Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “optimization platform”

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 199 records · Page 11
All sourcesNASA NTRSDOE OSTI
DOE OSTI2023

ProvLight: Efficient Workflow Provenance Capture on the Edge-to-Cloud Continuum

Modern scientific workflows require hybrid infrastructures combining numerous decentralized resources on the IoT/Edge interconnected to Cloud/HPC systems (aka the Computing Continuum) to enable their optimized execution. Understanding and optimizing the performance of such complex Edge-to-Cloud workflows is challenging. Capturing the provenance of key performance indicators, with their related data and processes, may assist in understanding and optimizing workflow executions. However, the capture overhead can be prohibitive, particularly in resource-constrained devices, such as the ones on the IoT/Edge.To address this challenge, based on a performance analysis of existing systems, we propose ProvLight, a tool to enable efficient provenance capture on the IoT/Edge. We leverage simplified data models, data compression and grouping, and lightweight transmission protocols to reduce overheads. We further integrate ProvLight into the E2Clab framework to enable workflow provenance capture across the Edge-to-Cloud Continuum. This integration makes E2Clab a promising platform for the performance optimization of applications through reproducible experiments.We validate ProvLight at a large scale with synthetic workloads on 64 real-life IoT/Edge devices in the FIT IoT LAB testbed. Evaluations show that ProvLight outperforms state-of-the-art systems like ProvLake and DfAnalyzer in resource-constrained devices. ProvLight is 26—37x faster to capture and transmit provenance data; uses 5—7x less CPU; 2x less memory; transmits 2x less data; and consumes 2—2.5x less energy. ProvLight [1] and E2Clab [2] are available as open-source tools.

Rosendo, Daniel↗
DOE OSTI2024

Developing multi-gene CRISPRa/i programs to accelerate DBTL cycles in ABF hosts engineered for chemical production

This project developed and implemented a modular CRISPR activation and interference (CRISPRa/i) platform to accelerate strain optimization and pathway development for industrially relevant microbial hosts. By integrating multiplexed transcriptional perturbation tools with data-driven Design–Build–Test–Learn (DBTL) workflows, the team achieved reductions in cycle time and enhanced production of industrial aromatics, particularly 4-aminocinnamic acid (4-ACA), in Pseudomonas putida. Key accomplishments included: ● Development of a robust, tunable CRISPRa/i system in P. putida that enabled efficient multi-target gene regulation via guide RNA (gRNA) programs ● Completion of two full DBTL cycles, guided by machine learning (ML) models trained on transcriptomic and performance data, reducing engineering time by over 30% ● Optimization of multi-gene regulatory programs to balance expression of host and pathway modules, improve 4-ACA titers, and resolve metabolic bottlenecks ● Demonstration of system portability through a limited proof-of-concept extension in Acinetobacter baylyi, underscoring the generalizability of the approach ● Evaluation of strain performance on lignocellulosic biomass-derived substrates, demonstrating the feasibility of converting renewable carbon into aromatic building blocks These results illustrate the feasibility of applying ML-guided CRISPRa/i perturbation strategies to accelerate strain development in complex microbial systems. The resulting tools and datasets contribute to DOE objectives by improving platform predictability, reducing development costs, and enabling broader access to sustainable, economically viable bioproduction technologies.

09 BIOMASS FUELS↗
DOE OSTI2022

Performance Evaluation of an Advanced Distributed Energy Resource Management Algorithm

This paper presents performance evaluation of a new distributed energy resource management system (DERMS) algorithm via an advanced hardware-in-the-loop (HIL) platform. The HIL platform provides realistic testing in a laboratory environment, including the accurate modeling of sub-transmission and distribution networks, the DERMS software controller, and 84 power hardware solar photovoltaic (PV) inverters, standard communication protocols, and a capacitor bank controller. The DERMS algorithm is also called, Grid-Optimization of Solar (GO-Solar) platform which includes predictive state estimation (PSE) and online multiple objective optimization (OMOO) to dispatch the legacy devices and distributed energy resources (e.g., PV). The voltage regulation performance is evaluated under three scenarios, volt-var smart inverter (baseline), and DERMS control for 100% and 30% of PV. The results show that controlling 30% of PV systems with the GO-Solar platform may provide the best balance of control performance and implementation cost.

distributed energy resource management system (DER↗
DOE OSTI2025

Web-Based Tools for Data-Informed Remedy Optimization: Software Theory and User Guide

