Minimum Description Length prevents overfitting in noisy data.
problem Learning from noisy data with overfitting risk.
method Minimum Description Length learning rule with tempered guarantees.
result Tempered agnostic finite sample learning guarantees and asymptotic behavior characterization.
Paper provides a performance guarantee for spectral clustering.
problem Finding the global solution to the minimum ratio cut problem.
method Two-step spectral clustering method with a rounding step, analyzed using two-to-infinity norm perturbation bounds.
result Spectral clustering is guaranteed to output the global solution under certain conditions.
Optimizes investment and consumption for post-retirement with minimum guarantee.
problem Maximizing final annuity with minimum guarantee during decumulation phase.
method Dynamic programming via Hamilton-Jacobi-Bellman (HJB) equation, finite difference method.
result Existence and uniqueness of classical solutions proved through dual transformation.
This paper improves indoor positioning accuracy by deploying reference nodes to ensure Line-of-Sight.
problem Systematic bias errors in indoor positioning due to non-LoS propagation.
method Model indoor service area as a graph, partition into cliques for reference nodes, set minimum distance and angle parameters.
result Guaranteed LoS to reference nodes improves indoor positioning accuracy and precision.
Study of participating policies with guaranteed minimum interest rate and surrender option.
problem Analyzing the value and optimal surrender strategy of participating policies with minimum interest rate guarantee and surrender option.
method Probabilistic analysis using optimal stopping and free boundary theory.
result Identification of an optimal surrender strategy involving stop-loss and too-good-to-persist boundaries.
New framework for DNN training guarantees convergence to global minimum.
problem Training deep neural networks to converge to global minimum.
method Reformulated minimization problem with recursive algorithmic framework, using bounded style assumptions.
result Convergence to an ε-(global) minimum with O(1/ε^3) gradient computations.
Combines CPPI and option-based strategy to ensure equity exposure.
problem Cash-in risk and equity market participation in CPPI.
method Two-step strategy: CPPI with guaranteed minimum equity exposure.
result Shows effectiveness through numerical analysis of option prices.
Research examines GMIB and reset options in variable annuities.
problem Understanding the value and rationality of GMIB and reset options.
method Exploration of various parameters affecting GMIB value and calculation of critical future interest rates for reset option rationality.
result Insight into how future market performance and interest rates influence policyholder and insurer actions.
Extends conformal prediction to contrastive learning for better coverage of positive samples.
problem Lack of principled guarantees on coverage in contrastive learning.
method Introduces minimum-volume covering sets with learnable constraints.
result Improves inclusion-exclusion trade-offs in positive and negative samples.
New guarantees for adaptive combinatorial maximization with various objectives.
problem Maximizing under cardinality constraints and minimum cost coverage in adaptive settings.
method Bayesian approach with comprehensive approximation guarantees for various utility functions.
result Maximal gain ratio is a new parameter that provides stronger approximation guarantees than greedy policies.
Proposes a new risk measurement method for risk-averse stochastic optimization.
problem Risk-averse stochastic optimization problems.
method Develops a risk measure based on argmin and minimum concepts.
result Guarantees the existence of solutions for the proposed problem.
Study shows unusual non-monotonic risk behavior in minimum-norm interpolants for various data scaling.
problem Understanding the risk behavior of minimum-norm interpolants in RKHS for different data scaling.
method Analysis of spectral properties of the random kernel matrix restricted to eigen-spaces of the population covariance operator.
result Minimum-norm interpolants in RKHS exhibit multiple descent in risk for d=nα with α∈(0,1). We design a non-convex second-order optimization algorithm that is guaranteed to return an approximate local minimum in time which scales linearly in the underlying dimension and the number of training examples. The time complexity of our algorithm to find an approximate local minimum is even faster than that of gradie…
New sampling method optimizes learning minimum mean among distributions.
problem Learning the minimum mean from a set of distributions.
method Developed Murphy Sampling, a novel approach.
result Murphy Sampling optimizes learning both low and high true minimums.
