulating from characteristic functions inaccessible in closed form.
method Generative neural network with Maximum-Mean-Discrepancy loss.
result Universal algorithm independent of dimensionality and function properties.
Paper proposes D3M to improve anti-spoofing detection by balancing loss function and using complementary features.
problem Improving automatic speaker verification systems against high-quality playback attacks.
method D3M uses a balanced focal loss function to dynamically scale loss based on sample traits, and combines three feature types for robust detection.
result D3M systems outperform conventional methods significantly, achieving min-tDCF of 0.0124 and EER of 0.55%.
We study minimax convergence rates of nonparametric density estimation under a large class of loss functions called "adversarial losses", which, besides classical losses, includes maximum mean discrepancy (MMD), Wasserstein distance, and total variation distance. These losses are closely related to the …
A new ensemble filter uses transport maps and MMD optimization for high-dimensional data assimilation.
problem High-dimensional data assimilation challenges in ensemble filtering.
method Optimized Maximum Mean Discrepancy (MMD) for transport map construction.
result Significant improvement in robustness and posterior approximation.
Generative model for TPPs using signatures and distributional discrepancies.
problem Limitations of signature methods for TPPs and lack of global sequence-level loss in neural models.
method Introduce interarrival embedding to lift jump paths to continuous paths of bounded variation, enabling signature methods for discrete event sequences. Develop sigTPP, a signature-based generative model trained on path-level loss.
result sigTPP achieves the best average rank across multiple metrics and outperforms or is within a standard error of the strongest baseline in 64% of dataset-metric pairs.
Enhances generative models stability and accuracy with BNPL, WMMD, and triple model.
problem Overfitting in GANs and noisy samples in VAEs.
method Bayesian non-parametric learning framework, integrating Wasserstein distance and maximum mean discrepancy.
result Superior performance across various generative tasks.
Improved kernel Stein discrepancy for large-scale data.
problem Efficiently testing probability distributions with kernel methods.
method Nyström approximation to reduce runtime complexity.
result Nyström-based KSD is -consistent and applicable for large datasets.
New autoencoder improves latent space learning by optimizing sliced Gromov-Wasserstein discrepancies.
problem Improving inner discrepancy between prior and posterior distributions in autoencoders.
method Proposed spherical sliced fused Gromov Wasserstein (SSFG) and variants (MSSFG, PSSFG) to find important directions.
result New autoencoders achieve favorable performance in latent manifold learning, image generation, and reconstruction.
DPC uses physics and neural nets to solve SDEs.
problem Solving stochastic differential equations with missing physics.
method Physics-data fusion with conditional maximum mean discrepancy (CMMD) loss.
result DPC achieves highly accurate solutions on benchmark examples.
Paper improves MMD estimation for analytical mean embeddings.
problem Improving MMD estimation for distributions with analytical mean embeddings.
method Proposes a tighter concentration result for MMD estimation under semi-explicit settings and extends to unbounded kernels.
result Demonstrates efficiency in real-world applications like index replication and calibration.
KSD Descent uses KSD to sample from a target distribution efficiently.
problem Sampling from complex target distributions efficiently.
method Wasserstein gradient flow of KSD, using L-BFGS optimization.
result KSD Descent can sample from a target distribution using a set of particles.
Framework learns to transform majority to minority samples for balanced classification.
problem Imbalanced classification leading to biased predictions.
method Minimizes MMD and uses triplet loss for global alignment and boundary awareness.
result Consistent improvements over classical and generative baselines in AUROC, G-mean, F1-score, and MCC.
A new deep generative model uses BSDEs for high-dimensional data generation.
problem Generating high-dimensional complex data, especially images.
method Combines BSDEs with deep neural networks for training with MMD loss.
result BSDE-Gen effectively generates high-dimensional data with stochasticity.
New method gives provable error bounds for neural nets under distribution shift.
problem Proving reliable error bounds for neural networks under distribution shift.
method Optimizing a classifier to disagree with another, using a new 'disagreement loss'.
result Valid error bounds with comparable accuracy to competitive methods.
A new method improves Bayesian inference for multimodal posteriors.
problem Insensitivity to well-separated modes in multimodal posteriors.
method Weighted Kernel Stein Discrepancy method.
result Significantly improved mode sensitivity compared to standard KSD-Bayes.
Recent unsupervised approaches to domain adaptation primarily focus on minimizing the gap between the source and the target domains through refining the feature generator, in order to learn a better alignment between the two domains. This minimization can be achieved via a domain classifier to detect target-domain feat…
A new metric detects non-Markovian states in partially observable environments.
problem Learning state representations in partially observable environments.
method Introducing the -discrepancy metric to detect non-Markovian states.
result The -discrepancy is zero for Markov processes and non-zero for partially observable environments.
