Gradient descent converges with arbitrary stepsize for separable data under Fenchel-Young losses.
problem Understanding the conditions under which gradient descent converges with arbitrary stepsize.
method Using Fenchel-Young losses and leveraging the classical perceptron argument to derive convergence rates.
result GD converges with arbitrary stepsize for a majority of Fenchel-Young losses, with better rates for specific loss functions.
The paper sets lower bounds for adversarial robustness in multiclass classification.
problem Adversarial robustness in multiclass classification with arbitrary loss functions.
method Dual and barycentric reformulations for robust risk minimization.
result Sharp lower bounds for adversarial risks are computed efficiently.
New algorithm handles bandit problems under translations and scales.
problem Adversarial multi-armed bandit problems with arbitrary translations and scales.
method Innovative online algorithm invariant to translations and scales, using universal prediction.
result Second-order regret bounds, unaffected by affine transformations of losses.
In this work we develop a new algorithm for regularized empirical risk minimization. Our method extends recent techniques of Shalev-Shwartz [02/2015], which enable a dual-free analysis of SDCA, to arbitrary mini-batching schemes. Moreover, our method is able to better utilize the information in the data defining the ER…
DDR method improves regression performance by predicting arbitrary quantiles.
problem Traditional regression methods produce biased mean predictions and lack robustness.
method Deep Distribution Regression (DDR) method that estimates arbitrary quantiles.
result DDR method outperforms traditional methods in mean and quantile prediction.
We consider deep linear networks with arbitrary convex differentiable loss. We provide a short and elementary proof of the fact that all local minima are global minima if the hidden layers are either 1) at least as wide as the input layer, or 2) at least as wide as the output layer. This result is the strongest possibl…
This paper introduces a new learning method for classification without true labels.
problem Learning with only complementary labels, not true class labels.
method Derives a novel framework for arbitrary losses and models, using unbiased risk estimation.
result Demonstrates improved risk estimator through non-negative correction and gradient ascent.
New algorithm tackles adversarial bandits with arbitrary strategies.
problem Adversarial bandit problem against arbitrary strategies.
method Adopted master-base framework using online mirror descent method (OMD). Proposed adaptive learning rates for OMD.
result Achieved improved regret bounds compared to previous methods.
Deep neural networks' loss surfaces contain every low-dimensional pattern.
problem Finding arbitrary low-dimensional patterns in neural network loss surfaces.
method Empirical and theoretical analysis of loss landscapes of deep neural networks.
result Deep universal approximators exhibit a property where arbitrary smooth patterns exist in their loss surfaces.
New algorithm tackles multi-armed bandit with arbitrary delays and general bounded losses.
problem Scale-free adversarial multi-armed bandit with arbitrary feedback delays.
method SFD-INF combines convex combination trick and doubling/skipping technique.
result Achieves adaptive regret bounds for non-negative and general scale-free losses.
Study predicts SGD test loss for structured features.
problem Understanding test loss dynamics in SGD for structured data.
method Solveable model of SGD on mean square loss for arbitrary covariance structure.
result Simpler Gaussian model accurately predicts test loss of nonlinear models.
Algorithm learns arbitrary ReLU neurons under Gaussian inputs.
problem Learn an arbitrary ReLU activation over Gaussian marginals.
method Statistical Query (SQ) algorithm that outputs a ReLU activation achieving O ( O P T ) + ε O(\mathrm{OPT}) + \varepsilon O ( OPT ) + ε loss. result First constant factor approximation for arbitrary bias in polynomial time.
Proposes a new framework for learning from labeled and unlabeled data.
problem Learning from unlabeled and multi-label samples with arbitrary loss functions.
method Multi-complementary and unlabeled learning framework.
result Effective estimation of classification risk with optimal convergence rate.
Two new algorithms improve performance in adversarial bandits with unbounded losses.
problem Adversarial Multi-Armed Bandits with unbounded losses.
method Developed UMAB-NN and UMAB-G for non-negative and general unbounded losses respectively.
result UMAB-NN achieves the first adaptive and scale-free regret bound for non-negative unbounded losses.
Most systems and learning algorithms optimize average performance or average loss -- one reason being computational complexity. However, many objectives of practical interest are more complex than simply average loss. This arises, for example, when balancing performance or loss with fairness across people. We prove tha…
We present surrogate regret bounds for arbitrary surrogate losses in the context of binary classification with label-dependent costs. Such bounds relate a classifier's risk, assessed with respect to a surrogate loss, to its cost-sensitive classification risk. Two approaches to surrogate regret bounds are developed. The…
New research shows many batch selection methods for training work just as well as full batch training.
problem Finding optimal batch selection methods for training.
method Analysis of mini-batch Gradient Descent (GD) and Stochastic GD (SGD) with various batch selection rules.
result All mini-batch schedules, including deterministic ones, generalize optimally for smooth Lipschitz-convex/nonconvex/strongly-convex loss functions.