This report documents the development and application of two web-based decision-support tools for pump-and-treat (P&T) groundwater remediation systems: PTOLEMY (Pump-and-Treat Optimized Location Evaluation to Maximize Yields) and OPTIMA (Optimization for Pump-and-Treat Implementation, Management, & Assessment). These tools enhance remedy design and management by leveraging advanced computational methods – specifically deep learning and multi-objective optimization – within a user-friendly platform. By integrating data-driven models with established hydrogeological knowledge, PTOLEMY and OPTIMA enable more efficient evaluation of well placement and operational strategies, helping site managers balance multiple remediation objectives under complex conditions. Both tools are implemented as modules within the SOCRATES (Suite Of Comprehensive Rapid Analysis Tools for Environmental Sites) web platform, which provides data access, visualization, and analytics to support remedy optimization across sites in the U.S. Department of Energy Office of Environmental Management complex. PTOLEMY is a rapid screening module designed to identify promising locations for new extraction wells. It employs a multi-channel three-dimensional convolutional neural network (MC3D-CNN) trained on high-fidelity simulation data to predict the relative performance (in terms of contaminant mass recovery) of potential well sites. Through an interactive web interface, PTOLEMY visualizes the probability of high performance across a site, highlighting areas where an extraction well is likely to yield above-threshold contaminant removal over a multi-year period. PTOLEMY’s map-based displays and exportable results support transparent communication of screening analyses. By focusing attention on the most favorable candidate locations, the tool augments traditional engineering judgment and physics-based modeling, providing a data informed basis for subsequent detailed evaluations. OPTIMA is a multi objective optimization module designed to find wellfield layouts and operating schedules that meet various cleanup goals. It quickly evaluates thousands of candidate setups – combinations of well locations, timing, and rates – and returns a small set of best trade-off options for comparison. At its core, OPTIMA uses a U-Net-based surrogate model – a deep-learning emulator of a groundwater flow and transport simulator – to dramatically accelerate scenario evaluations. Coupling this fast surrogate with the NSGA-II (Non-dominated Sorting Genetic Algorithm II) evolutionary algorithm, OPTIMA explores a wide decision space of well locations and schedules to identify Pareto-optimal solutions that trade off key objectives (e.g., minimizing cleanup time, maximizing contaminant mass removal, and minimizing plume extent). The tool outputs a family of optimal configurations and visualizes their trade-offs (Pareto frontiers of cleanup metrics and maps of optimized well placements). Site managers can use these results to understand the range of viable strategies and to select candidate designs for more detailed verification. OPTIMA is currently under active development and not yet fully released; this guide provides early documentation to support planning and gather user feedback.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2024

Highly Efficient Room‐Temperature Spin‐Orbit‐Torque Switching in a Van der Waals Heterostructure of Topological Insulator and Ferromagnet

Abstract All‐Van der Waals (vdW)‐material‐based heterostructures with atomically sharp interfaces offer a versatile platform for high‐performing spintronic functionalities at room temperature. One of the key components is vdW topological insulators (TIs), which can produce a strong spin‐orbit‐torque (SOT) through the spin‐momentum locking of their topological surface state (TSS). However, the relatively low conductance of the TSS introduces a current leakage problem through the bulk states of the TI or the adjacent ferromagnetic metal layers, reducing the interfacial charge‐to‐spin conversion efficiency ( q ICS ). Here, a vdW heterostructure is used consisting of atomically‐thin layers of a bulk‐insulating TI Sn‐doped Bi 1.1 Sb 0.9 Te 2 S 1 and a room‐temperature ferromagnet Fe 3 GaTe 2, to enhance the relative current ratio on the TSS up to ≈20%. The resulting q ICS reaches ≈1.65 nm −1 and the critical current density J c ≈0.9 × 10 6 Acm −2 at 300 K, surpassing the performance of TI‐based and heavy‐metal‐based SOT devices. These findings demonstrate that an all‐vdW heterostructure with thickness optimization offers a promising platform for efficient current‐controlled magnetization switching at room temperature.

