Research
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

4284126168 · Jun 202019922001200920172026
48 results for Noise Scheduling

This study shows how DDPM can be represented by the OU process.

problem Designing optimal noise schedules for DDPM.
method Formal equivalence between DDPM and OU process, heuristic designs based on Fisher Information.
result Fisher-Information-motivated schedule corresponds to cosine noise schedule.

ANT improves TS diffusion models by automatically determining noise schedules.

problem Suboptimal performance of TS diffusion models due to lack of domain-specific noise schedules.
method ANT proposes an adaptive noise schedule that automatically determines proper noise schedules for TS datasets based on their statistics.
result ANT achieves state-of-the-art performance on various TS tasks, including forecasting, refinement, and generation.

Researchers establish bounds for SGMs' KL and Wasserstein divergences under various noise schedules.

problem Estimating the error between target and estimated distributions in SGMs.
method Established upper bounds for KL divergence and Wasserstein distance, incorporating target distribution properties and SGM hyperparameters.
result Optimal noise schedules identified for SGMs, improving generative quality.

We improve private training accuracy with learning rate schedules and matrix factorizations.

problem Private training with learning rate schedules and correlated noise.
method General upper and lower bounds for learning rate schedules, memory-efficient constructions, and schedule-aware factorizations.
result Schedule-aware factorizations improve accuracy in private training.

MDMs train to decode tokens in a random order, which affects performance; we show they can be optimized for a favorable order.

problem Performance of MDMs is affected by the random order in which tokens are decoded.
method We show that MDMs can be optimized for a favorable decoding order by equipping their continuous-time variational objective with multivariate noise schedules.
result MDMs can be decomposed into a weighted auto-regressive losses over orders, making them auto-regressive models with learnable orders.

Generative models improve for multiscale scientific data with new noise and interpolation techniques.

problem Numerical challenges in generating high-fidelity samples for multiscale scientific data.
method Design of noise distributions and interpolation schedules in function space to ensure Lipschitz regularity and finite noise roughness.
result Scale-adaptive noise and interpolation schedules improve numerical efficiency and fidelity of generated samples.

Study finds optimal learning rate schedules for sub-100M quantization-aware training across bit-widths.

problem Optimal learning rate schedules for quantization-aware training depend on bit-width.
method Factorial grid testing over bit-width, warmdown fraction, LR magnitude, model size, and seed.
result INT6 QAT requires a different schedule than higher-precision training, falsifying the primary hypothesis.

Improved diffusion models achieve state-of-the-art likelihoods in image density estimation.

problem Improving likelihood-based performance of diffusion models.
method Joint optimization of noise schedule and model parameters, using signal-to-noise ratio simplification.
result State-of-the-art likelihoods on image density estimation benchmarks, faster optimization.

This work sets lower bounds on the number of score queries needed for diffusion sampling.

problem Establishing information-theoretic limits on the number of score evaluations required for diffusion sampling.
method Proving lower bounds on the number of adaptive score queries needed for sampling.
result Any sampling algorithm requires at least \(\widetilde{\Omega}(\sqrt{d})\) adaptive score queries for \(d\)-dimensional distributions.

We present a representation learning method that learns features at multiple different levels of scale. Working within the unsupervised framework of denoising autoencoders, we observe that when the input is heavily corrupted during training, the network tends to learn coarse-grained features, whereas when the input is …

2014-06-12abs ↗pdf ↗

Adjoint Matching improves flow and diffusion models with reward fine-tuning.

problem Improving generative models with reward fine-tuning.
method Casting reward fine-tuning as stochastic optimal control (SOC) and enforcing a specific noise schedule.
result Adjoint Matching outperforms existing SOC algorithms.

CWGD measures gradient diversity weighted by curvature, improving SGD convergence.

problem Gradient noise in high-curvature directions is underestimated by standard methods.
method CWGD weights gradient diversity by the inverse square root of the Hessian.
result CWGD-Cosine reduces optimization error by up to 20% compared to standard cosine annealing.

A method predicts GNS of transformer layers using normalization layer norms.

problem Estimating gradient noise scale with minimal variance.
method Simultaneously compute per-example gradient norms and parameter gradients.
result Total GNS is predicted well by normalization layer GNS.

Adaptive batch size schedules improve language model training efficiency and generalization.

problem Dilemma of choosing batch sizes in large-scale model training.
method General-purpose adaptive batch size schedules compatible with data and model parallelism.
result Adaptive batch size schedules outperform constant batch sizes and heuristic warmup schedules.

New framework optimizes deep learning training by deferring large batch sizes to late stages.

problem Optimizing batch size scheduling for deep learning training efficiency.
method Introduced the functional scaling law (FSL) framework to analyze and optimize batch size scheduling.
result Large batch sizes can be deferred to late training stages without sacrificing performance.

Diffusion models accurately recover mixture weights from generated samples despite score function insensitivity.

problem Score-based generative models often fail to learn correct relative mode amplitudes (mixture weights) from generated samples.
method Relate diffusion score matching (DSM) loss to mixture weight estimation error, define diffusion score sensitivity index (DSSI), and prove its governing role in mixture weight recovery.
result Generated samples can accurately recover mixture weights from the DSM loss, even when the target score is insensitive to mixture weights.

Masking diffusion outperforms other discrete diffusion models by incorporating jump times into the model.

problem Improving the performance of discrete diffusion models.
method Conditioning on the jump schedule of discrete Markov processes.
result Schedule-conditioned discrete diffusion (SCUD) models outperform classical and masking diffusion models.

