Search NASA⌕ Search

SEARCH · Search NASA

Results for “compressive neural network”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 37 records · Page 2
All sourcesNASA NTRSDOE OSTI
DOE OSTI2021

AutoAtlas: Neural Network for 3D Unsupervised Partitioning and Representation Learning

Here we present a novel neural network architecture called AutoAtlas for fully unsupervised partitioning and representation learning of 3D brain Magnetic Resonance Imaging (MRI) volumes. AutoAtlas consists of two neural network components: one neural network to perform multi-label partitioning based on local texture in the volume, and a second neural network to compress the information contained within each partition. We train both of these components simultaneously by optimizing a loss function that is designed to promote accurate reconstruction of each partition, while encouraging spatially smooth and contiguous partitioning, and discouraging relatively small partitions. We show that the partitions adapt to the subject specific structural variations of brain tissue while consistently appearing at similar spatial locations across subjects. AutoAtlas also produces very low dimensional features that represent local texture of each partition. We demonstrate prediction of metadata associated with each subject using the derived feature representations and compare the results to prediction using features derived from FreeSurfer anatomical parcellation. Since our features are intrinsically linked to distinct partitions, we can then map values of interest, such as partition-specific feature importance scores onto the brain for visualization.

42 ENGINEERING↗
DOE OSTI2023

In-pixel AI for lossy data compression at source for X-ray detectors

Integrating neural networks for data compression directly in the Read-Out Integrated Circuits (ROICs), i.e. the pixelated front-end, would result in a significant reduction in off-chip data transfer, overcoming the I/O bottleneck. Our ROIC test chip (AI-In-Pixel-65) is designed in a 65 nm Low Power CMOS process for the readout of pixelated X-ray detectors. Each pixel consists of an analog front-end for signal processing and a 10b analog-to-digital converter operating at 100KSPS. Here, we compare two non-reconfigurable techniques, Principal Component Analysis (PCA) and an AutoEncoder (AE) as lossy data compression engines implemented within the pixelated area. The PCA algorithm achieves 50$×$ compression, adds one clock cycle latency, and results in a 21% increase in the pixel area. The AE achieves 70$×$ compression, adds 30 clock cycle latency, and results in a similar area increase.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗
DOE OSTI2025

Exploring Black-box Adversarial Attacks on Low-rank Constrained Neural Networks

Low-rank compression has been shown as an effective tool to reduce parameter counts of convolutional and vision transformer architectures; however, low-rank training often reduces model robustness to adversarial perturbations. In this work, we explore the effects of low-rank training on black-box attacks, where attacked images are generated without knowledge of the low-rank parameters. We find that low-rank training is not sufficient as a black-box defense and can sometimes produce worse than expected as compared to baseline models. Influencing the spectrum of the low-rank models during training, which is known to increase model robustness against white-box attacks, improves black-box performance as well.

Schnake, Stefan [ORNL] (ORCID:0000000215183538)↗
DOE OSTI2022

Towards Compact Neural Networks via End-to-End Training: A Bayesian Tensor Approach with Automatic Rank Determination

Post-training model compression can reduce the inference costs of deep neural networks, but uncompressed training still consumes enormous hardware resources and energy. To enable low-energy training on edge devices, it is highly desirable to directly train a compact neural network from scratch with a low memory cost. Low-rank tensor decomposition is an effective approach to reduce the memory and computing costs of large neural networks. However, directly training low-rank tensorized neural networks is a very challenging task because it is hard to determine a proper tensor rank a priori, and the tensor rank controls both model complexity and accuracy. Here, this paper presents a novel end-to-end framework for low-rank tensorized training. We first develop a Bayesian model that supports various low-rank tensor formats (e.g., CANDECOMP/PARAFAC, Tucker, tensor-train, and tensor-train matrix) and reduces neural network parameters with automatic rank determination during training. Then we develop a customized Bayesian solver to train large-scale tensorized neural networks. Our training methods shows orders-of-magnitude parameter reduction and little accuracy loss (or even better accuracy) in the experiments. On a very large deep learning recommendation system with over 4.2 ×10 9 model parameters, our method can reduce the parameter number to 1.6 ×10 5 automatically in the training process (i.e., by 2.6 ×10 4 times) while achieving almost the same accuracy. Code is available at https://github.com/colehawkins/bayesian-tensor-rank-determination.