2D ferromagnet↗
DOE OSTI2021

Island scanning pattern optimization for residual deformation mitigation in laser powder bed fusion via sequential inherent strain method and sensitivity analysis

Laser powder bed fusion (L-PBF) has emerged as one of the mainstream additive manufacturing approaches for fabricating metal parts with complex geometries and intricate internal structures. However, large deformation associated with rapid heating and cooling can lead to build failure and requires post-processing which may increase manufacturing cost and prolong the production period. Here in this work, an island scanning pattern design method is proposed to optimize the scanning direction of each island in order to reduce part deformation after cutting off the build platform. The objective of this optimization is to minimize the upward bending of the part after sectioning, which allows the part deformation to satisfy the tolerance requirement or reduce the post heat treatment time. Inherent strain method is employed in the sequential finite element analysis consisting of layer-by-layer activations and sectioning for fast residual distortion prediction. Full sequential sensitivity analysis for the formulated optimization is provided to update the island scanning directions. To show the feasibility and effectiveness of the proposed method, the scanning patterns of a block structure and a connecting rod were designed and parts were fabricated using an open architecture L-PBF machine. The fabrication experiments demonstrated that the residual deformation of both parts fabricated by optimized scanning pattern can be reduced by over 50% compared to the initial scanning patterns, which demonstrate the effectiveness of the proposed method.

36 MATERIALS SCIENCE↗
DOE OSTI2016

Economic optimization of operations for hybrid energy systems under variable markets

We prosed a hybrid energy systems (HES) which is an important element to enable increasing penetration of clean energy. Our paper investigates the operations flexibility of HES, and develops a methodology for operations optimization for maximizing economic value based on predicted renewable generation and market information. A multi-environment computational platform for performing such operations optimization is also developed. In order to compensate for prediction error, a control strategy is accordingly designed to operate a standby energy storage element (ESE) to avoid energy imbalance within HES. The proposed operations optimizer allows systematic control of energy conversion for maximal economic value. Simulation results of two specific HES configurations are included to illustrate the proposed methodology and computational capability. These results demonstrate the economic viability of HES under proposed operations optimizer, suggesting the diversion of energy for alternative energy output while participating in the ancillary service market. Economic advantages of such operations optimizer and associated flexible operations are illustrated by comparing the economic performance of flexible operations against that of constant operations. Sensitivity analysis with respect to market variability and prediction error, are also performed.

29 ENERGY PLANNING, POLICY, AND ECONOMY↗
DOE OSTI2016

Towards a Scalable and Adaptive Application Support Platform for Large-Scale Distributed E-Sciences in High-Performance Network Environments

The advent of large-scale collaborative scientific applications has demonstrated the potential for broad scientific communities to pool globally distributed resources to produce unprecedented data acquisition, movement, and analysis. System resources including supercomputers, data repositories, computing facilities, network infrastructures, storage systems, and display devices have been increasingly deployed at national laboratories and academic institutes. These resources are typically shared by large communities of users over Internet or dedicated networks and hence exhibit an inherent dynamic nature in their availability, accessibility, capacity, and stability. Scientific applications using either experimental facilities or computation-based simulations with various physical, chemical, climatic, and biological models feature diverse scientific workflows as simple as linear pipelines or as complex as a directed acyclic graphs, which must be executed and supported over wide-area networks with massively distributed resources. Application users oftentimes need to manually configure their computing tasks over networks in an ad hoc manner, hence significantly limiting the productivity of scientists and constraining the utilization of resources. The success of these large-scale distributed applications requires a highly adaptive and massively scalable workflow platform that provides automated and optimized computing and networking services. This project is to design and develop a generic Scientific Workflow Automation and Management Platform (SWAMP), which contains a web-based user interface specially tailored for a target application, a set of user libraries, and several easy-to-use computing and networking toolkits for application scientists to conveniently assemble, execute, monitor, and control complex computing workflows in heterogeneous high-performance network environments. SWAMP will enable the automation and management of the entire process of scientific workflows with the convenience of a few mouse clicks while hiding the implementation and technical details from end users. Particularly, we will consider two types of applications with distinct performance requirements: data-centric and service-centric applications. For data-centric applications, the main workflow task involves large-volume data generation, catalog, storage, and movement typically from supercomputers or experimental facilities to a team of geographically distributed users; while for service-centric applications, the main focus of workflow is on data archiving, preprocessing, filtering, synthesis, visualization, and other application-specific analysis. We will conduct a comprehensive comparison of existing workflow systems and choose the best suited one with open-source code, a flexible system structure, and a large user base as the starting point for our development. Based on the chosen system, we will develop and integrate new components including a black box design of computing modules, performance monitoring and prediction, and workflow optimization and reconfiguration, which are missing from existing workflow systems. A modular design for separating specification, execution, and monitoring aspects will be adopted to establish a common generic infrastructure suited for a wide spectrum of science applications. We will further design and develop efficient workflow mapping and scheduling algorithms to optimize the workflow performance in terms of minimum end-to-end delay, maximum frame rate, and highest reliability. We will develop and demonstrate the SWAMP system in a local environment, the grid network, and the 100Gpbs Advanced Network Initiative (ANI) testbed. The demonstration will target scientific applications in climate modeling and high energy physics and the functions to be demonstrated include workflow deployment, execution, steering, and reconfiguration. Throughout the project period, we will work closely with the science communities in the fields of climate modeling and high energy physics including Spallation Neutron Source (SNS) and Large Hadron Collider (LHC) projects to mature the system for production use.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2023