In this paper we present a numerical valuation of variable annuities with combined Guaranteed Minimum Withdrawal Benefit (GMWB) and Guaranteed Minimum Death Benefit (GMDB) under optimal policyholder behaviour solved as an optimal stochastic control problem. This product simultaneously deals with financial risk, mortali…
In this paper we explore an identity in distribution of hitting times of a finite variation process (Yor's process) and a diffusion process (geometric Brownian motion with affine drift), which arise from various applications in financial mathematics. As a result, we provide analytical solutions to the fair charge of va…
Paper establishes generalization bounds for representation learning using Minimum Description Length.
problem Designing efficient statistical supervised learning algorithms that generalize well to unseen data.
method Developed a compressibility framework using Minimum Description Length (MDL) to derive upper bounds on generalization error.
result Established the first theoretical generalization bounds for Information Bottleneck type encoders and representation learning.
Gradient noise improves privacy-protected optimization performance.
problem Improving privacy in convex optimization while maintaining utility.
method We analyze the effect of gradient perturbation on differentially private convex optimization, focusing on expected curvature.
result Gradient perturbation can achieve a significantly improved utility guarantee for differentially private convex optimization.
New NTK bounds show deep networks with minimum over-parameterization can still memorize and optimize.
problem Understanding memorization and optimization in sub-linear over-parameterized deep networks.
method Lower bound on NTK eigenvalues for deep networks with minimum over-parameterization.
result Deep networks with minimum over-parameterization can still be powerful memorizers and optimizers.
New method for NMF without tuning parameter.
problem Finding latent structures in noisy data matrices.
method Inspired by square-root lasso, proposes a tuning-free minimum-volume NMF.
result Optimal tuning parameter value is noise level-independent.
We analyze stochastic gradient descent for optimizing non-convex functions. In many cases for non-convex functions the goal is to find a reasonable local minimum, and the main concern is that gradient updates are trapped in saddle points. In this paper we identify strict saddle property for non-convex problem that allo…
Uniform convergence of interpolators proven for Gaussian data.
problem Interpolation learning in high-dimensional linear regression with Gaussian data.
method Generic uniform convergence guarantee in terms of Gaussian width.
result Consistency of interpolators for minimum-norm and near-minimal-norm cases.
Efficiently finds sparse solutions to max-plus equations for convex regression.
problem Finding sparse solutions to max-plus equations for convex multivariate regression.
method Polynomial-time algorithm for sparse approximate solutions.
result Optimal piecewise-linear fitting with minimum number of regions.
Combines online learning algorithms to achieve better performance.
problem Improving online learning algorithms with varying guarantees.
method Adding iterates of two parameter-free algorithms to create a new algorithm with improved regret.
result Generates efficient algorithms that adapt to multiple norms and maintain dimension-free guarantees.
Study shows unique sharp local minimum in ℓ1-minimization for dictionary learning.
problem Global recovery of a dictionary from random linear combinations of atoms.
method Norm condition, explicit bound, perturbation-based test, Block Coordinate Descent algorithm.
result Reference dictionary is the unique sharp local minimum of the ℓ1 objective function. Efficient private algorithms for estimating block models and mixture models.
problem Estimating block models and mixture models in high-dimensional settings.
method General tools for designing efficient private estimation algorithms.
result First efficient private algorithms for weak and exact recovery of stochastic block models.
This paper proves IRM minimizes o.o.d. risk under certain conditions.
problem Deep networks can fail to generalize to new domains with different distributions.
method Proves IRM minimizes o.o.d. risk through a bi-level optimization problem.
result IRM minimizes o.o.d. risk under specific conditions.
We use smoothed analysis techniques to provide guarantees on the training loss of Multilayer Neural Networks (MNNs) at differentiable local minima. Specifically, we examine MNNs with piecewise linear activation functions, quadratic loss and a single output, under mild over-parametrization. We prove that for a MNN with …
Paper assesses GMMB in VAs using FST for accurate net liability calculations.
problem Risk management of GMMB under stochastic mortality and regime-switching.
method Net liability model with FST algorithm for accurate numeric solutions.
result FST algorithm provides reliable results for net liability of GMMB.
A new method selects penalties for high-dimensional models using the MDL principle.
problem Selecting optimal penalties for high-dimensional regularization models.
method MDL-RS method that minimizes a tight upper bound of LNML in high-dimensional spaces.
result Improves generalization performance of regularized estimates, especially with redundant parameters.
Develops a new weighted Laplacian method for graph problems.
problem Graph partitioning and balanced minimum cut problems.
method Weighted Laplacian method based on graph theory and PDEs.
result Established equivalence relations among graph problems.