compact neural networks↗
DOE OSTI2021

A Reconfigurable Neural Network ASIC for Detector Front-End Data Compression at the HL-LHC

Despite advances in the programmable logic capabilities of modern trigger systems, a significant bottleneck remains in the amount of data to be transported from the detector to off-detector logic where trigger decisions are made. We demonstrate that a neural network (NN) autoencoder model can be implemented in a radiation-tolerant application-specific integrated circuit (ASIC) to perform lossy data compression alleviating the data transmission problem while preserving critical information of the detector energy profile. For our application, we consider the high-granularity calorimeter from the Compact Muon Solenoid (CMS) experiment at the CERN Large Hadron Collider. The advantage of the machine learning approach is in the flexibility and configurability of the algorithm. By changing the NN weights, a unique data compression algorithm can be deployed for each sensor in different detector regions and changing detector or collider conditions. To meet area, performance, and power constraints, we perform quantization-aware training to create an optimized NN hardware implementation. The design is achieved through the use of high-level synthesis tools and the hls4ml framework and was processed through synthesis and physical layout flows based on a low-power (LP)-CMOS 65-nm technology node. The flow anticipates 200 Mrad of ionizing radiation to select gates and reports a total area of 3.6 mm 2 and consumes 95 mW of power. The simulated energy consumption per inference is 2.4 nJ. Furthermore, this is the first radiation-tolerant on-detector ASIC implementation of an NN that has been designed for particle physics applications.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗
DOE OSTI2025

Neural architecture codesign for fast physics applications

We develop a pipeline to streamline neural architecture codesign for physics applications to reduce the need for ML expertise when designing models for novel tasks. Our method employs neural architecture search and network compression in a two-stage approach to discover hardware efficient models. This approach consists of a global search stage that explores a wide range of architectures while considering hardware constraints, followed by a local search stage that fine-tunes and compresses the most promising candidates. We exceed performance on various tasks and show further speedup through model compression techniques such as quantization-aware-training and neural network pruning. We synthesize the optimal models to high level synthesis code for FPGA deployment with the hls4ml library. Additionally, our hierarchical search space provides greater flexibility in optimization, which can easily extend to other tasks and domains. We demonstrate this with two case studies: Bragg peak finding in materials science and jet classification in high energy physics, achieving models with improved accuracy, smaller latencies, or reduced resource utilization relative to the baseline models.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗
DOE OSTI2025

Dynamic Low-Rank Training with Spectral Regularization: Achieving Robustness in Compressed Representations

Deployment of neural networks on resource-constrained devices demands models that are both compact and robust to adversarial inputs. However, compression and adversarial robustness often conflict. In this work, we introduce a dynamical low-rank training scheme enhanced with a novel spectral regularizer that controls the condition number of the low-rank core in each layer. This approach mitigates the sensitivity of compressed models to adversarial perturbations without sacrificing clean accuracy. The method is model- and data-agnostic, computationally efficient, and supports rank adaptivity to automatically compress the network at hand. Extensive experiments across standard architectures, datasets, and adversarial attacks show the regularized networks can achieve over 94\% compression while recovering or improving adversarial accuracy relative to uncompressed baselines.