Sustainable ammonia synthesis from nitrogen wet with sea water by single-step plasma catalysis

Ammonia synthesis at ambient conditions employing intermittent distributed green sources of energy and feedstocks is globally sought to replace the centralized Haber-Bosch (H-B) process operating at high temperature and pressure. We report herein for the first time an effective and sustainable ammonia synthesis pathway from N 2 wet with seawater vapor over spherical SiO 2 and M/SiO 2 (M: Ag, Cu, and Co) catalysts driven by non-thermal plasma (NTP). Experimental results indicate that the presence of a catalyst is required for ammonia production from seawater vapor and N 2 . The Co/SiO 2 catalyst delivered the highest ammonia synthesis rate (r NH 3 ) of 3.7 mmol g cat -1 h -1 and energy yield of 3.2 g NH 3 ∙kW -1 ∙h -1 at a relatively low input power of 2 W. The extraction of H atoms from H 2 O molecules plays an important role in the ammonia synthesis from seawater vapor. Finally, this work unfolds a novel platform for the subsequent optimization of sustainable ammonia production from endless resources such as seawater and N 2 through catalytic non-thermal plasma potentially powered by renewable sources.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2022

GPU-accelerated DNS of compressible turbulent flows

Here, this paper explores strategies to transform an existing CPU-based high-performance computational fluid dynamics solver, HyPar, for compressible flow simulations on emerging exascale heterogeneous (CPU+GPU) computing platforms. The scientific motivation for developing a GPU-enhanced version of HyPar is to simulate canonical turbulent flows at the highest resolution possible on such platforms. We show that optimizing memory operations and thread blocks results in 200x speedup of computationally intensive kernels compared with a CPU core. Using multiple GPUs and CUDA-aware MPI communication, we demonstrate both strong and weak scaling of our GPU-based HyPar implementation on the NVIDIA Volta V100 GPUs. We simulate the decay of homogeneous isotropic turbulence in a triply periodic box on grids with up to 1024 3 points (5.3 billion degrees of freedom) and on up to 1,024 GPUs. We compare the wall times for CPU-only and CPU+GPU simulations. The results presented in the paper are obtained on the Summit and Lassen supercomputers at Oak Ridge and Lawrence Livermore National Laboratories, respectively.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2022

A high-fidelity building performance simulation test bed for the development and evaluation of advanced controls

We present an open-source building performance simulation test bed, the Advanced Controls Test Bed (ACTB), that interfaces high-fidelity Spawn of EnergyPlus building models, with advanced controllers implemented in Python. Additionally, the ACTB leverages the Building Optimization Testing and Alfalfa platforms for managing simulations, providing an external clock, a representational state transfer (REST) application programming interface (API), and key performance indicators for evaluating the effectiveness of control strategies. The REST API allows the development of external controllers programmed in languages such as Python, which provides flexibility and a rich choice of scientific libraries for designing control sequences. We present three test cases based on the U.S. Department of Energy's Reference Small Office Building to demonstrate the ACTB's capabilities: (a) rule-based controls compliant with ASHRAE Guideline 36 control sequences; (b) an economic model predictive control implemented using do-mpc; and (c) a deep Q-network reinforcement learning agent implemented using OpenAI Gym.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗
DOE OSTI2024

Flexible silicon photonic architecture for accelerating distributed deep learning

The increasing size and complexity of deep learning (DL) models have led to the wide adoption of distributed training methods in datacenters (DCs) and high-performance computing (HPC) systems. However, communication among distributed computing units (CUs) has emerged as a major bottleneck in the training process. In this study, we propose Flex-SiPAC, a flexible silicon photonic accelerated compute cluster designed to accelerate multi-tenant distributed DL training workloads. Flex-SiPAC takes a co-design approach that combines a silicon photonic hardware platform with a tailored collective algorithm, optimized to leverage the unique physical properties of the architecture. The hardware platform integrates a novel wavelength-reconfigurable transceiver design and a micro-resonator-based wavelength-reconfigurable switch, enabling the system to achieve flexible bandwidth steering in the wavelength domain. The collective algorithm is designed to support reconfigurable topologies, enabling efficient all-reduce communications that are commonly used in DL training. The feasibility of the Flex-SiPAC architecture is demonstrated through two testbed experiments. First, an optical testbed experiment demonstrates the flexible routing of wavelengths by shuffling an array of input wavelengths using a custom-designed spatial-wavelength selective switch. Second, a four-GPU testbed running two DL workloads shows a 23% improvement in job completion time compared to a similarly sized leaf-spine topology. We further evaluate Flex-SiPAC using large-scale simulations, which show that Flex-SiPAC is able to reduce the communication time by 26% to 29% compared to state-of-the-art compute clusters under representative collective operations.

