Paper proposes NASAIC framework for co-designing neural architectures and heterogeneous ASICs.
problem Designing efficient neural architectures and ASICs for multiple tasks.
method Build ASIC templates and propose NASAIC framework for simultaneous design of architectures and ASICs.
result NASAIC ensures design specifications and maximizes accuracy with minimal performance loss.
Optimized neural networks for Edge TPU achieve high accuracy in real-time image classification.
problem Designing neural networks for hardware accelerators to achieve optimal performance.
method Hardware-aware neural architecture search and model customization for Edge TPU.
result Improved accuracy-latency tradeoff on Pixel 4's Edge TPU compared to existing models.
DANCE optimizes neural network and accelerator design for faster, more efficient DNN execution.
problem Challenges in optimizing neural network and accelerator design for efficient DNN execution.
method Differentiable approach to co-exploration of accelerator and network architecture design.
result Significantly shorter time to achieve superior accuracy and hardware cost metrics.
Morph accelerates 3D CNNs for video recognition, reducing energy consumption and improving performance.
problem Efficiently accelerating 3D CNNs for video recognition is challenging due to their large memory footprint and higher dimensionality.
method Designing a flexible accelerator called Morph that adapts to different spatial and temporal tiling strategies, and codesigning a software infrastructure to control the hardware.
result Morph achieves up to 3.4x reduction in energy consumption and up to 5.1x improvement in performance/watt compared to a baseline 3D CNN accelerator.
AGBM accelerates GBM with theoretical guarantees.
problem Accumulation of errors in GBM's momentum term.
method Incorporates Nesterov's acceleration techniques and a corrected pseudo residual.
result First GBM type with theoretically-justified accelerated convergence rate.
TinyCNN accelerates CNN models on embedded FPGA with 15x speedup.
problem Limited memory on embedded FPGAs restricts CNN performance.
method Software and hardware design tool for FPGA resource-aware CNN accelerator.
result 3% accuracy loss with 15.75x speedup on image classification.
Accelerates policy optimization in RL with optimistic and adaptive updates.
problem Improving policy optimization methods in reinforcement learning.
method Integrates foresight into policy improvement step via optimistic and adaptive updates.
result Designs an optimistic policy gradient algorithm, adaptive via meta-gradient learning.
Memory-augmented neural networks (MANNs) are designed for question-answering tasks. It is difficult to run a MANN effectively on accelerators designed for other neural networks (NNs), in particular on mobile devices, because MANNs require recurrent data paths and various types of operations related to external memory a…
Bayesian DOE accelerates experimental design with improved efficiency.
problem Enhancing experimental design efficiency and reliability.
method Bayesian framework, conditional density estimation, informative data selection.
result Significantly improved computational efficiency of experimental design.
Paper accelerates diffusion models, improving sampling speed.
problem Low sampling speed in score-based diffusion models.
method Design of novel training-free algorithms for deterministic and stochastic samplers.
result Accelerated samplers converge faster with improved rates.
New adaptive and accelerated SGD methods achieve optimal convergence rates.
problem Optimizing convergence rates of stochastic gradient descent methods.
method Integrates diagonal scaling and momentum into accelerated SGD.
result Achieves optimal sampling and iteration complexity for smooth stochastic optimization.
We consider the problem of accelerating distributed optimization in multi-agent networks by sequentially adding edges. Specifically, we extend the distributed dual averaging (DDA) subgradient algorithm to evolving networks of growing connectivity and analyze the corresponding improvement in convergence rate. It is know…
Computer vision performances have been significantly improved in recent years by Convolutional Neural Networks(CNN). Currently, applications using CNN algorithms are deployed mainly on general purpose hardwares, such as CPUs, GPUs or FPGAs. However, power consumption, speed, accuracy, memory footprint, and die size sho…
ANADDH uses deep learning to improve volatility risk management.
problem Traditional Vega hedging strategies are inadequate for rapidly changing markets.
method Combines distributional reinforcement learning with adaptive Nesterov acceleration.
result Significant performance gains over existing hedging techniques.
A new method accelerates deep neural network training using minimal margin score.
problem Training deep neural networks is computationally expensive.
method Introduces minimal margin score (MMS) for selecting samples.
result Significant acceleration in training deep neural networks.
ASVRG accelerates stochastic variance reduction methods with simplicity and efficiency.
problem Efficiently solving convex and non-convex optimization problems.
method Accelerated proximal stochastic variance reduced gradient (ASVRG) method with momentum acceleration.
result ASVRG achieves best known oracle complexities for strongly and non-strongly convex objectives.