In this paper, we review pricing of variable annuity living and death guarantees offered to retail investors in many countries. Investors purchase these products to take advantage of market growth and protect savings. We present pricing of these products via an optimal stochastic control framework, and review the exist…
PAC-Bayesian theory applied to learning optimization algorithms with generalization guarantees.
problem Learning optimization algorithms with provable generalization guarantees and explicit trade-offs.
method PAC-Bayes theory applied to learning-to-optimize, reformulating the learning procedure into a one-dimensional minimization problem.
result Learned optimization algorithms outperform deterministic worst-case analysis algorithms, even in the limit case of guaranteed convergence.
Study optimal investment strategy for pension schemes to hedge longevity risk.
problem Hedging longevity risk in defined contribution pension schemes.
method Transformed optimal investment problem into an unconstrained problem using dynamic programming and numerical studies.
result Longevity risk significantly impacts investment strategies, supporting the use of mortality-linked securities.
Variable annuities (VA) are popular insurance products. VAs provides the insured with a guaranteed accumulation rate on their premium at maturity. In addition, the insured may receive extra benefit if returns of underlying funds are high enough. Here we consider a special case of VA with high-water mark feature and Gua…
We propose a sampling scheme suitable for reducing a data set prior to selecting a hypothesis with minimum empirical risk. The sampling only considers a subset of the ultimate (unknown) hypothesis set, but can nonetheless guarantee that the final excess risk will compare favorably with utilizing the entire original dat…
Study on inflection points of plane curve shadows with fixed embedded shapes.
problem Minimum number of inflection points in plane curves with fixed embedded shadows.
method Finite coorientation problem on building polygons, dynamic programming, universal lower bound, tree-necklace shadows.
result Exact formula for minimum number of normalized inflections for tree-like shadows.
This paper proves adding neurons eliminates all bad local minima in deep learning.
problem Eliminating all suboptimal local minima in deep learning models.
method Adding one special neuron per output unit eliminates all suboptimal local minima of deep neural networks.
result At every local minimum, the original neural network parameters are a global minimum.
In this paper, we study the price of Variable Annuity Guarantees, especially of Guaranteed Annuity Options (GAO) and Guaranteed Minimum Income Benefit (GMIB), and this in the settings of a derivative pricing model where the underlying spot (the fund) is locally governed by a geometric Brownian motion with local volatil…
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 large collection of financial contracts offering guaranteed minimum benefits are often posed as control problems, in which at any point in the solution domain, a control is able to take any one of an uncountable number of values from the admissible set. Often, such contracts specify that the holder exert control at a…
A new robust PCA estimator combining M-estimators and minimum divergence estimators.
problem Adverse effect of outlying observations in PCA for high-dimensional data.
method Minimum density power divergence estimator combined with a computationally efficient algorithm.
result High breakdown guarantee regardless of data dimension with theoretical support and practical applications.
We give two provably accurate feature-selection techniques for the linear SVM. The algorithms run in deterministic and randomized time respectively. Our algorithms can be used in an unsupervised or supervised setting. The supervised approach is based on sampling features from support vectors. We prove that the margin i…
AdaLoss optimizes adaptive learning rates for efficient convergence in various models.
problem Efficiently optimizing adaptive learning rates for gradient descent methods.
method AdaLoss uses loss function information to dynamically adjust step sizes.
result AdaLoss achieves linear convergence in linear regression and robust global convergence in neural networks.
Symmetric nonnegative matrix factorization (SymNMF) has important applications in data analytics problems such as document clustering, community detection and image segmentation. In this paper, we propose a novel nonconvex variable splitting method for solving SymNMF. The proposed algorithm is guaranteed to converge to…
Local search heuristics for non-convex optimizations are popular in applied machine learning. However, in general it is hard to guarantee that such algorithms even converge to a local minimum, due to the existence of complicated saddle point structures in high dimensions. Many functions have degenerate saddle points su…
New PAC-Bayesian bounds for deep networks without stochastic or compressed parameters.
problem Generalization of overparameterized deep networks.
method PAC-Bayesian framework that leverages noise-resilience of flat minima.
result Generalization guarantee for deterministic, uncompressed networks.
Nonparametric detection of existence of an anomalous structure over a network is investigated. Nodes corresponding to the anomalous structure (if one exists) receive samples generated by a distribution q, which is different from a distribution p generating samples for other nodes. If an anomalous structure does not exi…