We consider the behavior of gradient flow and of discrete and noisy gradient descent. It is commonly noted that the addition of noise to the process of discrete gradient descent can affect the trajectory of gradient descent. In previous work, we observed such effects. There, we considered the case where the minima had …
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.
168,695 papers · 148 categories
Trend · papers per month
48 results for “gradient behavior”
We study how the behavior of deep policy gradient algorithms reflects the conceptual framework motivating their development. To this end, we propose a fine-grained analysis of state-of-the-art methods based on key elements of this framework: gradient estimation, value prediction, and optimization landscapes. Our result…
Study on Einstein solitons with bounds and asymptotic behavior.
problem Understanding the properties of Einstein solitons.
method Computed lower bounds for scalar curvature, established asymptotic behavior, proved finiteness of fundamental group and weighted volume.
result Established finiteness of fundamental group and weighted volume for gradient shrinking Einstein solitons.
We introduce a new approach for comparing reinforcement learning policies, using Wasserstein distances (WDs) in a newly defined latent behavioral space. We show that by utilizing the dual formulation of the WD, we can learn score functions over policy behaviors that can in turn be used to lead policy optimization towar…
New example of non-Kähler soliton with Kähler-like behavior at infinity.
problem Constructing non-Kähler expanding gradient Ricci solitons.
method Asymptotically conical (AC) construction with Kähler tangent cone at infinity.
result Example of a non-Kähler soliton with a Kähler-like behavior at infinity.
This paper investigates asymptotic behaviors of gradient descent algorithms (particularly accelerated gradient descent and stochastic gradient descent) in the context of stochastic optimization arising in statistics and machine learning where objective functions are estimated from available data. We show that these alg…
The study examines 4D steady gradient Ricci solitons with nonnegative curvature away from a compact set.
problem Analyzing noncompact steady gradient Ricci solitons with nonnegative curvature operator.
method Examining the asymptotic behavior of noncompact κ-noncollapsed steady gradient Ricci solitons with nonnegative curvature operator away from a compact set.
result 4D noncompact κ-noncollapsed steady gradient Ricci solitons with nonnegative sectional curvature must be a Bryant Ricci soliton up to scaling.
Wide networks with polynomial activations have proven asymptotic behavior.
problem Understanding the behavior of neural networks in the large width limit.
method Proving a conjecture for deep networks with polynomial activation functions.
result Tight bounds on the behavior of wide networks during stochastic gradient descent and derivation of their finite-width dynamics.
In this note, we observe the behavior of gradient flow and discrete and noisy gradient descent in some simple settings. It is commonly noted that addition of noise to gradient descent can affect the trajectory of gradient descent. Here, we run some computer experiments for gradient descent on some simple functions, and…
The paper studies Möbius energy gradient of helix pairs and finds limiting behavior as coiling ratio increases.
problem Characterizing the limiting behavior of Möbius energy gradient for symmetric helix pairs.
method Complex asymptotics
result The gradient diverges in opposing directions based on radius, approaching 1/2 as coiling ratio increases.
The asymptotic behavior of the stochastic gradient algorithm with a biased gradient estimator is analyzed. Relying on arguments based on the dynamic system theory (chain-recurrence) and the differential geometry (Yomdin theorem and Lojasiewicz inequality), tight bounds on the asymptotic bias of the iterates generated b…
Stochastic gradient descent (SGD) has been widely used in machine learning due to its computational efficiency and favorable generalization properties. Recently, it has been empirically demonstrated that the gradient noise in several deep learning settings admits a non-Gaussian, heavy-tailed behavior. This suggests tha…
This paper presents a stochastic behavior analysis of a kernel-based stochastic restricted-gradient descent method. The restricted gradient gives a steepest ascent direction within the so-called dictionary subspace. The analysis provides the transient and steady state performance in the mean squared error criterion. It…
Gradient descent on neural nets often operates at the Edge of Stability, where loss behavior is complex but loss decreases over time.
problem Understanding the optimization dynamics of neural networks at the Edge of Stability.
method Empirical demonstration of gradient descent behavior in neural network training.
result Gradient descent on neural networks typically occurs at the Edge of Stability, where loss behavior is non-monotonic but loss decreases over time.
Study on asymptotic behavior of Taub-NUT type solitons and construction of new ALF Calabi-Yau metrics.
problem Asymptotic behavior of steady gradient Kähler-Ricci solitons of Taub-NUT type.
method Determination of asymptotic cone, special case analysis, and construction of new metrics using Tian-Yau-Hein method.
result Construction of new ALF Calabi-Yau metrics on quotients of the Taub-NUT type soliton.
