With the increase in the amount of data and the expansion of model scale, distributed parallel training becomes an important and successful technique to address the optimization challenges. Nevertheless, although distributed stochastic gradient descent (SGD) algorithms can achieve a linear iteration speedup, they are l…
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
48 results for “Training Speedup”
Novel periodic momentum SGD method for decentralized training with linear speedup.
problem Lack of effective momentum schema in decentralized training methods.
method Proposes a novel periodic decentralized momentum SGD method.
result Achieves linear speedup in decentralized training.
Predicting structured outputs can be computationally onerous due to the combinatorially large output spaces. In this paper, we focus on reducing the prediction time of a trained black-box structured classifier without losing accuracy. To do so, we train a speedup classifier that learns to mimic a black-box classifier u…
The impact of the maximally possible batch size (for the better runtime) on performance of graphic processing units (GPU) and tensor processing units (TPU) during training and inference phases is investigated. The numerous runs of the selected deep neural network (DNN) were performed on the standard MNIST and Fashion-M…
Speeds up deep neural networks training by 10x using GPU concurrency.
problem Training deep residual neural networks efficiently.
method Layer-wise parallel training with GPU concurrency and Nonlinear Multigrid.
result 10.2x speedup over traditional techniques.
System learns optimizer hyperparameters to generalize across tasks.
problem Lack of generalization in learning optimizers for neural networks.
method Generalization-first approach, learning optimizer hyperparameters.
result System outperforms Adam on various tasks, including unseen modalities.
Distributed Non-Convex Optimization with Sublinear Speedup under Intermittent Client Availabilitystat.ML
Federated learning algorithm improves with intermittent client availability.
problem Performance degradation in Federated Averaging due to client availability changes.
method Federated Latest Averaging (FedLaAvg) uses latest gradients from all clients, even when unavailable.
result FedLaAvg achieves sublinear speedup compared to classical Federated Averaging.
SliceOut speeds up deep learning training without sacrificing accuracy.
problem Frequent model re-training and large model training workloads in deep learning.
method SliceOut uses dropout-inspired scheme to drop contiguous sets of units at random, leveraging GPU memory layout.
result 10-40% speedups and memory reduction with minimal accuracy loss.
Recurrent neural networks (RNN) have been successfully applied to various sequential decision-making tasks, natural language processing applications, and time-series predictions. Such networks are usually trained through back-propagation through time (BPTT) which is prohibitively expensive, especially when the length o…
GPU optimization speeds up large-scale classification tasks.
problem Efficiently training large-scale classification models on GPUs.
method Judecious GPU-optimization principles applied to TRON algorithm.
result Significant speedups for logistic regression and SVM classification.
Neural networks offer high-accuracy solutions to a range of problems, but are costly to run in production systems because of computational and memory requirements during a forward pass. Given a trained network, we propose a techique called Deep Learning Approximation to build a faster network in a tiny fraction of the …
State-of-the-art implementations of boosting, such as XGBoost and LightGBM, can process large training sets extremely fast. However, this performance requires that the memory size is sufficient to hold a 2-3 multiple of the training set size. This paper presents an alternative approach to implementing the boosted trees…
Critical Parameters for Scalable Distributed Learning with Large Batches and Asynchronous Updatescs.LG
New insights explain speedup saturation in distributed learning with large batches and delays.
problem Understanding and optimizing speedup in distributed learning with large batches and delays.
method Theoretical analysis of strongly convex, convex, and non-convex settings, considering data sparsity.
result Identification of a data-dependent parameter explaining speedup saturation in both batch size and gradient staleness.
The Graph Convolutional Network (GCN) model and its variants are powerful graph embedding tools for facilitating classification and clustering on graphs. However, a major challenge is to reduce the complexity of layered GCNs and make them parallelizable and scalable on very large graphs -- state-of the art techniques a…
A new training method speeds up ResNet training by 3x with minimal accuracy loss.
problem Training ResNets is slow due to dependencies between modules.
method Serial-parallel hybrid training strategy with data augmentation and downsampling.
result Significant speedup over traditional methods with comparable accuracy.
This work compares and evaluates various sampling methods for neural language models.
problem Lack of systematic comparison and myths about sampling methods.
method Monte Carlo sampling, importance sampling, compensated partial summation, noise contrastive estimation.
result All sampling methods can perform equally well if posterior probabilities are corrected.