Piecewise linear activations create many spurious local minima in neural networks.
problem Understanding the loss surface of neural networks with piecewise linear activations.
method Proved the existence of infinite spurious local minima and partitioned the loss surface into smooth cells.
result Piecewise linear activations create many spurious local minima that are invariant under a continuous path.
This paper calculates worst-case target semi-variances for uncertain losses.
problem Managing risk when loss distribution is uncertain and only partial information is known.
method Derives worst-case target semi-variances for symmetric or non-negative losses under uncertainty sets representing investor's undesirable scenarios.
result Closed-form expressions for worst-case target semi-variances are derived.
Global implicit function theorem for Fréchet spaces, solving derivative loss problems.
problem Solving initial value problems with derivative loss in Fréchet spaces.
method Global implicit function theorems for Keller's C c 1 C_c^1 C c 1 -mappings in Fréchet spaces, applied through submersions and transversality. result Global existence and uniqueness of solutions to initial value problems with derivative loss.
Easyllp simplifies LLP, achieving low task loss at individual instance level.
problem Weakly supervised classification with label proportions.
method Flexible debiasing approach based on aggregate labels, operating on arbitrary loss functions.
result Accurately estimates expected loss at individual level, with provable guarantees.
The study establishes SQ lower bounds for learning halfspaces and ReLUs under Gaussian marginals.
problem Agnostically learning halfspaces and ReLUs under Gaussian marginals.
method Statistical Query (SQ) lower bounds analysis.
result Proves SQ lower bounds of d p o l y ( 1 / ε ) d^{\mathrm{poly}(1/ε)} d poly ( 1/ ε ) for both problems. Flexible tree ensemble learning framework supports arbitrary loss functions and multi-task learning.
problem Limited modeling capabilities of existing tree ensemble learning toolkits.
method Differentiable tree ensembles with tensor-based formulation for efficient training.
result Our framework leads to 100x more compact and 23% more expressive tree ensembles.
We study proper losses for discrete generative models without knowing the target distribution.
problem Evaluating generative models in the discrete setting without direct access to the target distribution.
method Define and construct black-box proper losses using statistical estimation theory.
result Black-box proper losses must be of polynomial form and involve more samples than the polynomial degree.
Gradient descent implicitly follows regularization for general losses.
problem The implicit bias of gradient descent methods in machine learning.
method Empirical risk minimization over linear predictors with arbitrary convex, strictly decreasing losses.
result Gradient descent and regularization paths converge to the same direction for non-attained risks.
A novel gradient-based method optimizes decision trees for complex tasks.
problem Training decision trees with arbitrary differentiable loss functions.
method Gradient-based optimization using first and second derivatives of loss functions.
result Improves accuracy and flexibility in decision tree optimization.
Fine-grained analysis of gradient descent with momentum provides modified loss equations.
problem Understanding the dynamics of gradient descent with momentum.
method Fine-grained analysis and derivation of modified loss equations.
result Global approximation bounds and continuous modified equations for HB.
New bounds on minimax regret for sequential probability assignment using logarithmic loss.
problem Minimizing regret in sequential probability assignment against arbitrary experts.
method Using self-concordance property of logarithmic loss to derive tight bounds.
result Tight bounds on minimax regret for various expert classes.
ALIS uses probabilistic sampling to minimize true loss in active learning.
problem Efficiently choosing data points to label in active learning.
method Proposes ALIS algorithm with optimal sampling distribution.
result Derives upper bounds on true loss for any sampling procedure.
In this paper we consider Fourier transform techniques to efficiently compute the Value-at-Risk and the Conditional Value-at-Risk of an arbitrary loss random variable, characterized by having a computable generalized characteristic function. We exploit the property of these risk measures of being the solution of an ele…
Unified approach for robust low rank matrix estimation with adversaries.
problem Robust low rank matrix estimation in the presence of adversaries.
method Unified approach combining Huber loss and nuclear norm penalization.
result Sharp estimation error bounds for matrix compressed sensing and completion.
Reduces dynamic regret to static problem in RKHS.
problem Minimizing cumulative loss in online convex optimization.
method Reduces dynamic regret to static regret problem in RKHS.
result Optimal dynamic regret guarantees for linear losses and new bounds for exp-concave and improper linear regression.