Proposes a continuous flow model to understand and control instability in gradient descent for deep learning.
problem Understanding and controlling the instability of gradient descent in deep learning.
method Introduces the Principal Flow (PF), a continuous time flow that approximates gradient descent dynamics.
result The PF captures divergent and oscillatory behaviors of gradient descent, including escaping local minima and saddle points.
We prove that the evolution of weight vectors in online gradient descent can encode arbitrary polynomial-space computations, even in very simple learning settings. Our results imply that, under weak complexity-theoretic assumptions, it is impossible to reason efficiently about the fine-grained behavior of online gradie…
Random feature model shows slow self-correction of generalization gap.
problem Slow deterioration of generalization error in random feature model.
method Examined the dynamic behavior of gradient descent in the model's resonance regime.
result Gradient descent exhibits a self-correction mechanism, reducing generalization gap over time.
The paper studies stochastic gradient descent with infinite variance gradients.
problem Theoretical properties of SGD with infinite variance gradients.
method Establish asymptotic behavior of SGD with infinite variance gradients.
result Asymptotic distribution of SGD is characterized as a stationary distribution of an Ornstein-Uhlenbeck process driven by a stable Lévy process.
Understanding the behavior of stochastic gradient descent (SGD) in the context of deep neural networks has raised lots of concerns recently. Along this line, we study a general form of gradient based optimization dynamics with unbiased noise, which unifies SGD and standard Langevin dynamics. Through investigating this …
Gradient descent dynamics in neural networks show quenching and activation phases.
problem Understanding training dynamics in neural networks.
method Numerical and phenomenological study of gradient descent algorithm for two-layer neural networks.
result Gradient descent dynamics exhibit quenching and activation phases in under-parametrized networks.
Gradient descent fails to train two-layer ReLU networks, leading to poor performance.
problem Gradient descent training of two-layer ReLU networks initialized by He et al. (2015) fails to find optimal solutions.
method Gradient descent on a least-squares loss for training two-layer (Leaky)ReLU networks.
result Gradient descent only finds bad local minima, leading to linear regression for non-linear target functions.
Riemannian gradient descent escapes some spurious critical points on low-rank matrix manifold.
problem Spurious critical points on the boundary of low-rank matrix manifold.
method Riemannian gradient descent with dynamical low-rank approximation and rescaled gradient flow.
result Riemannian gradient descent escapes some spurious critical points on the boundary of the manifold.
In this paper we propose a novel gradient algorithm to learn a policy from an expert's observed behavior assuming that the expert behaves optimally with respect to some unknown reward function of a Markovian Decision Problem. The algorithm's aim is to find a reward function such that the resulting optimal policy matche…
Study compares dropout and l2 regularization in linear models.
problem Understanding the statistical behavior of dropout and l2 regularization in linear models.
method Derives non-asymptotic bounds for gradient descent iterates with dropout and compares them to l2 regularization.
result Indicates a more subtle relationship between dropout and l2 regularization, highlighting interactions between dynamics and randomness.
Infinitesimal boosting converges to a deterministic process in large sample limit.
problem Characterizing the asymptotic behavior of infinitesimal gradient boosting in large sample sizes.
method Proving convergence to a deterministic process using large sample theory and differential equations.
result The test error decreases over time in the population limit.
We study the problem of off-policy policy optimization in Markov decision processes, and develop a novel off-policy policy gradient method. Prior off-policy policy gradient approaches have generally ignored the mismatch between the distribution of states visited under the behavior policy used to collect data, and what …
Study reveals biases in gradient descent for GLNs, improving neural network performance.
problem Understanding and improving the inductive biases of deep neural networks.
method Derive infinite-time training limit of gated linear networks and generalize to other networks.
result Theoretical framework captures key inductive biases of ReLU networks.
Deterministic GD can behave stochastically in large learning rates for multiscale functions.
problem Understanding deterministic GD's stochastic behavior in large learning rates for multiscale objectives.
method Established a sufficient condition for deterministic GD to converge to a rescaled Gibbs distribution in large learning rates for multiscale functions.
result Deterministic GD can converge to a statistical distribution in large learning rates for multiscale functions.
The study examines the dynamic behavior of RMSprop and Adam algorithms.
problem Understanding the training loss curve patterns of adaptive gradient algorithms.
method Careful numerical experiments and theoretical explanations using the signGD flow.
result Adam converges smoother and faster when momentum factors are close to each other.
SAD Neural Networks: Divergent Gradient Flows and Asymptotic Optimality via o-minimal Structurescs.LG
Gradient flows of neural networks converge to optimal values or diverge, with thresholds and asymptotic behaviors.
problem Understanding the convergence and divergence of gradient flows in neural networks.
method Analysis of gradient flows on loss landscapes of neural networks using o-minimal structures.
result Gradient flows either converge to optimal values or diverge to infinity, with thresholds and asymptotic behaviors.
