Optimal search for change point anomaly in multiple processes.
problem Detecting a change point in an anomalous process among multiple normal processes.
method Deterministic search algorithm balancing sample complexity and detection accuracy.
result Asymptotically optimal in minimizing Bayes risk.
Under Markovian assumptions, we leverage a Central Limit Theorem (CLT) for the empirical measure in the test statistic of the composite hypothesis Hoeffding test so as to establish weak convergence results for the test statistic, and, thereby, derive a new estimator for the threshold needed by the test. We first show t…
Differential privacy has seen remarkable success as a rigorous and practical formalization of data privacy in the past decade. This privacy definition and its divergence based relaxations, however, have several acknowledged weaknesses, either in handling composition of private algorithms or in analyzing important primi…
Detects anomalies without training data using deep learning.
problem Detect anomalies in data without labeled training data.
method Inverse Generative Adversarial Network (GAN) for semi-supervised learning.
result Successfully detects anomalies in data without labeled training data.
While statistics focusses on hypothesis testing and on estimating (properties of) the true sampling distribution, in machine learning the performance of learning algorithms on future data is the primary issue. In this paper we bridge the gap with a general principle (PHI) that identifies hypotheses with best predictive…
Study robust hypothesis testing under Hellinger distance, proving lower bounds and providing tests.
problem Testing close variants of specified distributions robustly to Hellinger distance.
method Lower bound on slack factor, testing with Hellinger balls, symmetric chi-squared distance analysis.
result Lower bound on slack factor quantifies robustness under misspecification.
Paper generalizes paracomposition and change of variables for paradifferential operators.
problem Generalizing paracomposition and change of variables for paradifferential operators in low regularity settings.
method Drops diffeomorphism hypothesis, estimates in Sobolev and Zygmund spaces, discusses pull-back of pseudodifferential and paradifferential operators.
result Sharp estimates for composition in Sobolev and Zygmund spaces, change of variables in paradifferential operators.
PEAK tests means of multiple data streams with sequential betting.
problem Testing means of multiple data streams with nonparametric methods.
method Sequential, nonparametric testing using a betting scheme.
result PEAK provides up to 85% reduction in samples for stopping.
We study the power of interactivity in local differential privacy. First, we focus on the difference between fully interactive and sequentially interactive protocols. Sequentially interactive protocols may query users adaptively in sequence, but they cannot return to previously queried users. The vast majority of exist…
The problem of multi-hypothesis testing with controlled sensing of observations is considered. The distribution of observations collected under each control is assumed to follow a single-parameter exponential family distribution. The goal is to design a policy to find the true hypothesis with minimum expected delay whi…
We study the portfolio problem of maximizing the outperformance probability over a random benchmark through dynamic trading with a fixed initial capital. Under a general incomplete market framework, this stochastic control problem can be formulated as a composite pure hypothesis testing problem. We analyze the connecti…
Unified framework controls false discovery rate in bandit multiple testing.
problem Designing adaptive algorithms to identify true discoveries in multiple hypothesis testing.
method Unified modular framework using e-processes for FDR control in arbitrary settings.
result Unified framework ensures FDR control for dependent and simultaneous arm queries.
Develops GLRT for defending against adversarial attacks in hypothesis testing.
problem Adversarial attacks on machine learning models causing misclassification.
method Generalized likelihood ratio test applied to composite hypothesis testing problem.
result GLRT approach yields competitive robustness-accuracy tradeoff under various attacks.
Transformers learn to generalize unseen tasks by composing self-attention layers.
problem How Transformers generalize to unseen, out-of-distribution tasks.
method Examined synthetic examples and pretrained LLMs, focusing on induction heads and latent subspace.
result Transformers can learn hidden rules for unseen tasks by composing self-attention layers, achieving OOD generalization.
We show that the problem of recognizing that a knot diagram represents a specific torus knot, or any torus knot at all, is in the complexity class NP∩co-NP, assuming the generalized Riemann hypothesis. We also show that satellite knot detection is in NP under the same assumption, and t…
Paper proposes kernel-based tests for model misspecification.
problem Determining if a model is misspecified.
method Minimum distance estimators based on MMD and KSD.
result Correct test level maintained without data splitting.
