In this work we study loss functions for learning and evaluating probability distributions over large discrete domains. Unlike classification or regression where a wide variety of loss functions are used, in the distribution learning and density estimation literature, very few losses outside the dominant ar…
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.
168,695 papers · 148 categories
Trend · papers per month
48 results for “desirable loss distribution”
This work explores alternative learning criteria beyond traditional risk.
problem The trade-offs of optimizing for average performance in machine learning.
method Survey and introduction of non-traditional learning criteria.
result Emphasizes the importance of considering the loss distribution for desirable performance.
Paper introduces a new topological loss for better convergence.
problem Optimizing topological losses for model's desired topological behavior.
method Introduces a new regularized topology-aware loss function.
result Guarantees efficient optimization of the new loss function.
Estimation of importance sampling weights for off-policy evaluation of contextual bandits often results in imbalance - a mismatch between the desired and the actual distribution of state-action pairs after weighting. In this work we present balanced off-policy evaluation (B-OPE), a generic method for estimating weights…
Proposes a method to improve classification robustness against label noise.
problem Improving classification robustness against label noise.
method Extends a simple regression approach to classification, modeling labels as samples from a logistic-normal distribution.
result The method improves robustness against label noise in classification.
Generative models learn complex data from low-dimensional manifolds.
problem Theoretical justification for generative models on manifold structures.
method Prove statistical guarantees of generative networks under Wasserstein-1 loss, considering intrinsic dimensionality.
result Generative networks converge to zero at a fast rate depending on intrinsic dimensionality, not ambient data dimension.
In structural credit risk models, default events and the ensuing losses are both derived from the asset values at maturity. Hence it is of utmost importance to choose a distribution for these asset values which is in accordance with empirical data. At the same time, it is desirable to still preserve some analytical tra…
Operational risk models commonly employ maximum likelihood estimation (MLE) to fit loss data to heavy-tailed distributions. Yet several desirable properties of MLE (e.g. asymptotic normality) are generally valid only for large sample-sizes, a situation rarely encountered in operational risk. In this paper, we study how…
Significant advances have been made recently on training neural networks, where the main challenge is in solving an optimization problem with abundant critical points. However, existing approaches to address this issue crucially rely on a restrictive assumption: the training data is drawn from a Gaussian distribution. …
Generative models often misrepresent class frequencies; this paper calibrates them.
problem Miscalibration of class frequencies in generative models.
method Formulated as constrained optimization, using surrogate objectives to approximate constraints.
result Significant reduction in calibration error across various models and applications.
Neural Processes (NPs) are a class of models that learn a mapping from a context set of input-output pairs to a distribution over functions. They are traditionally trained using maximum likelihood with a KL divergence regularization term. We show that there are desirable classes of problems where NPs, with this loss, f…
The paper proposes a new method to measure risk with fine-grained tail sensitivity.
problem Risk measures that do not account for tail sensitivity are insufficient for machine learning systems.
method The approach involves specifying a reference distribution with desired tail behavior and constructing risk measures compatible with this upper probability.
result Risk measures with fine-grained tail sensitivity can replace the expectation operator in machine learning systems.
Wasserstein GANs fail to approximate Wasserstein distance, leading to their success.
problem Approximating Wasserstein distance in deep generative models.
method Analysis of differences between theoretical setup and training reality.
result Wasserstein GANs' success is due to their failure to approximate Wasserstein distance.
Conditional Distribution Function Estimation Using Neural Networks for Censored and Uncensored Datastat.ME
Estimates conditional distribution function using neural networks for censored and uncensored data.
problem Estimating conditional distribution function for censored and uncensored data.
method Neural network algorithm based on Cox regression with time-dependent covariates, using full likelihood with unconstrained optimization.
result Proposed method yields more accurate estimates than existing methods when model assumptions are violated.
On Next-Token Prediction in LLMs: How End Goals Determine the Consistency of Decoding Algorithmsstat.ML
This paper examines how different decoding algorithms for LLMs align with various goals.
problem Consistency of decoding algorithms with different goals in LLMs.
method Analysis of greedy, lookahead, random sampling, and temperature-scaled random sampling algorithms.
result Random sampling is consistent with the true probability distribution, but other goals require optimal algorithms for specific probability distributions.
Generative models learn distributions, new method finds inputs matching desired conditional distributions.
problem Designing inputs that produce specific conditional distributions, not just points.
method Conditional Distribution Matching (CDM) and MLGD-F algorithm.
result MLGD-F reliably recovers inputs matching diverse user-specified conditional distributions.