3D steady gradient Ricci solitons are all O(2)-symmetric.
problem Characterizing 3D steady gradient Ricci solitons.
method Analyzing asymptotic behavior and using O(2) symmetry.
result All 3D steady gradient Ricci solitons are O(2)-symmetric.
Training generative adversarial networks (GANs) often suffers from cyclic behaviors of iterates. Based on a simple intuition that the direction of centripetal acceleration of an object moving in uniform circular motion is toward the center of the circle, we present the Simultaneous Centripetal Acceleration (SCA) method…
This paper uses SDEs to analyze GANs training and long-run behavior.
problem Understanding the training process and long-run behavior of GANs.
method Established SDE approximations for GANs training and analyzed long-run behavior via invariant measures.
result The long-run behavior of GANs training can be studied via the invariant measures of its SDE approximations.
The paper explores how structured representations influence learning dynamics in neural networks.
problem Understanding the training dynamics of deep neural networks.
method Investigates a family of enriched transformation layers with constrained pathways and adaptive corrections.
result Improved robustness, smoother optimization, and scalable depth behavior are achieved through structured representations.
We survey some of the recent progress on complete gradient shrinking Ricci solitons, including the classifications in dimension three and asymptotic behavior of potential functions as well as volume growths of geodesic balls in higher dimensions. This article is written for the conference proceedings dedicated to Yau's…
In this paper, we study two kind of L^2 norm preserved non-local heat flows on closed manifolds. We first study the global existence, stability and asymptotic behavior to such non-local heat flows. Next we give the gradient estimates of positive solutions to these heat flows.
MaxVA improves Adam's step sizes by maximizing gradient variance.
problem Unstable or extreme adaptive learning rates in Adam.
method Maximizing the variance of gradient coordinates in Adam's running mean of squared gradients.
result Faster adaptation and more desirable convergence behaviors than Adam.
Gradient rollback explains neural models by identifying influential training examples.
problem Explain predictions of neural black-box models, especially in applications requiring user trust.
method Gradient rollback, a general approach for influence estimation applicable to neural models with limited parameter updates.
result Gradient rollback provides faithful explanations for neural models, including knowledge graph embedding methods.
Constructs expanding gradient Ricci solitons with unique properties.
problem Creating expanding gradient Ricci solitons with specific characteristics.
method Combining previous work with localized maximum principle.
result Constructs various examples of expanding gradient Ricci solitons with positive curvature and exotic curvature decay.
Survey on gradient Ricci solitons in 4D, focusing on geometry and classification.
problem Understanding gradient Ricci solitons in four dimensions.
method Geometric analysis and classification of solitons.
result Recent results on classification and rigidity of gradient Ricci solitons in 4D.
We make policy optimization algorithms batch size-invariant by decoupling proximal and behavior policies.
problem Some policy optimization algorithms do not have batch size-invariance, leading to inefficiencies.
method We decouple the proximal policy from the behavior policy to achieve batch size-invariance.
result Our approach makes policy optimization algorithms more efficient and allows them to use stale data more effectively.
Robust algorithm for distributed optimization resistant to Byzantine failures.
problem Resilient optimization in the presence of unreliable agents.
method Temporal and spatial robust aggregation, gradient normalization.
result Convergence for strongly convex and non-convex functions.
We investigate the low-energy behavior of the gradient flow of the norm of the Riemannian curvature on four-manifolds. Specifically, we show long time existence and exponential convergence to a metric of constant sectional curvature when the initial metric has positive Yamabe constant and small initial energy.
Experience replay (ER) is a fundamental component of off-policy deep reinforcement learning (RL). ER recalls experiences from past iterations to compute gradient estimates for the current policy, increasing data-efficiency. However, the accuracy of such updates may deteriorate when the policy diverges from past behavio…
We propose a simple and general variant of the standard reparameterized gradient estimator for the variational evidence lower bound. Specifically, we remove a part of the total derivative with respect to the variational parameters that corresponds to the score function. Removing this term produces an unbiased gradient …
Deep neural networks are usually trained with stochastic gradient descent (SGD), which minimizes objective function using very rough approximations of gradient, only averaging to the real gradient. Standard approaches like momentum or ADAM only consider a single direction, and do not try to model distance from extremum…
We conduct mathematical analysis on the effect of batch normalization (BN) on gradient backpropogation in residual network training, which is believed to play a critical role in addressing the gradient vanishing/explosion problem, in this work. By analyzing the mean and variance behavior of the input and the gradient i…