Adaptive stochastic gradient methods such as AdaGrad have gained popularity in particular for training deep neural networks. The most commonly used and studied variant maintains a diagonal matrix approximation to second order information by accumulating past gradients which are used to tune the step size adaptively. In…
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.
168,742 papers · 148 categories
Trend · papers per month
40 results for “full-matrix”
Adaptive regularization methods pre-multiply a descent direction by a preconditioning matrix. Due to the large number of parameters of machine learning problems, full-matrix preconditioning methods are prohibitively expensive. We show how to modify full-matrix adaptive regularization in order to make it practical and e…
Black-box optimization is primarily important for many compute-intensive applications, including reinforcement learning (RL), robot control, etc. This paper presents a novel theoretical framework for black-box optimization, in which our method performs stochastic update with the implicit natural gradient of an exponent…
Algorithm reduces regret in online learning with varying norms.
problem Online convex optimization with changing norms.
method Adaptive online learning algorithm that adjusts to varying norms without tuning.
result Achieves improved regret bounds for full-matrix AdaGrad.
The adaptive gradient online learning method known as AdaGrad has seen widespread use in the machine learning community in stochastic and adversarial online learning problems and more recently in deep learning methods. The method's full-matrix incarnation offers much better theoretical guarantees and potentially better…
Efficient optimization method reduces Full AdaGrad complexity.
problem Computational challenges in Full AdaGrad for large-scale applications.
method Recursive method and streaming variant for parameter updates.
result Significant reduction in complexity and resource demands.
This work improves knowledge distillation by transferring full kernel matrices efficiently.
problem Efficiently transferring full pairwise similarity matrices for model compression in deep learning.
method The authors propose a method to transfer the full similarity matrix effectively using the Nyström method, decomposing it into partial matrices.
result The difference between the full kernel matrices of teacher and student can be well bounded by partial matrices, improving optimization efficiency.
The Hilbert manifold consisting of positive invertible (unitized) Hilbert-Schmidt operators has a rich structure and geometry. The geometry of unitary orbits is studied from the topological and metric viewpoints: we seek for conditions that ensure the existence of a smooth local structure for the set $…
New framework estimates eigenvalues of kernel matrices without full matrix construction.
problem Estimating eigenvalues of large kernel matrices efficiently.
method Eigenvalue quantile estimation framework for kernel matrices with quick decay.
result Validates framework with empirical evidence and proves interlacing theorem.
We provide an online convex optimization algorithm with regret that interpolates between the regret of an algorithm using an optimal preconditioning matrix and one using a diagonal preconditioning matrix. Our regret bound is never worse than that obtained by diagonal preconditioning, and in certain setting even surpass…
In this paper, we show that the bundle method can be applied to solve semidefinite programming problems with a low rank solution without ever constructing a full matrix. To accomplish this, we use recent results from randomly sketching matrix optimization problems and from the analysis of bundle methods. Under strong d…
Abstract: Determinants and formulas for operators on various spaces.
problem Determinants and formulas for operators on different algebras and spaces.
method Use of Poincaré type determinants, invariant operators, and full matrix-symbols.
result Explicit formulas for determinants of elliptic operators and periodic pseudo-differential operators.
We show that a simple randomized sketch of the matrix multiplicative weight (MMW) update enjoys (in expectation) the same regret bounds as MMW, up to a small constant factor. Unlike MMW, where every step requires full matrix exponentiation, our steps require only a single product of the form , which the Lanczos …
Study on overlaps of singular vectors in Gaussian matrix submatrices.
problem Analyzing overlaps of singular vectors in submatrices of Gaussian matrices.
method Utilizes dynamics of singular vectors and specific resolvents for Brownian trajectories.
result Explicit forms for limiting rescaled mean squared overlaps in the bulk of spectra.
IRCUR accelerates RPCA by using CUR decomposition for efficient low rank estimation.
problem Dimension reduction in robust principal component analysis.
method IRCUR employs CUR decomposition to update the low rank component efficiently.
result IRCUR achieves significant computational efficiency compared to existing algorithms.
Study eigenvector overlaps in large Gaussian matrices, simplifying for GOE.
problem Investigate eigenvector overlaps in large Gaussian matrices.
method Analysis of eigenvector flow under Dyson Brownian motion.
result Explicit computation of limiting rescaled mean squared overlaps.
PACE-GGM uses Gaussian mechanism for private covariance estimation.
problem Private estimation of covariance matrices in high dimensions.
method Data-adaptive selection of entries, Gaussian mechanism, maximum-entropy reconstruction.
result Consistent improvements in estimation error compared to Gaussian mechanism and baselines.
A new model calculates optimal clearing payments in dynamic financial networks.
problem Determining fair clearing payments in networks with potential defaults.
method Extends Eisenberg-Noe model to multiple time periods, solving linear programs for optimal payments.
result Proves the model satisfies the priority of debt claims requirement and finds unique optimal payments.
Clarifies the structure of quantum states using algebraic methods.
problem Unclear stratification of quantum states in physics literature.
method Analyzes the state space S(A) of a finite-dimensional C*-algebra A, focusing on unitary orbits and their properties.
result Identifies a natural Whitney stratification of the state space into matrices of fixed rank, providing a pseudo-manifold structure.
We introduce Independently Recurrent Long Short-term Memory cells: IndyLSTMs. These differ from regular LSTM cells in that the recurrent weights are not modeled as a full matrix, but as a diagonal matrix, i.e.\ the output and state of each LSTM cell depends on the inputs and its own output/state, as opposed to the inpu…
KrADagrad: Kronecker Approximation-Domination Gradient Preconditioned Stochastic Optimizationstat.ML
A new optimization method reduces memory and compute requirements for deep learning.
problem Memory and compute constraints in second-order stochastic optimizers for deep learning.
method Proposes KrAD, a novel factorization to approximate inverse Fisher matrix without inversion, leading to KrADagrad.
result Improves performance over Shampoo for 32-bit precision and comparable/generalization on real datasets.