New measure shows how LSTM models compose hierarchical representations.
problem Understanding how LSTM models capture compositional structure in language.
method Novel measure of interdependence between word meanings in LSTM internal gates.
result High interdependence can hurt generalization and reveals hierarchical structure learning.
Many recent works have discussed the propensity, or lack thereof, for emergent languages to exhibit properties of natural languages. A favorite in the literature is learning compositionality. We note that most of those works have focused on communicative bandwidth as being of primary importance. While important, it is …
This paper shows that scientific discovery can be efficiently learned via compositional function trees, reducing the sample complexity.
problem Statistical and computational intractability of scientific discovery via symbolic regression.
method PAC learning approach focusing on compositional function trees built from a finite vocabulary of smooth operators.
result The Rademacher complexity and excess risk are controlled by depth and Lipschitz constants of the base operators, leading to finite-union bounds and high-probability risk bounds.
We extend knot Floer homology to string links in D^{2} \times I and to d-based links in arbitrary three manifolds, without any hypothesis on the null-homology of the components. As for knot Floer homology we obtain a description of the Euler characteristic of the resulting homology groups (in D^{2} \times I) in terms o…
Paper defends machine learning models from adversarial attacks using GLRT.
problem Adversarial attacks on machine learning models leading to misclassification.
method Generalized likelihood ratio test (GLRT) for robust classification.
result GLRT yields performance competitive with minimax approach under worst-case attacks, and better trade-off under weaker attacks.
With model uncertainty characterized by a convex, possibly non-dominated set of probability measures, the agent minimizes the cost of hedging a path dependent contingent claim with given expected success ratio, in a discrete-time, semi-static market of stocks and options. Based on duality results which link quantile he…
Surrogate Data Analysis (SDA) is a statistical hypothesis testing framework for the determination of weak chaos in time series dynamics. Existing SDA procedures do not account properly for the rich structures observed in stock return sequences, attributed to the presence of heteroscedasticity, seasonal effects and outl…
Study excess capacity in neural networks using Rademacher complexity.
problem Understanding how much capacity deep networks have beyond what's needed for classification.
method Unified Rademacher complexity bounds for function composition and convolutional layers, considering Lipschitz constants and initialization norms.
result There is substantial excess capacity per task, and capacity can be kept similar across different tasks.
We introduce a novel approach, requiring only mild assumptions, for the characterization of deep neural networks at initialization. Our approach applies both to fully-connected and convolutional networks and easily incorporates batch normalization and skip-connections. Our key insight is to consider the evolution with …
We propose a robust inferential procedure for assessing uncertainties of parameter estimation in high-dimensional linear models, where the dimension p can grow exponentially fast with the sample size n. Our method combines the de-biasing technique with the composite quantile function to construct an estimator that …
New method improves MMD estimation without convexity assumptions.
problem Lack of theoretical guarantees for MMD estimation algorithms.
method Preconditioned gradient descent (PGD) scheme for MMD optimization.
result PGD scheme converges globally under specific conditions.
A statistical framework for removing unwanted data domains in machine learning.
problem Removing unwanted data domains in machine learning while preserving desired performance.
method Modeling domains as probability distributions and using hypothesis testing to select samples to remove.
result Characterization of allowable edited data distributions and removal-preservation Pareto frontiers for various distribution families.
New framework resolves central limit behavior in differential privacy.
problem Choosing appropriate privacy metrics in hypothesis testing.
method Infinitely divisible limit experiments and Le Cam's theory.
result Characterizes all limiting baseline trade-off functions in differential privacy.
This work is an analytical and numerical study of the composition of several fractals into one and of the relation between the composite dimension and the dimensions of the component fractals. In the case of composition of standard IFS with segments of equal size, the composite dimension can be expressed as a function …
AdaPT-GMM improves multiple testing power with covariates.
problem Powerful and robust multiple testing with covariates.
method Covariate-assisted Gaussian mixture model with adaptive thresholding.
result AdaPT-GMM delivers high power in various scenarios.
New geometric approach for analyzing compositional data like gut microbiomes.
problem Analyzing non-negative compositional data with relative values only.
method Reinterpret compositional data as quotient topology of a sphere, using spherical harmonics and reflection group actions.
result Construction of Reproducing Kernel Hilbert Space (RKHS) for compositional data.
Study on deep neural networks using branching processes and Mehler's formula.
problem Understanding the mathematical role of activation functions in compositional neural networks.
method Connection between compositional kernels and branching processes via Mehler's formula; new random features algorithm.
result Explicit formulas for eigenvalues of compositional kernels quantify complexity.