Schotthoefer, Steffen [ORNL] (ORCID:00000002156965↗
DOE OSTI2020

CSB-RNN: A Faster-Than-Realtime RNN Acceleration Framework with Compressed Structured Blocks

Recurrent Neural Networks (RNN) is widely applied to temporal sequence analysis, where real-time performance is usually in demand. However, RNN suffers a heavy computational workload as the model comes with a large weight matrix. To alleviate the pain, model compression (pruning) schemes have been proposed for RNN that pruning the redundant (near-zero) weight-values. On the one hand, the non-structured pruning methods achieve a considerable pruning rate while bringing the computational irregularity, which is un-friendly to parallel-hardware. On the other hand, the existing structured pruning methods consider the hardware parallelism; However, they suffer a poor pruning rate due to the restrict constraints on pruning structure. This paper presents CSB-RNN, an optimized full-stack RNN framework with the novel compressed structured block (CSB) technique. The CSB-pruned RNN model comes with both fine-granularity that benefits the pruning rate and regular structure that facilitates the hardware-parallelism. Further, we propose a novel hardware architecture for inferencing the CSB-pruned model. Different from conventional parallel hardware, this architecture solves the block-workload imbalance issue and achieves an over 95% hardware utilization. With the experiments on 10 RNN models in 5 application domains, the CSB-RNN realizes 7×-20× lossless compression and up to 50× acceptable lossy-compression, which is 2×-7× to the prior art. With the addition of the novel hardware, the compressed-RNN inference reaches a super real-time latency of 10-400µs with FPGA implementation.

Shi, Runbin↗
DOE OSTI2021

All Optical Neural Networks for Low Power Edge Computing

We developed a simplistic physics-based model of an all-optical neural network that mimics the encoder part of an autoencoder neural network for image compression. Our approach relies on the generation of a MATLAB-based model for both data compression and decompression and utilizes MATLAB's built-in autoencoder networks in combination with simple propagation of optical fields between layers constituting phase elements via Fourier transform. We optimize the phase elements using the particle swarm optimization technique and using our model, we demonstrate a compression ratio of 25% for 2828-pixel input images containing numeric digits from 0 to 9.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2023

Differentiable Earth mover’s distance for data compression at the high-luminosity LHC

Abstract The Earth mover’s distance (EMD) is a useful metric for image recognition and classification, but its usual implementations are not differentiable or too slow to be used as a loss function for training other algorithms via gradient descent. In this paper, we train a convolutional neural network (CNN) to learn a differentiable, fast approximation of the EMD and demonstrate that it can be used as a substitute for computing-intensive EMD implementations. We apply this differentiable approximation in the training of an autoencoder-inspired neural network (encoder NN) for data compression at the high-luminosity LHC at CERN The goal of this encoder NN is to compress the data while preserving the information related to the distribution of energy deposits in particle detectors. We demonstrate that the performance of our encoder NN trained using the differentiable EMD CNN surpasses that of training with loss functions based on mean squared error.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗
DOE OSTI2024

DLSIA: Deep Learning for Scientific Image Analysis

DLSIA (Deep Learning for Scientific Image Analysis) is a Python-based machine learning library that empowers scientists and researchers across diverse scientific domains with a range of customizable convolutional neural network (CNN) architectures for a wide variety of tasks in image analysis to be used in downstream data processing. DLSIA features easy-to-use architectures, such as autoencoders, tunable U-Nets and parameter-lean mixed-scale dense networks (MSDNets). Additionally, this article introduces sparse mixed-scale networks (SMSNets), generated using random graphs, sparse connections and dilated convolutions connecting different length scales. For verification, several DLSIA-instantiated networks and training scripts are employed in multiple applications, including inpainting for X-ray scattering data using U-Nets and MSDNets, segmenting 3D fibers in X-ray tomographic reconstructions of concrete using an ensemble of SMSNets, and leveraging autoencoder latent spaces for data compression and clustering. As experimental data continue to grow in scale and complexity, DLSIA provides accessible CNN construction and abstracts CNN complexities, allowing scientists to tailor their machine learning approaches, accelerate discoveries, foster interdisciplinary collaboration and advance research in scientific image analysis.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2024

Distilling particle knowledge for fast reconstruction at high-energy physics experiments