Bézier-GAN optimizes airfoil design by reducing shape complexity.
problem High computational cost in aerodynamic shape optimization.
method Generative adversarial networks (GANs) to learn compact shape representations.
result Empirically accelerates optimization convergence by at least two times.
Improves error predictors by selecting features based on intermediate accelerator results.
problem Designing accurate error predictors for approximate computing.
method Scheduling-aware feature selection leveraging accelerator intermediate results.
result Significant improvements in prediction accuracy compared to prior works.
Deep neural networks (DNN) are increasingly being accelerated on application-specific hardware such as the Google TPU designed especially for deep learning. Timing speculation is a promising approach to further increase the energy efficiency of DNN accelerators. Architectural exploration for timing speculation requires…
Accelerates stochastic optimization for convex and strongly convex problems.
problem Improving convergence rates in noisy stochastic optimization.
method Extends Catalyst approach to stochastic settings, handles inexact proximal operators.
result Achieves optimal worst-case complexity for noise-dominated regions.
With the rise of self-driving vehicles comes the risk of accidents and the need for higher safety, and protection for pedestrian detection in the following scenarios: imminent crashes, thus the car should crash into an object and avoid the pedestrian, and in the case of road intersections, where it is important for the…
There is widespread sentiment that it is not possible to effectively utilize fast gradient methods (e.g. Nesterov's acceleration, conjugate gradient, heavy ball) for the purposes of stochastic optimization due to their instability and error accumulation, a notion made precise in d'Aspremont 2008 and Devolder, Glineur, …
AutoAssist accelerates deep neural network training by filtering out less informative instances.
problem Efficiently training deep neural networks with millions of instances.
method AutoAssist uses a shrinking operation to filter out less informative instances, accelerating training.
result AutoAssist reduces training time by 40% for ResNet and 30% for transformer models.
Framework for accelerated gradient flows in Bayesian inverse problems.
problem Design efficient MCMC algorithms for Bayesian inverse problems.
method Nesterov's accelerated gradient flows in probability space, considering various information metrics.
result Proved convergence properties and proposed sampling-efficient algorithms for different metrics.
This paper tackles co-design of neural hardware and software to improve efficiency.
problem Designing efficient deep learning systems that consider both hardware and software optimizations together.
method Developed a constrained Bayesian optimization framework to automatically identify profitable design points in the joint hardware/software design space.
result Improved energy-delay product by 18% (ResNet) and 40% (DQN) over hand-tuned systems.
Stealthy hardware Trojan exploits DLA architecture vulnerabilities.
problem Security of DLA deployed on hardware accelerators.
method Input Interception Attack (IIA) exploiting statistical properties of DLA outputs.
result Stealthy Trojan can trigger with some definiteness.
We accelerate PMD algorithms for reinforcement learning using functional methods.
problem Improving reinforcement learning algorithms for large-scale optimization.
method Proposed a momentum-based update for PMD algorithms leveraging duality.
result The proposed method accelerates reinforcement learning algorithms.
SwitchML accelerates machine learning training by aggregating model updates in the network.
problem Training machine learning models in parallel consumes a lot of data.
method Designing a communication primitive that uses a programmable switch to aggregate model updates.
result SwitchML reduces data exchange by up to 5.5 times, speeding up training.
PABO optimizes DNN and hardware hyperparameters for efficient edge device acceleration.
problem Joint optimization of DNN accuracy and hardware cost for edge devices.
method Bayesian optimization with pseudo agent-based approach for memristive crossbar accelerators.
result PABO achieves significant speed-ups and superior performance compared to state-of-the-art methods.
Gradient tree boosting is a prediction algorithm that sequentially produces a model in the form of linear combinations of decision trees, by solving an infinite-dimensional optimization problem. We combine gradient boosting and Nesterov's accelerated descent to design a new algorithm, which we call AGB (for Accelerated…
A new transformer model accelerates training with optimization techniques.
problem Training deep neural networks efficiently and effectively.
method Interprets transformer layers as optimization steps, applying Nesterov acceleration.
result The new model outperforms existing models on benchmark datasets.