We study strictly proper scoring rules in the Reproducing Kernel Hilbert Space. We propose a general Kernel Scoring rule and associated Kernel Divergence. We consider conditions under which the Kernel Score is strictly proper. We then demonstrate that the Kernel Score includes the Maximum Mean Discrepancy as a special …
The paper extends mixability theory to function-valued forecasts, proving various loss functions are mixable.
problem Efficient aggregation of functional and probabilistic forecasts in online prediction games.
method Adapting mixable and exponentially concave loss functions to function-valued forecasts.
result Various loss functions used for probabilistic forecasting are mixable (exp-concave).
We unify f-divergences, Bregman divergences, surrogate loss bounds (regret bounds), proper scoring rules, matching losses, cost curves, ROC-curves and information. We do this by systematically studying integral and variational representations of these objects and in so doing identify their primitives which all are rela…
AGMMNs improve learning of copula models by adaptively selecting kernels.
problem Learning dependence structures in copula models.
method Adaptive bandwidth selection for MMD in GMMNs, increasing kernels based on validation loss.
result AGMMNs significantly improve training performance over GMMNs and parametric models.
Paper develops a framework to optimize neural networks using weighted metrics.
problem Discrepancy between maximizing weighted classification scores and minimizing loss function.
method Formalizes weighted classification metrics and constructs corresponding losses.
result Framework includes well-established approaches like cost-sensitive learning and weighted cross entropy.
Optimizes kernel discrepancies by selecting subsets efficiently.
problem Improving kernel discrepancies for QMC methods.
method Introduces a novel subset selection algorithm for kernel discrepancies.
result Efficiently generates low-discrepancy samples from various distributions.
Improved speech enhancement using diffusion models with MSE loss.
problem Efficient incorporation of noisy speech in generative speech enhancement.
method Augmented diffusion-based generative model with a MSE loss for enhanced speech.
result Proposed method improves speech enhancement performance compared to original diffusion model.
We propose a principled method for gradient-based regularization of the critic of GAN-like models trained by adversarially optimizing the kernel of a Maximum Mean Discrepancy (MMD). We show that controlling the gradient of the critic is vital to having a sensible loss function, and devise a method to enforce exact, ana…
Generative adversarial nets (GANs) are widely used to learn the data sampling process and their performance may heavily depend on the loss functions, given a limited computational budget. This study revisits MMD-GAN that uses the maximum mean discrepancy (MMD) as the loss function for GAN and makes two contributions. F…
EPD method accurately captures parameter distributions from RCS data.
problem Limitations of traditional methods in estimating parameter distributions from RCS data.
method EPD method generates synthetic trajectories, estimates parameters, and selects parameters based on discrepancy.
result EPD provides accurate distribution of parameters without data loss.
In some misspecified settings, the posterior distribution in Bayesian statistics may lead to inconsistent estimates. To fix this issue, it has been suggested to replace the likelihood by a pseudo-likelihood, that is the exponential of a loss function enjoying suitable robustness properties. In this paper, we build a ps…
New discrepancy function compares discrete probability measures considering space geometry.
problem Comparing discrete probability measures in a geometrically meaningful way.
method Proposes the Fourier Discrepancy Function, proving convexity, differentiability, and providing gradient formula.
result Proves the Fourier Discrepancy is convex, twice differentiable, and provides an explicit gradient formula.
This paper introduces localized discrepancy theories for unsupervised domain adaptation.
problem Improving generalization bounds for unsupervised domain adaptation.
method Localized discrepancies defined on the hypothesis space after localization, leading to smaller and asymmetric values.
result Improved generalization bounds and sample complexity reduction.
Improves deep learning training by matching mini-batch distributions.
problem Overfitting and noise in mini-batch training.
method ITDM, which matches the moments of mini-batch distributions to reduce overfitting.
result ITDM reduces overfitting and improves DNN training.
The article introduces practical estimators for kernel discrepancies.
problem Estimating kernel discrepancies accurately and efficiently.
method Presented various estimators for MMD, HSIC, and KSD, including V-statistics, U-statistics, and incomplete U-statistics. Stressed the importance of kernel bandwidth and introduced adaptive estimators.
result Adaptive estimators combining multiple estimators with various kernels address the problem of kernel selection.
The study optimizes bounds for comparing training and population loss.
problem Optimizing bounds for comparing training and population loss.
method Derives generic information-theoretic and PAC-Bayesian generalization bounds using convex comparator functions.
result The tightest possible bound is obtained with the comparator being the convex conjugate of the CGF of the bounding distribution.