Training deep networks is expensive and time-consuming with the training period increasing with data size and growth in model parameters. In this paper, we provide a framework for distributed training of deep networks over a cluster of CPUs in Apache Spark. The framework implements both Data Parallelism and Model Paral…
New study finds optimal hyperparameter tuning crucial for fair optimizer comparisons.
problem Inadequate hyperparameter tuning and misleading evaluation setups hinder fair comparisons of optimizers.
method Systematic study of ten optimizers across four model scales and data-to-model ratios.
result Optimal hyperparameters for one optimizer may be suboptimal for another, and many claimed speedups are lower than expected.
A central task in the field of quantum computing is to find applications where quantum computer could provide exponential speedup over any classical computer. Machine learning represents an important field with broad applications where quantum computer may offer significant speedup. Several quantum algorithms for discr…
A new multilevel framework speeds up ResNet training.
problem Training deep residual networks (ResNets) is time-consuming.
method Formulates ResNets as dynamical systems and uses time-dependent optimal control problems.
result Enhanced training of ResNets with multilevel auxiliary networks achieves significant speedup.
COPML framework securely trains models across multiple data owners without revealing individual data.
problem Privacy-preserving collaborative machine learning with multiple data owners.
method Securely encodes data, distributes computation, performs distributed training.
result Achieves up to 16x speedup in training time while maintaining strong privacy.
Convolutional Neural Network (CNN) based Deep Learning (DL) has achieved great progress in many real-life applications. Meanwhile, due to the complex model structures against strict latency and memory restriction, the implementation of CNN models on the resource-limited platforms is becoming more challenging. This work…
In unsupervised learning, collecting more data is not always a costly process unlike the training. For example, it is not hard to enlarge the 40GB WebText used for training GPT-2 by modifying its sampling methodology considering how many webpages there are in the Internet. On the other hand, given that training on this…
Orders-of-magnitude speedup in atmospheric chemistry modeling through neural network-based emulationphysics.ao-ph
Chemical transport models (CTMs), which simulate air pollution transport, transformation, and removal, are computationally expensive, largely because of the computational intensity of the chemical mechanisms: systems of coupled differential equations representing atmospheric chemistry. Here we investigate the potential…
ML-EM method speeds up diffusion model sampling.
problem Efficiently sampling from complex diffusion models.
method Multilevel Euler-Maruyama method with UNet approximations.
result Polynomial speedup in sampling from diffusion models.
VRL-SGD reduces communication complexity in non-identical data settings.
problem Training machine learning models with non-identical data distribution.
method VRL-SGD, which eliminates gradient variance dependency and achieves linear speedup with lower communication complexity.
result VRL-SGD reduces communication complexity from $O(T^{ rac{3}{4}} N^{ rac{3}{4}})$ to $O(T^{ rac{1}{2}} N^{ rac{3}{2}})$.
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.
TSSM splits neural networks for parallel training with minimal accuracy loss.
problem Accuracy degradation in parallel training of deep neural networks.
method TSSM reformulates alternating minimization to achieve parallelism with minimal accuracy loss.
result TSSM achieves significant speedup without accuracy loss on multiple datasets.
Quantum computing speeds up linear regression training.
problem Reducing training time for machine learning models.
method Formulated regression problem as QUBO, used D-Wave 2000Q for adiabatic optimization.
result Quantum approach achieves up to 2.8x speedup on larger datasets.
Paper finds sharpness differences in transformer blocks accelerating LLM training.
problem Understanding and accelerating large language model pre-training.
method Uncovering sharpness disparity across transformer blocks and proposing Blockwise Learning Rate.
result Blockwise Learning Rate strategy accelerates LLM pre-training with lower loss and speedup.
Deploying deep learning (DL) models across multiple compute devices to train large and complex models continues to grow in importance because of the demand for faster and more frequent training. Data parallelism (DP) is the most widely used parallelization strategy, but as the number of devices in data parallel trainin…
SIFT reduces training time by selecting samples with approximate losses.
problem Reducing training time by selecting samples with large approximate losses.
method Developed SIFT which uses early exiting to obtain approximate losses with intermediate layer representations for sample selection.
result SIFT achieves significant gains in training time and number of backpropagation steps without optimized implementation.
FROCC uses random projections for fast one-class classification.
problem Efficient one-class classification for large datasets.
method Random projection of data onto random unit vectors, with region bounding.
result Significant improvement in ROC performance with substantial speedup.