The paper decomposes probabilistic scores into reliability, uncertainty, and information loss.
problem Understanding the reliability and uncertainty of probabilistic predictions.
method Developed decomposition identities for proper losses, quantifying reliability, residual uncertainty, and information gain.
result A three-term identity for classification scores, revealing miscalibration, grouping term, and feature-level uncertainty.
We consider on-line density estimation with a parameterized density from the exponential family. The on-line algorithm receives one example at a time and maintains a parameter that is essentially an average of the past examples. After receiving an example the algorithm incurs a loss which is the negative log-likelihood…
This paper introduces SRPR for robust phase retrieval with smoothed loss functions.
problem Robust phase retrieval from noisy quadratic measurements with corruptions.
method Smoothed robust phase retrieval (SRPR) using convolution-type smoothed loss functions.
result SRPR has no spurious local solutions and benign landscape under corruptions.
Online SGD from random init solves non-smooth, non-convex phase retrieval.
problem Solving phase retrieval with non-smooth, non-convex loss functions.
method Online stochastic gradient descent (SGD) with constant step size, starting from arbitrary initialization.
result SGD converges from arbitrary initializations for the amplitude squared loss objective.
Active learning is a type of sequential design for supervised machine learning, in which the learning algorithm sequentially requests the labels of selected instances from a large pool of unlabeled data points. The objective is to produce a classifier of relatively low risk, as measured under the 0-1 loss, ideally usin…
SGD converges globally to logistic loss minima for two-layer nets.
problem Global convergence of SGD for logistic loss on two-layer neural nets.
method Demonstrates existence of Frobenius norm regularized logistic loss functions as Villani functions, proving convergence and exponential rate.
result SGD converges globally to the global minima of appropriately regularized logistic empirical risk of depth 2 nets.
New algorithm minimizes cumulative loss in dynamic linear bandits without prior knowledge of comparator switches.
problem Minimizing cumulative loss in dynamic linear bandits with unknown number of switches.
method Combining several bandit algorithms to adapt to unknown number of switches without prior knowledge.
result First algorithm achieving optimal regret guarantee of O ( d ( 1 + S T ) T ) \mathcal{O}\big(\sqrt{d(1+S_T) T}\big) O ( d ( 1 + S T ) T ) up to poly-logarithmic terms. This paper establishes minimax rates for online regression with arbitrary classes of functions and general losses. We show that below a certain threshold for the complexity of the function class, the minimax rates depend on both the curvature of the loss function and the sequential complexities of the class. Above this…
Dynamic regret minimization is shown equivalent to static regret minimization for linear losses.
problem Dynamic regret minimization in online convex optimization.
method Equivalence between dynamic and static regret minimization for linear losses.
result Dynamic regret minimization is equivalent to static regret minimization for linear losses.
Extends loss function analysis to infinite dimensions for better machine learning.
problem Difficulty in separating machine learning problem study from loss function properties.
method Generalizes proper-composite representation to infinite dimensions, characterizing canonical link.
result Simple characterisation of canonical link in infinite dimensional setting.
We establish linear regret bounds for convex smooth losses using Fenchel-Young losses.
problem Establishing linear regret bounds for convex smooth losses.
method Constructing a convex smooth surrogate loss using Fenchel-Young losses generated by the convolutional negentropy.
result We derive a smooth loss with a linear surrogate regret bound.
The paper analyzes the implicit bias of SGD near loss manifold and provides new insights.
problem Understanding the implicit bias of SGD near loss manifolds in overparametrized models.
method Adapting ideas from Katzenberger (1991) to analyze SGD dynamics using a stochastic differential equation (SDE).
result SGD with label noise locally decreases the sharpness of loss, leading to a global analysis of implicit bias.
Study ruin probabilities in risk processes on stochastic networks.
problem Ruin probabilities in risk processes on stochastic networks.
method Classification of agents by types, Poisson process for loss propagation, explicit ruin probabilities for infinite network size.
result Explicit ruin probabilities for agents of any type in infinite network size.
New method trains generative models without discriminators, improving stability and accuracy.
problem Training implicit generative models with adversarial discriminators leads to instability and mode-dropping.
method Invariant statistical loss function, avoiding discriminators.
result Successfully trains generative models for various complex distributions without mode-dropping.
Efficient online learning with pairwise loss functions is a crucial component in building large-scale learning system that maximizes the area under the Receiver Operator Characteristic (ROC) curve. In this paper we investigate the generalization performance of online learning algorithms with pairwise loss functions. We…
This work studies applications and generalizations of a simple estimation technique that provides exponential concentration under heavy-tailed distributions, assuming only bounded low-order moments. We show that the technique can be used for approximate minimization of smooth and strongly convex losses, and specificall…