Knowledge distillation is a form of model compression that allows artificial neural networks of different sizes to learn from one another. Its main application is the compactification of large deep neural networks to free up computational resources, in particular on edge devices. In this article, we consider proton-proton collisions at the High-Luminosity Large Hadron Collider (HL-LHC) and demonstrate a successful knowledge transfer from an event-level graph neural network (GNN) to a particle-level small deep neural network (DNN). Our algorithm, DistillNet, is a DNN that is trained to learn about the provenance of particles, as provided by the soft labels that are the GNN outputs, to predict whether or not a particle originates from the primary interaction vertex. The results indicate that for this problem, which is one of the main challenges at the HL-LHC, there is minimal loss during the transfer of knowledge to the small student network, while improving significantly the computational resource needs compared to the teacher. This is demonstrated for the distilled student network on a CPU, as well as for a quantized and pruned student network deployed on a field programmable gate array. Our study proves that knowledge transfer between networks of different complexity can be used for fast artificial intelligence (AI) in high-energy physics that improves the expressiveness of observables over non-AI-based reconstruction algorithms. Such an approach can become essential at the HL-LHC experiments, e.g. to comply with the resource budget of their trigger stages.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗
DOE OSTI2022

Latent Space Dynamics Identification

LaSDI is a data-driven physical simulation software that forms a latent space for a given high-fidelity model and discovers a set of ordinary differential equations for the latent space dynamics. It allows a fast and accurate solution process, which is useful for multi-query decision making applications, such as design optimization and uncertainty quantification. The performance of the LaSDI framework is demonstrated on four different problems, i.e., 1D and 2D Burgers equations, nonlinear heat conduction, and radial advection problems. Both linear and nonlinear compression techniques, such as neural network and proper orthogonal decomposition, are used to form a latent space. A concept of local dynamics identification procedure is introduced to enable a parametric model, which enhances the accuracy level over a given parameter space.

Fries, William↗
DOE OSTI2024

Symmetry-Based Structured Matrices for Efficient Approximately Equivariant Networks

There has been much recent interest in designing symmetry-aware neural networks (NNs) exhibiting relaxed equivariance. Such NNs aim to interpolate between being exactly equivariant and being fully flexible, affording consistent performance benefits. In a separate line of work, certain structured parameter matrices -- those with displacement structure, characterized by low displacement rank (LDR) -- have been used to design small-footprint NNs. Displacement structure enables fast function and gradient evaluation, but permits accurate approximations via compression primarily to classical convolutional neural networks (CNNs). In this work, we propose a general framework -- based on a novel construction of symmetry-based structured matrices -- to build approximately equivariant NNs with significantly reduced parameter counts. Our framework integrates the two aforementioned lines of work via the use of so-called Group Matrices (GMs), a forgotten precursor to the modern notion of regular representations of finite groups. GMs allow the design of structured matrices -- resembling LDR matrices -- which generalize the linear operations of a classical CNN from cyclic groups to general finite groups and their homogeneous spaces. We show that GMs can be employed to extend all the elementary operations of CNNs to general discrete groups. Further, the theory of structured matrices based on GMs provides a generalization of LDR theory focussed on matrices with cyclic structure, providing a tool for implementing approximate equivariance for discrete groups. We test GM-based architectures on a variety of tasks in the presence of relaxed symmetry. We report that our framework consistently performs competitively compared to approximately equivariant NNs, and other structured matrix-based compression frameworks, sometimes with a one or two orders of magnitude lower parameter count.

Samudre, Ashwin↗
DOE OSTI2021

Analyzing acoustic emission data to identify cracking modes in cement paste using an artificial neural network

This research is focused on the identification of cracking mechanisms for cement paste using acoustic emission data, recorded from compression and notched four-point bending tests. A procedure is developed for analyzing the data by employing an agglomerative hierarchical clustering method, an artificial neural network, and a ray-tracing source location algorithm. An agglomerative hierarchical clustering method is utilized to cluster the AE data from a compression test using frequency-dependent features. A neural network is trained using the compression test data and applied to the AE data emitted during the four-point bending test. The clustered data from the four-point bending test is localized using a ray-tracing algorithm. Based on the occurrence and locations of the clustered events and signal feature analyses, potential cracking mechanisms are identified and assigned.

36 MATERIALS SCIENCE↗