We describe several algorithms for matrix completion and matrix approximation when only some of its entries are known. The approximation constraint can be any whose approximated solution is known for the full matrix. For low rank approximations, similar algorithms appears recently in the literature under different name…
Adaptive gradient approaches that automatically adjust the learning rate on a per-feature basis have been very popular for training deep networks. This rich class of algorithms includes Adagrad, RMSprop, Adam, and recent extensions. All these algorithms have adopted diagonal matrix adaptation, due to the prohibitive co…
This paper proposes a numerical method for pricing foreign exchange (FX) options in a model which deals with stochastic interest rates and stochastic volatility of the FX rate. The model considers four stochastic drivers, each represented by an Itô's diffusion with time--dependent drift, and with a full matrix of corre…
A new method solves optimization problems on the generalized Stiefel manifold using random estimates of B.
problem Optimization over the generalized Stiefel manifold in applications like CCA, ICA, and GEVP.
method Cheap stochastic iterative method that converges to critical points on the manifold.
result The method achieves the same convergence rates as Riemannian optimization but with lower per-iteration cost.
Theory explains deep nonlinear networks' plateaus and transitions.
problem Understanding long plateaus and feature acquisition transitions in deep nonlinear networks.
method Derived an exact identity for Frobenius norms, classified activation functions, and reduced matrix flow to a scalar ODE.
result Escape time law for deep nonlinear networks, where is the number of bottleneck layers.
CUR matrix decomposition is a randomized algorithm that can efficiently compute the low rank approximation for a given rectangle matrix. One limitation with the existing CUR algorithms is that they require an access to the full matrix A for computing U. In this work, we aim to alleviate this limitation. In particular, …
New method speeds up deep learning optimization.
problem Scalable second-order optimization for deep learning.
method Second-order optimization with algorithmic and numerical improvements.
result Significant convergence and wall-clock time improvements.
New method finds efficient low-rank neural networks during training.
problem High memory and computational demands of neural networks.
method Restricts weight matrices to a low-rank manifold and updates low-rank factors.
result Significantly reduced time and memory resources required for training and evaluation.
This paper investigates Shampoo's heuristics and decouples preconditioner updates.
problem Improving Shampoo's heuristics for training neural networks.
method Decomposing preconditioner updates, correcting eigenvalues, and adapting eigenbasis computation frequency.
result Principled techniques to remove Shampoo's heuristics and improve training algorithms.
Sketchy reduces memory and compute requirements for adaptive regularization in deep learning.
problem Prohibitive memory and running time for adaptive regularization methods in deep learning.
method Low-rank sketching approach using Frequent Directions (FD) to reduce memory and compute requirements.
result Efficient interpolation between resource requirements and degradation in regret guarantees with rank .
Paper develops a bootstrap method for estimating sketched SVD errors.
problem Lack of tools for accurately estimating sketched SVD errors.
method Develops a fully data-driven bootstrap method for numerical error estimation.
result Allows users to adaptively predict extra work needed for desired error tolerance.
MetaGrad adapts multiple learning rates for faster online optimization.
problem Online convex optimization with varying function types.
method MetaGrad uses multiple adaptive learning rates based on empirical performance.
result MetaGrad achieves faster rates for various function types.
Novel algorithm speeds up log-determinant estimation for large matrices.
problem Efficiently estimating log-determinants of large positive definite matrices under memory constraints.
method Hierarchical algorithm based on block-wise computation of LDL decomposition.
result Accurate estimation of NTK log-determinants from a tiny fraction of the full dataset.
DEO uses gradient information to escape saddle points in neural networks.
problem Training deep neural networks struggles with flat regions and saddle points.
method Dimer-Enhanced Optimization (DEO) uses gradient information to estimate curvature and escape saddle points.
result DEO improves training efficiency and performance compared to standard first-order methods.
DOCKSTRING simplifies docking simulations for better drug design benchmarks.
problem Lack of meaningful benchmarks for ligand design.
method Open-source Python package for docking scores, extensive dataset, and pharmaceutically-relevant tasks.
result Docking scores are more appropriate benchmarks than simple physicochemical properties.
cGAP visualizes high-dimensional categorical data with interpretable geometric structure.
problem Lack of visualization tools for high-dimensional categorical data.
method cGAP uses Homogeneity Analysis (HOMALS) to embed data in a 3D space and maps it to colors.
result cGAP reveals coherent clusters, outliers, and local-to-global structure in categorical data.
cGAP visualizes high-dimensional categorical data with interpretable geometric structure.
problem Lack of visualization tools for high-dimensional categorical data.
method cGAP uses Homogeneity Analysis (HOMALS) to embed data in a 3D space and maps it to colors for visualization.
result cGAP reveals coherent clusters, outliers, and local-to-global structure in categorical data.
New method estimates large matrices' spectra from small sub-matrices.
problem Estimating large matrices' spectra when full matrix-vector products are not available.
method Free decompression based on free probability theory.
result Estimates eigenspectrum of impalpable matrices from small sub-matrices.
Large deep neural network (DNN) models pose the key challenge to energy efficiency due to the significantly higher energy consumption of off-chip DRAM accesses than arithmetic or SRAM operations. It motivates the intensive research on model compression with two main approaches. Weight pruning leverages the redundancy i…