Fewer degrees of freedom can train deep networks, showing a sharp phase transition.
problem Training deep networks with fewer degrees of freedom than parameters.
method Examined success probability of hitting training loss sub-level sets within random subspaces.
result Threshold training dimension increases as desired final loss decreases.
Generalized Huber Loss for Robust Learning and its Efficient Minimization for a Robust Statisticsstat.ML
Proposes a new Huber loss combining absolute and quadratic properties.
problem Improving robustness in learning models.
method Introduces a generalized Huber loss with a log-exp transform and provides an efficient minimization algorithm.
result Shows that the new loss function can be minimized efficiently.
Proposes a new loss function for robust learning.
problem Creating a robust loss function for machine learning.
method Extended pseudo Huber loss with log-exp transform and logistic function.
result Linear convergence algorithm for minimizer finding.
In the UK betting market, bookmakers often offer a free coupon to new customers. These free coupons allow the customer to place extra bets, at lower risk, in combination with the usual betting odds. We are interested in whether a customer can exploit these free coupons in order to make a sure gain, and if so, how the c…
We present -loss, , a tunable loss function for binary classification that bridges log-loss () and - loss (). We prove that -loss has an equivalent margin-based form and is classification-calibrated, two desirable properties for a good surrogate loss function for the ideal y…
UAMM uses external market prices to improve AMM efficiency and reduce liquidity provider risk.
problem Traditional AMMs lack consideration of external markets and risk management.
method UAMM calculates prices by incorporating external market prices and impermanent loss, maintaining constant product curve properties.
result UAMM eliminates arbitrage opportunities when external market prices are efficient, reducing liquidity provider risk.
Hi-fi priors enhance BNNs by learning flexible activations.
problem Challenging to impose function-space priors on BNNs.
method Optimization techniques to learn flexible activations.
result BNNs with flexible activations can achieve desired priors.
Paper introduces privacy-preserving inventory policy learning for feature-based newsvendor with unknown demand.
problem Privacy-preserving inventory policy learning for feature-based newsvendor with unknown demand distribution and nonsmooth loss function.
method Developed a clipped noisy gradient descent algorithm based on convolution smoothing for optimal inventory estimation within f-differential privacy framework.
result Achieved privacy-preserving optimal inventory policy with provable privacy guarantees and desirable statistical precision.
Paper extends FFT-based differential privacy method to heterogeneous compositions.
problem Computing accurate differential privacy guarantees for mixed mechanisms.
method Uses Fast Fourier Transform (FFT) for error analysis and parameter selection.
result Provides tighter bounds for heterogeneous compositions compared to homogeneous cases.
Paper proposes a method to efficiently cluster stretched mixtures.
problem Clustering stretched elliptical mixtures using standard methods like PCA and k-means fails.
method Proposes a non-convex program to transform data into a one-dimensional point cloud.
result Efficient first-order algorithm achieves near-optimal statistical precision.
Sharpe et al. proposed the idea of having an expected utility maximizer choose a probability distribution for future wealth as an input to her investment problem instead of a utility function. They developed a computer program, called The Distribution Builder, as one way to elicit such a distribution. In a single-perio…
Max-margin learning is a powerful approach to building classifiers and structured output predictors. Recent work on max-margin supervised topic models has successfully integrated it with Bayesian topic models to discover discriminative latent semantic structures and make accurate predictions for unseen testing data. Ho…
Selective removal of data subsets can efficiently unlearn unwanted distributions.
problem Efficiently removing unwanted data subsets without losing important information.
method Formalized as distributional unlearning, using Kullback-Leibler divergence constraints to select a small subset of data.
result Proposed method achieves corresponding log-loss bounds and is quadratically more sample-efficient than random removal.
Heavy Lasso improves robustness in high-dimensional linear regression with heavy-tailed errors.
problem Challenges of classical Lasso in handling heavy-tailed noise and outliers.
method Data-augmented soft-thresholding with Student's t-distribution loss.
result Heavy Lasso achieves comparable rates to Huber loss under theoretical bounds.
The gain-loss ratio is known to enjoy very good properties from a normative point of view. As a confirmation, we show that the best market gain-loss ratio in the presence of a random endowment is an acceptability index and we provide its dual representation for general probability spaces. However, the gain-loss ratio w…
ANGLE tackles circular data regression, improving predictive performance.
problem Geometrically misleading traditional regression for circular data.
method Generative map optimized via GCES loss for non-parametric distributional regression.
result Unified toolbox for circular statistics challenges.
