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.
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.
Seesaw optimizes training by balancing learning rate and batch size, accelerating model pretraining.
problem Optimizing training efficiency for large language models with adaptive optimizers.
method Develops a principled framework for batch-size scheduling, introducing Seesaw which multiplies learning rate by 1/√2 and doubles batch size.
result Empirically, Seesaw reduces wall-clock time by approximately 36% compared to cosine decay, matching theoretical limits.
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.
Bayesian method optimizes rescheduling for multipurpose batch processes with incomplete look-ahead information.
problem Optimizing rescheduling for multipurpose batch processes under incomplete look-ahead information.
method Proposes a Bayesian dynamic scheduling method that learns from disturbances and updates schedules online.
result Achieves statistically better long-term costs and system nervousness compared to existing periodic rescheduling strategies.
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.
New analysis reveals batch size effects on stochastic conditional gradient methods.
problem Understanding the role of batch size in stochastic conditional gradient methods.
method Deriving a new analysis focusing on momentum-based stochastic conditional gradient algorithms (e.g., Scion).
result Increasing batch size initially improves optimization accuracy but can degrade performance beyond a critical threshold.
We investigate the dynamical and convergent properties of stochastic gradient descent (SGD) applied to Deep Neural Networks (DNNs). Characterizing the relation between learning rate, batch size and the properties of the final minima, such as width or generalization, remains an open question. In order to tackle this pro…
This paper introduces sample-averaged Q-learning for better RL performance.
problem Improving reinforcement learning algorithms by managing uncertainty.
method Integrates statistical inference into Q-learning through sample averaging and functional central limit theorem.
result Establishes a unified theoretical foundation for sample-averaged Q-learning.
New convergence results for NGVI with various step sizes and sample sizes.
problem Understanding convergence of stochastic NGVI for various schedules.
method Projected stochastic NGVI for exponential family variational distributions.
result Geometric convergence and $\mathcal{O}\left(\frac{1}{T^ρ}
ight)$ rates for different schedules.
We find optimal learning rate schedules for a random feature model.
problem Choosing optimal learning rates for deep learning models.
method We analyze a powerlaw random feature model trained with SGD, considering optimal schedules as numerical and analytical problems.
result We discover two regimes: easy and hard phases, with different optimal learning rate schedules.
Large-batch SGD is important for scaling training of deep neural networks. However, without fine-tuning hyperparameter schedules, the generalization of the model may be hampered. We propose to use batch augmentation: replicating instances of samples within the same batch with different data augmentations. Batch augment…
New method shows how order of gradient updates impacts stability and convergence in deep learning.
problem Training deep learning models can be unstable and computationally expensive.
method Theoretical analysis and experiments with backward-SGD.
result The order of gradient updates affects stability and convergence, leading to improved performance.
Adjusting the learning rate schedule in stochastic gradient methods is an important unresolved problem which requires tuning in practice. If certain parameters of the loss function such as smoothness or strong convexity constants are known, theoretical learning rate schedules can be applied. However, in practice, such …
With an increasing demand for training powers for deep learning algorithms and the rapid growth of computation resources in data centers, it is desirable to dynamically schedule different distributed deep learning tasks to maximize resource utilization and reduce cost. In this process, different tasks may receive varyi…
Mini-batch stochastic gradient descent and variants thereof have become standard for large-scale empirical risk minimization like the training of neural networks. These methods are usually used with a constant batch size chosen by simple empirical inspection. The batch size significantly influences the behavior of the …
The paper optimizes LLM inference systems through queueing theory.
problem Efficient LLM inference for AI agents under various routing topologies.
method Developed a fluid-limit framework for multi-class batched processing networks under K-FCFS scheduling.
result Proved that work-conserving scheduling algorithms maximize throughput for LLM inference.
Paper proposes faster certified robust training methods with short warmup.
problem Certified robust training methods require long warmup schedules, making training costly.
method Proposes three improvements: new weight initialization, BN, and regularization.
result Achieves 65.03% verified error on CIFAR-10 and 82.36% on TinyImageNet with short warmup.
BaNk-UCB tackles batched nonparametric bandits with k-NN regression and UCB.
problem Sequential decision-making with limited online feedback in domains like medicine and marketing.
method Combines k-NN regression with UCB principle for fully nonparametric, adaptive, and simple implementation.
result Near-optimal regret guarantees under Lipschitz smoothness and margin assumptions, with minimax-optimal rates.
Bayesian optimization method for permutations accelerates combinatorial search.
problem Optimizing expensive-to-evaluate objectives on permutation problems.
method LAW2ORDER, a batch Bayesian optimization method based on the acquisition weighted kernel.
result LAW2ORDER achieves sublinear batch cumulative regret, demonstrating accelerated search.
Noise in SGD helps deep nets generalize better, even with smaller batch sizes.
problem The generalization benefit of using noise in SGD over large batch sizes.
method Carefully designed experiments and rigorous hyperparameter sweeps on various models.
result Small or moderately large batch sizes outperform very large batches on test sets.
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.
New method SF-AdamW trains large models without decay phases or memory overhead.
problem Inadequate fixed compute budgets for large-scale training.
method Schedule-Free (SF) method revisited and refined.
result SF-AdamW effectively navigates loss landscape without decay phases or memory overhead.