A new method optimizes diffusion models for fine-tuning tasks efficiently.

problem Optimizing diffusion models for downstream tasks using nested bilevel structures.
method Formalizes the challenge as a generative bilevel optimization problem and introduces a first-order bilevel framework.
result Our method outperforms existing fine-tuning and hyperparameter search baselines.

S-SGD adds symmetrical noise to weights to avoid sharp minima in deep learning.

problem SGD does not always converge to a flat minimum, leading to poor generalization.
method Symmetrical weight noise injection in SGD.
result S-SGD outperforms conventional SGD and weight-noise injection methods in large batch training.

Improved analysis for fair federated learning reduces dependence on noise floor.

problem Asymptotic stationarity in group fair federated learning with reduced noise floor dependence.
method DS FedProxGrad framework with inexact local proximal solutions and fairness regularization.
result Algorithm converges asymptotically to stationarity without dependence on a noise floor.

This paper proposes a method to select project schedules with the lowest risk.

problem Selecting schedules that meet project deadlines while minimizing risk.
method Integrating aleatory uncertainty into project scheduling to quantify and compare risks.
result Proposes a method to select schedules with the lowest risk.

ScheduleFree+ improves large language model training without schedules or learning rates.

problem Scaling up Schedule-Free Learning to large language models.
method Learning-rate-free and schedule-free method for training large language models.
result ScheduleFree+ outperforms SOTA schedules by 31% at 1000 tokens per parameter.

Optimal learning rates decay to zero in easy tasks and maintain a warmup phase in hard tasks.

problem Optimizing learning rates under functional scaling laws for model training.
method Deriving optimal learning-rate schedules based on exponents ss and ββ.
result Sharp phase transition between easy and hard tasks, with different decay behaviors.

The paper optimizes interpolation schedules in generative models to improve sampling accuracy.

problem Improving sampling accuracy in generative models with fewer resources.
method Minimizing the averaged squared Lipschitzness of the drift field, using transfer formulas.
result Designed schedules yield more accurate fine-scale statistics at fixed integrator budget.

New method converts and optimizes sampling schedules for generative models.

problem Optimizing sampling schedules for generative models like flows and diffusions.
method Unified framework for stochastic interpolants, including point mass schedules.
result Demonstrated efficient generation of images with fewer steps.

Improved SEG method converges to Nash equilibrium in bilinear games.

problem Stochastic bilinear minimax optimization problem
method Stochastic ExtraGradient (SEG) method with constant step size, iteration averaging, and scheduled restarting.
result Provable convergence to Nash equilibrium under standard settings, optimal convergence rate in interpolation setting.

We propose a variance reduction framework for variational inference using the Multilevel Monte Carlo (MLMC) method. Our framework is built on reparameterized gradient estimators and "recycles" parameters obtained from past update history in optimization. In addition, our framework provides a new optimization algorithm …

2019-02-01abs ↗pdf ↗

This paper proposes a system-agnostic policy for dynamic scheduling.

problem Dynamic scheduling in changing systems is challenging due to system-specific optimal policies.
method Descriptive policy that learns a system-agnostic scheduling principle.
result System-agnostic meta-learning enables adaptation to unseen system characteristics.

The paper presents a multi-power law for predicting loss curves across different learning rate schedules.

problem Understanding and optimizing the relationship between model performance and hyperparameters, especially learning rates.
method Proposes a multi-power law that combines power laws based on the sum of learning rates and additional laws for loss reduction due to decay.
result The multi-power law accurately predicts loss curves for unseen learning rate schedules and finds a schedule that outperforms cosine learning rate.

BDDMs eliminate noise conditioning in diffusion models, simplifying training and sampling.

problem Noise conditioning in diffusion models is ad hoc and requires unprincipled noise embeddings.
method Introduce blind denoising diffusion models (BDDMs) that do not require noise conditioning.
result BDDMs simplify training and sampling by eliminating noise conditioning.

More and more companies have deployed machine learning (ML) clusters, where deep learning (DL) models are trained for providing various AI-driven services. Efficient resource scheduling is essential for maximal utilization of expensive DL clusters. Existing cluster schedulers either are agnostic to ML workload characte…

2019-09-13abs ↗pdf ↗

Optimal learning rate schedules for SGD in changing data distributions.

problem Minimizing regret in online learning with changing data distributions.
method Characterized optimal schedules for linear regression, proposed schedules for general convex and non-convex losses, and defined a notion of regret for non-convex losses.
result Upper and lower bounds for regret with constants for convex losses, and an upper bound on total expected regret for non-convex losses.

SignSGD outperforms SGD in linear regression with optimal scaling laws under PLRF model.

problem Improving linear regression performance with signSGD under power-law random features.
method Analysis of signSGD risk under PLRF model, comparison with SGD, identification of unique effects.
result SignSGD can have a steeper compute-optimal slope than SGD in noisy regimes, especially with WSD schedule.

Enhances multi-project scheduling with multiple priority rules.

problem Resource allocation in multi-project scheduling with limited time and resources.
method Simulation-based approach using composite priority rules.
result Increased probability of finding schedules with shortest duration.

Learning-rate schedules for large models match optimization theory closely, leading to better training.

problem Improving training of large models with optimal learning rates.
method Used a bound from non-smooth convex optimization theory to match learning-rate schedules with practical benefits.
result Extending the learning-rate schedule with optimal learning-rate and transferring it across schedules improves model training.