arXiv research

Structured subsampling improves privacy in deep time series forecasting.

Privacy is enhanced by synthetic data release even with unlimited data.

New method amplifies hidden structure in complex networks.

New sampling scheme improves privacy in DP-SGD without sacrificing utility.

Improves privacy amplification by shuffling for differential privacy.

We study the phenomenon of bias amplification in classifiers, wherein a machine learning model learns to predict classes with a greater disparity than the underlying ground truth. We demonstrate that bias amplification can arise via an inductive bias in gradient descent methods that results in the overestimation of the…

Mitigates gender bias amplification in model predictions.

Gradient amplification boosts deep learning model performance without increasing training time.

Privacy amplification improved through contraction coefficients and $E_γ$-divergence.

New method amplifies privacy in decentralized learning without centralized communication.

This study quantifies systemic importance in global banks using a continuous framework that amplifies localized shocks.

Paper analyzes tech adoption in financial networks, finding key leadership and diffusion dynamics.

AdaPrivate-TS: A differentially private Thompson Sampling algorithm for contextual bandits

Latent-state environments with long horizons, such as those faced by recommender systems, pose significant challenges for reinforcement learning (RL). In this work, we identify and analyze several key hurdles for RL in such environments, including belief state error and small action advantage. We develop a general prin…

New algorithm mitigates misspecification amplification in regression models with covariate shift.

A fundamental result in differential privacy states that the privacy guarantees of a mechanism are preserved by any post-processing of its output. In this paper we investigate under what conditions stochastic post-processing can amplify the privacy of a mechanism. By interpreting post-processing as the application of a…

Develops a framework for distilling flow models from few steps.

Many real world learning tasks involve complex or hard-to-specify objectives, and using an easier-to-specify proxy can lead to poor performance or misaligned behavior. One solution is to have humans provide a training signal by demonstrating or judging performance, but this approach fails if the task is too complicated…

New method for privacy amplification without sampling for matrix factorization.

Unified framework for subsampling mechanisms with tighter privacy guarantees.

The paper develops a theory explaining how machine learning models can amplify biases.

Differential privacy comes equipped with multiple analytical tools for the design of private data analyses. One important tool is the so-called "privacy amplification by subsampling" principle, which ensures that a differentially private mechanism run on a random subsample of a population provides higher privacy guaran…

Language is increasingly being used to define rich visual recognition problems with supporting image collections sourced from the web. Structured prediction models are used in these tasks to take advantage of correlations between co-occurring labels and visual input but risk inadvertently encoding social biases found i…

We mathematically compare four competing definitions of group-level nondiscrimination: demographic parity, equalized odds, predictive parity, and calibration. Using the theoretical framework of Friedler et al., we study the properties of each definition under various worldviews, which are assumptions about how, if at a…

FLAME improves privacy in federated learning without trusted parties.

Missing data enhances privacy in differential privacy.

Sharp pseudospectral bounds prevent transient amplification in coupled gradient descent.

The study explains how market-makers' hedging affects stock volatility during gamma-squeeze events.

Most of the econometric and econophysics models have been borrowed from the statistical physics, and as a cosequence, a new interdisciplinary science called econophysics has emerged. In this paper we planned to extend the analogy between different economic processes or phenomena and processes and phenomena from differe…

Synthetic data can amplify privacy in linear regression models.

We find empirically a characteristic sharp peak-flat trough pattern in a large set of commodity prices. We argue that the sharp peak structure reflects an endogenous inter-market organization, and that peaks may be seen as local ``singularities'' resulting from imitation and herding. These findings impose a novel strin…

The paper analyzes and improves privacy in machine learning through importance sampling.

Enhances privacy in machine learning through Rényi Pufferfish mechanisms.

Quantum algorithm solves best arm identification problem faster.

Momentum SGD fails to track nonstationary optima due to drift amplification.

Quantum method speeds up risk estimation for insurance tail risks.

This work studies differential privacy in the context of the recently proposed shuffle model. Unlike in the local model, where the server collecting privatized data from users can track back an input to a specific user, in the shuffle model users submit their privatized inputs to a server anonymously. This setup yields…

Study shows how sentiment shocks affect equity markets, revealing asymmetries and state-dependent effects.

Paper proves privacy guarantees for shuffled and online PNSGD, reducing noise over time.

Behavior cloning training instabilities amplified by SGD noise over long horizons.

Simpler, faster algorithm for uniformity testing in the shuffle model.

New method improves experimental design under model misspecification.

Given data drawn from an unknown distribution, $D$, to what extent is it possible to ``amplify'' this dataset and output an even larger set of samples that appear to have been drawn from $D$? We formalize this question as follows: an $(n,m)$ $\text{amplification procedure}$ takes as input $n$ independent draws from an …

Improved guarantees for misspecified kernelized bandit optimization.

Optimal privacy and accuracy in distributed mean estimation with compression.

Generative AI amplifies data without increasing information, with a mathematical limit.

The paper addresses bias amplification in prediction and decision-making using causal analysis.

Estimating properties of discrete distributions is a fundamental problem in statistical learning. We design the first unified, linear-time, competitive, property estimator that for a wide class of properties and for all underlying distributions uses just $2n$ samples to achieve the performance attained by the empirical…