We establish conditions for compositional generalization in machine learning.
problem Achieving compositional generalization in machine learning models.
method We reformulate compositionality as a property of the data-generating process and derive mild conditions on the training distribution and model architecture.
result Our theoretical framework enables compositional generalization under mild conditions.
This paper proves hyperbolicity of virtual knot compositions.
problem Proving hyperbolicity of virtual knot compositions.
method Exploring the composition of hyperbolic virtual knots.
result Strong lower bounds on the volume of compositions.
Develops methods for causal inference in compositional data using instrumental variables.
problem Interpreting summary statistics like diversity indices as causal effects in compositional data.
method Statistical data transformations and regression techniques tailored for compositional data.
result Advantages and limitations of the proposed methods demonstrated on synthetic and real microbiome data.
In classical field theory, the composite fibred manifolds Y -> Z -> X provides the adequate mathematical formulation of gauge models with broken symmetries, e.g., the gauge gravitation theory. This work is devoted to connections on composite fibred manifolds. In particular, we get the horizontal splitting of the vertic…
New adaptive test for NPIV models controls size and has superior power.
problem Testing inequality and equality restrictions in nonparametric IV models.
method Adaptive hypothesis test based on modified leave-one-out sample quadratic distance.
result Adaptive test attains the adaptive minimax rate of testing in L2. The p-index improves investment performance for NYSE stocks but not for SSE stocks.
problem Improving investment performance for stocks using the p-index.
method Comparing different p-ratio strategies and empirical efficient frontiers for SSE and NYSE stocks.
result The p-index enhances investment performance for NYSE stocks but not for SSE stocks.
This paper shows how to construct sequential tests with power one against weakly compact sets in Polish spaces.
problem Testing composite null hypotheses involving weakly compact sets in Polish spaces.
method Develops sequential tests for i.i.d. laws in Polish spaces, providing a sufficient condition for power one.
result Power-one sequential tests exist for weakly compact sets against their complements in i.i.d. laws in Polish spaces.
Model predicts composite structures assembly quality with input uncertainty.
problem Accurate prediction of dimensional deviations and residual stress in composite structures assembly.
method Neural Network Gaussian Process considering input uncertainty.
result NNGPIU model outperforms other methods for nonsmooth, nonlinear responses.
New algorithm reduces complexity for optimizing complex machine learning tasks.
problem Optimizing complex machine learning objectives like reinforcement learning and portfolio management.
method Developed SARAH-Compositional algorithm using Stochastic Recursive Gradient Descent.
result Achieved optimal IFO complexity bounds for stochastic compositional optimization.
This paper introduces compositional data analysis for financial ratios, improving industry-level analysis.
problem Statistical issues with standard financial ratios at industry level.
method Compositional data analysis techniques for financial ratios.
result Improved analysis of financial ratios using compositional data methods.
Study challenges neural models in compositional learning tasks.
problem Challenges in neural models for compositional and relational learning.
method Introduced ConceptWorld environment for generating images from compositional concepts, tested various neural architectures.
result Neural models struggle with longer compositional chains and substitutivity tests.
SCL discovers compositional structures in analogical reasoning tasks.
problem Discovering compositional structures in analogical reasoning tasks like Raven's Progressive Matrices.
method Proposes Scattering Compositional Learner (SCL) that composes neural networks in sequence.
result Achieves state-of-the-art performance on RPM datasets with significant improvements.
New filters match advanced composition for adaptive privacy, with practical constants.
problem Limitations of existing adaptive composition methods.
method Constructed new filters and odometers that match advanced composition rates, including constants.
result Achieved fully adaptive privacy with practical filters and odometers.
Paper develops momentum schemes with variance reduction for non-convex composition optimization.
problem Lack of convergence guarantee and efficient momentum design in existing algorithms.
method Develops various momentum schemes with SPIDER-based variance reduction.
result Achieves near-optimal sample complexity and linear convergence rate.
Extends knockoff filter for composite null hypotheses in variable selection.
problem Handling composite null hypotheses in variable selection.
method Developed two methods for composite inference with knockoffs: S-OLS and FRPP.
result Proposed heuristic variants of S-OLS outperforming BH procedure for composite nulls.