Sparsity helps reduce the computational complexity of deep neural networks by skipping zeros. Taking advantage of sparsity is listed as a high priority in next generation DNN accelerators such as TPU. The structure of sparsity, i.e., the granularity of pruning, affects the efficiency of hardware accelerator design as w…
ASVGD accelerates SVGD for efficient sampling.
problem Slow SVGD in high-dimensional sampling.
method Accelerated gradient flow in a metric space of probability densities, using Nesterov's method and momentum-based updates.
result ASVGD outperforms SVGD and other methods in sampling efficiency.
Accelerates data loading in deep neural network training by 30x.
problem Data loading is a bottleneck in deep neural network training.
method Locality-aware data loading method using software caches.
result More than 30x speedup in data loading.
This paper speeds up simulations of hypersonic reentry by combining traditional and neural methods.
problem Accurately simulating hypersonic reentry with chemical reactions is computationally expensive.
method A hybrid simulation code combining a traditional fluid dynamics solver with a neural network approximating chemical reactions.
result Achieved significant acceleration factors (imes10 to imes18.6) while maintaining accuracy. Study shows BNN inference accelerators are vulnerable to soft errors, causing significant misclassification.
problem Vulnerability of BNN inference accelerators to soft errors affecting image classification accuracy.
method Fault injection experiments on CIFAR-10 and MNIST datasets to assess impact of soft errors.
result Accuracy of image classifier can drop by up to 76.70% in worst-case scenarios.
TorchKM: A GPU-Oriented Library for Kernel Learning and Model Selection
problem Kernel learning and model selection
method GPU acceleration
result Competitive predictive performance with speedups
VegasFlow accelerates complex simulations across various hardware platforms.
problem Complex calculations and simulations requiring high-dimensional integrals.
method Monte Carlo integration techniques using Vegas algorithm and TensorFlow.
result Significantly faster performance on various hardware platforms.
We present a design and implementation of the Thomas algorithm optimized for hardware acceleration on an FPGA, the Thomas Core. The hardware-based algorithm combined with the custom data flow and low level parallelism available in an FPGA reduces the overall complexity from 8N down to 5N serial arithmetic operations, a…
GrateTile optimizes CNN feature map storage for efficient data access.
problem Efficient storage and access of sparse CNN feature maps.
method Divides feature maps into uneven-sized subtensors, compresses and stores them in a compressed yet accessible format.
result Average 55% DRAM bandwidth reduction with minimal indexing overhead.
Accelerates Birkhoff projection for manifold-constrained hyper-connections with high accuracy and speed.
problem Inaccurate and slow Birkhoff projection in mHC implementations.
method Dual formulation, Newton's method, implicit differentiation, warp-level CUDA kernel.
result Substantial speedups and accuracy improvements in doubly stochastic projections.
Bayesian optimization identifies optimal alloy formulations.
problem Accelerated discovery in materials science with autonomous systems.
method Bayesian optimization over problem formulation space.
result Framework converges on optimal alloy formulations.
Graph neural networks speed up nonnegative matrix factorization.
problem Efficiently factorize nonnegative matrices for various applications.
method Developed a graph neural network that combines bipartite self-attention with ADMM updates.
result Significant acceleration achieved in nonnegative matrix factorization.
Poor (even random) starting points for learning/training/optimization are common in machine learning. In many settings, the method of Robbins and Monro (online stochastic gradient descent) is known to be optimal for good starting points, but may not be optimal for poor starting points -- indeed, for poor starting point…
In this paper, we propose a novel progressive parameter pruning method for Convolutional Neural Network acceleration, named Structured Probabilistic Pruning (SPP), which effectively prunes weights of convolutional layers in a probabilistic manner. Unlike existing deterministic pruning approaches, where unimportant weig…
A new accelerated method with simpler momentum update rules.
problem Optimizing parameters in machine learning models.
method Proposes a novel accelerated stochastic gradient method with simpler momentum update rules.
result The method outperforms Sgdm and Adam in practical problems.
Accelerates deep neural network training with a generalized BN approach.
problem Conventional Batch Normalization (BN) struggles with convergence speed and error rate.
method Introduces Generalized Batch Normalization (GBN) using alternative deviation measures and statistics.
result GBN accelerates training and often improves error rate compared to conventional BN.
ORIGAMI accelerates ML algorithms by splitting compute tasks between in-memory and off-chip accelerators.
problem Memory bandwidth bottleneck in ML processing.
method Heterogeneous in-memory accelerators and off-chip compute platform, pattern-matching for compute patterns, computation-splitting compiler.
result ORIGAMI outperforms state-of-the-art accelerators in performance and energy-efficiency.