It is common practice to decay the learning rate. Here we show one can usually obtain the same learning curve on both training and test sets by instead increasing the batch size during training. This procedure is successful for stochastic gradient descent (SGD), SGD with momentum, Nesterov momentum, and Adam. It reache…
A new method for graph neural networks speeds up inference and training.
problem Challenges in constructing mini-batches for large graphs in graph neural networks.
method Theoretical model of batch construction via maximizing influence score of nodes on outputs.
result Accelerates inference by up to 130x compared to previous methods.
A major challenge in current optimization research for deep learning is to automatically find optimal step sizes for each update step. The optimal step size is closely related to the shape of the loss in the update step direction. However, this shape has not yet been examined in detail. This work shows empirically that…
GOLS-I automatically determines learning rates for various neural network training algorithms.
problem Adapting learning rates in stochastic training algorithms for neural networks.
method Gradient-Only Line Search (GOLS-I) for automatically setting learning rates.
result GOLS-I learning rate schedules are competitive with manually tuned rates across multiple algorithms, architectures, datasets, and loss functions.
Intriguing empirical evidence exists that deep learning can work well with exoticschedules for varying the learning rate. This paper suggests that the phenomenon may be due to Batch Normalization or BN, which is ubiquitous and provides benefits in optimization and generalization across all standard architectures. The f…
We determine the critical batch size for large language models and find it scales with data size, not model size.
problem Determining the optimal batch size for large-scale model training.
method We propose a measure of critical batch size, pre-trained models, and systematic hyper-parameter sweeps.
result The critical batch size scales primarily with data size, not model size.
New algorithm reduces costs and latency for large language model inference.
problem Optimizing inference costs and latency for large language models with GPU constraints.
method Formulated as an online scheduling problem with endogenous memory growth, introduced fluid model and WAIT algorithms.
result Reduced costs and latency, especially in near-overloaded and overloaded regimes.
Accelerates BERT pretraining from 3 days to 54 minutes.
problem Long training time of BERT due to large mini-batch sizes.
method LANS method and learning rate scheduler for large mini-batch training.
result Achieved fastest BERT training time of 54 minutes.
Step sizes in neural network training are largely determined using predetermined rules such as fixed learning rates and learning rate schedules. These require user input or expensive global optimization strategies to determine their functional form and associated hyperparameters. Line searches are capable of adaptively…
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.
SBA improves neural network generalization by dynamically augmenting data.
problem Lack of direct supervision in data augmentation for generalization.
method Stochastic Batch Augmentation with dynamic soft label regularization.
result SBA improves generalization and speeds up training.
Recent work has shown how to train Convolutional Neural Networks (CNNs) rapidly on large image datasets, then transfer the knowledge gained from these models to a variety of tasks. Following [Radford 2017], in this work, we demonstrate similar scalability and transfer for Recurrent Neural Networks (RNNs) for Natural La…
Normalization methods are a central building block in the deep learning toolbox. They accelerate and stabilize training, while decreasing the dependence on manually tuned learning rate schedules. When learning from multi-modal distributions, the effectiveness of batch normalization (BN), arguably the most prominent nor…
Simplified analysis of SGD for linear regression with weight averaging.
problem Understanding SGD optimization in linear regression models.
method Simplified analysis using linear algebra tools, bypassing complex operator manipulations.
result Recovery of bias and variance bounds for SGD in linear regression.
A new framework reduces traffic congestion by 36%.
problem Efficient traffic signal control in a metropolitan area.
method Two-stage framework combining Evolution Strategies and multi-agent reinforcement learning.
result Reduces traffic congestion by 36%.
The paper analyzes early stopping in linear regression and shows it's equivalent to ridge regularization.
problem Understanding the effect of early stopping on linear regression models.
method Characterization of gradient descent dynamics and analysis of excess risk.
result Early stopped solution is equivalent to minimum norm solution for a generalized ridge regularized problem.
Study online learning with delays and capacity constraints, achieving optimal regret bounds.
problem Online learning with delays and capacity constraints.
method Novel scheduling and preemptive techniques, matching upper and lower bounds.
result Achieves optimal regret bounds across all capacity levels.
New method schedules learning rate without stopping time, outperforming existing methods.
problem Learning rate schedules that require stopping time are outperformed.
method Schedule-Free approach that avoids stopping time and introduces no additional hyper-parameters.
result Exhibits state-of-the-art performance across various problems.
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.
Study examines how data augmentation impacts optimization in linear regression.
problem Understanding how data augmentation schedules affect optimization in linear regression.
method Analyzed the effect of augmentation on optimization in linear regression with MSE loss, using classical convex optimization and recent work on implicit bias.
result Proved that under certain joint schedules for learning rate and augmentation scheme, augmented gradient descent converges and characterized the resulting minimum.
Optimizes financial auditor schedules to reduce time and costs.
problem Efficiently scheduling financial auditors with multiple constraints.
method Used Integer Linear Programming and compared two exact formulations.
result Multi-commodity network flow formulation is 24 times faster.
Subset pretraining speeds up neural network training.
problem Efficiently training neural networks with partial gradients.
method Use small subsets to approximate the loss minimum of the full training set.
result Subset minima can approximate the full training set's loss minimum efficiently.
Cosine schedule is optimal for discrete diffusion models.
problem Choosing the best discretization schedule for diffusion models.
method Optimized using Fisher-Rao geometry.
result Cosine schedule is Fisher-Rao optimal.
Optimal learning rate schedules derived for various tasks.
problem Inadequate learning rate schedules in practice compared to theory.
method Refined analysis of learning rate schedules for optimization algorithms.
result Derives new problem-adaptive learning rate schedules.
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.