COMP-AMS optimizes distributed training with compressed gradients, achieving similar accuracy with less communication.
problem Efficiently training large-scale models in distributed environments with reduced communication costs.
method Distributed optimization framework using gradient averaging and adaptive AMSGrad, with gradient compression and error feedback.
result COMP-AMS achieves the same convergence rate and linear speedup as standard AMSGrad with less communication.
In machine learning, asynchronous parallel stochastic gradient descent (APSGD) is broadly used to speed up the training process through multi-workers. Meanwhile, the time delay of stale gradients in asynchronous algorithms is generally proportional to the total number of workers, which brings additional deviation from …
ProxSkip achieves linear speedup in distributed non-convex optimization.
problem Achieving linear speedup in distributed non-convex optimization.
method Unified convergence analysis for stochastic non-convex, convex, and strongly convex problems.
result ProxSkip achieves linear speedup in the number of nodes under stochastic gradients.
Efficient algorithm finds fast Transformer models.
problem Slow inference time of Transformer models.
method Decompose Transformer architecture into components, use sampling-based one-shot search.
result Achieved 10% to 30% speedup on pre-trained BERT and 70% on top of a previous state-of-the-art model.
Improved neural framework for scaling entropic MOT with significant computational gains.
problem High computational overhead in multimarginal optimal transport.
method Neural Entropic MOT (NEMOT) using mini-batch training to reduce complexity.
result Significant speedups and feasibility improvements for multimarginal data.
MACER accelerates error repair by modularly identifying and applying fixes.
problem Automated compilation error repair for novice programmers.
method Modular segregation of repair process into identification and application, using discriminative learning techniques.
result MACER outperforms existing methods by 20% on popular errors and is competitive on all error types.
Quantum computing promises faster finance algorithms.
problem Solving finance problems faster than classical methods.
method Quantum computing applications to finance, including Monte Carlo, portfolio optimization, and machine learning.
result Quantum speedups for finance problems, especially Monte Carlo and portfolio optimization.
A fast method for Lasso and Logistic Lasso problems.
problem Solving Lasso and Logistic Lasso regression problems efficiently.
method Iterative active set approach using solver updates.
result 31.41 times faster on average for compressed sensing.
In an era when the performance of a single compute device plateaus, software must be designed to scale on massively parallel systems for better runtime performance. However, in the context of training deep learning models, the popular back-propagation (BP) algorithm imposes a strong sequential dependency in the process…
A new method for automatic gradient tree boosting using information theory.
problem Automatic selection of tree complexity and number in gradient boosting.
method Optimism of greedy leaf splitting procedure modeled as a Cox-Ingersoll-Ross process, leading to an information criterion for model selection.
result The method achieves significant speedups (10-1400) compared to xgboost without sacrificing predictive power.
Quantum computing offers a quadratic speedup for estimating non-linear functionals.
problem Estimating non-linear functionals of probability distributions.
method Proposes a quantum-inside-quantum Monte Carlo algorithm for a broad class of non-linear estimation problems.
result Achieves a quadratic speedup for non-linear estimation problems, including nested conditional expectations and stochastic optimization.
Federated Q-learning achieves linear speedup with heterogeneity, improving sample complexity.
problem Collaborative learning in distributed RL settings with limited data sharing.
method Analyzes synchronous and asynchronous federated Q-learning, proposes importance averaging.
result Achieves linear speedup with heterogeneity, robust to local trajectory heterogeneity.
The early layers of a deep neural net have the fewest parameters, but take up the most computation. In this extended abstract, we propose to only train the hidden layers for a set portion of the training run, freezing them out one-by-one and excluding them from the backward pass. Through experiments on CIFAR, we empiri…
GPA improves LLM training speed by 8.71% for Llama-160M models.
problem Training Large Language Models (LLMs) with high memory overhead and slow convergence.
method Generalized Primal Averaging (GPA) extends Nesterov's method to eliminate memory-intensive two-loop structure.
result GPA achieves up to 10.13% speedup over AdamW in training Llama-1B model.
Paper speeds up large foundation models for time series data.
problem Resource-intensive foundation models limit accessibility.
method Dimensionality reduction techniques, including PCA and neural network adapters.
result Up to 10x speedup and 4.5x more datasets fit on a single GPU.