Model calculates capital requirements for multi-line insurance companies.
problem Measuring and capitalizing on incurred claims risk for multi-line property and casualty insurers.
method Stochastic model integrating accident semester, development lag effects, autocorrelation, and hierarchical copula.
result Model accurately reproduces empirical loss ratio dynamics and quantifies overall portfolio risk.
Proposes a deep ordinal regression framework using optimal transport loss and unimodal output probabilities.
problem Lack of unimodal output probabilities in recent ordinal regression models.
method Introduces a deep learning framework based on optimal transport loss and unimodal output distribution, inspired by the Proportional Odds model.
result Demonstrates improved performance and unimodal output probabilities on real-world datasets compared to existing methods.
Cross-entropy loss together with softmax is arguably one of the most common used supervision components in convolutional neural networks (CNNs). Despite its simplicity, popularity and excellent performance, the component does not explicitly encourage discriminative learning of features. In this paper, we propose a gene…
In high-dimensional classification settings, we wish to seek a balance between high power and ensuring control over a desired loss function. In many settings, the points most likely to be misclassified are those who lie near the decision boundary of the given classification method. Often, these uninformative points sho…
Support vector machines (SVMs) are special kernel based methods and belong to the most successful learning methods since more than a decade. SVMs can informally be described as a kind of regularized M-estimators for functions and have demonstrated their usefulness in many complicated real-life problems. During the last…
On Cyclical MCMC Samplingstat.CO
Cyclical MCMC tackles high-dimensional multimodal distributions, showing convergence under certain conditions.
problem High-dimensional multimodal posterior distributions in deep learning.
method Cyclical MCMC framework that tracks tempered versions of the target distribution over time.
result Cyclical MCMC converges to the target distribution under fast mixing kernels but fails in slow mixing cases.
While optimizing convex objective (loss) functions has been a powerhouse for machine learning for at least two decades, non-convex loss functions have attracted fast growing interests recently, due to many desirable properties such as superior robustness and classification accuracy, compared with their convex counterpa…
SEMF predicts prediction intervals for ML models using latent variables.
problem Uncertainty quantification in ML models, especially for diverse data distributions.
method Supervised Expectation-Maximization Framework (SEMF) extending EM algorithm for latent variable modeling.
result SEMF produces narrower prediction intervals with desired coverage probability.
This paper extends performative prediction to nonlinear cases.
problem Performative prediction's effectiveness is limited by linear assumptions in real-world applications.
method Formulated a maximum margin approach loss function and extended it to nonlinear spaces using kernel methods.
result Derived conditions for performative stability in both linear and nonlinear cases.
Paper proposes new loss functions for training energy networks.
problem Challenges in computing gradients for training energy networks.
method Proposes generalized Fenchel-Young losses for efficient gradient computation.
result Demonstrates the calibration of excess risk for linear-concave energies.
Stochastic GD converges linearly for CV@R learning under certain conditions.
problem Optimizing CV@R in statistical learning with non-convex loss functions.
method Stochastic Gradient Descent with Polyak-Łojasiewicz condition.
result Stochastic GD achieves linear convergence for CV@R learning.
Paper proposes a method to design molecules with specific properties.
problem Designing molecules with desired chemical and biological properties.
method Energy-based model in latent space, SGDS algorithm for gradual distribution shifting.
result Method achieves strong performances on various molecule design tasks.
Data containing human or social attributes may over- or under-represent groups with respect to salient social attributes such as gender or race, which can lead to biases in downstream applications. This paper presents an algorithmic framework that can be used as a data preprocessing method towards mitigating such bias.…
We propose a robust inferential procedure for assessing uncertainties of parameter estimation in high-dimensional linear models, where the dimension can grow exponentially fast with the sample size . Our method combines the de-biasing technique with the composite quantile function to construct an estimator that …
Study improves generalization bounds for linear regression across tasks.
problem Improving generalization in high-dimensional regression problems.
method Distribution-dependent bounds on generalization error for L1, L2, and elastic net regularization.
result Generalization bounds improve with data distribution niceness and do not degrade with feature dimension.
We present a fully-supervized method for learning to segment data structured by an adjacency graph. We introduce the graph-structured contrastive loss, a loss function structured by a ground truth segmentation. It promotes learning vertex embeddings which are homogeneous within desired segments, and have high contrast …