Wu, Zhenguo (ORCID:0000000322847985)↗
DOE OSTI2020

Elucidating algal-bacterial community interactions by tracking volatile biomarkers (Final Report)

This feasibility study project aimed to lay further groundwork for studying and understanding how microorganisms interact with each other at the molecular level using model algal-bacterial co-cultures. A better understanding of such interactions is critical to optimizing and operating biotic platforms for a number of applications, including bioenergy, algal bioproducts and agriculture, and carbon capture. Here we explored the feasibility of characterizing and following these interactions by identifying and tracking volatile metabolites and biomarkers in the complex gaseous headspace of such cultures. We set up an experimental platform using a co-culture of Phaeodactylum tricornutum and Marinobacter subspecies 3-2 as a model system, used solid-phase microextraction fibers for volatile collection at various time points throughout culture growth, and employed gas chromatography mass spectrometry-based instrumentation available at LLNL for compound identification and quantification. We found that it is feasible to determine volatile profiles from the growth of Phaeodactylum tricornutum and Marinobacter subspecies 3-2 using our methodology and monitor profile changes over time. We also observed that volatiles distinguishing between cultures of the individual species and the co-cultures could be detected. We expect that results of this project will feed into the ongoing LLNL’s DOE BER funded Biofuels Scientific Focus Area (SFA) on algal-bacterial interactions and pave the way for future expanded volatile studies supported by DOE. The approach and technologies developed here would also be applicable and transferable to other microbial communities under study, such as cyanobacteria, rhizosphere communities and biofilms.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2020

Improving the performance of medical imaging applications using SYCL

As opposed to the Open Computing Language (OpenCL) programming model, in which host and device codes are generally written in different programming languages, SYCL can combine host and device codes for an application in a type-safe way to improve development productivity and performance portability. Hence, this report shows the experimental results of applying the SYCL programming model to medical imaging applications for a study on performance portability and programming productivity between OpenCL and SYCL. Rodinia is a widely used open-source benchmark suite for heterogeneous computing. We choose two medical imaging applications (Heart Wall and Particle Filter) in the benchmark suite, migrate the OpenCL implementations of the applications to the SYCL implementations, and evaluate their performance and productivity on Intel® microprocessors that contain a central processing unit (CPU) and an integrated graphics processing unit (GPU). The maturing SYCL compilers, which are based on a conformant implementation of the SYCL 1.2.1 Khronos specification, have been actively optimized for Intel® computing platforms. The experimental results are promising in terms of the raw performance and productivity. Although the SYCL implementation of the Heart Wall application does not execute successfully on a CPU, the SYCL implementations of the application can achieve comparable or better performance than the OpenCL implementation on Intel® integrated GPUs. For the Particle Filter application, the performance difference between the SYCL and OpenCL implementations is comparable on the GPUs for most cases, but the SYCL implementations are on average 4X faster than the OpenCL implementations on Intel® Xeon® CPUs. For programming productivity, we arguably use lines of code as a way to measure programming productivity in software. The SYCL programs reduce the lines of code of the OpenCL programs by 52% and 38% for the Heart Wall and Particle Filter, respectively. The results indicate that SYCL is a promising programming model for heterogeneous computing with the maturing compilers. We organize the remainder of the report as follows. Section II introduces the SYCL programming model, compares the major differences between an OpenCL program and a SYCL program, and describes the characteristics of the two applications. Section III describes the SYCL programming model in more details, and shows the SYCL implementation of a kernel function in the Particle Filter application as an example. In Section IV, we evaluate the performance of the applications on the CPUs and GPUs. Section V concludes the report.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2020

Defect Kinetics and Control for Module Reliability

Potential induced degradation is currently one of the most important module degradation mechanisms. It has been suggested that stacking faults decorated with sodium from the module glass are responsible for this effect and authors have also shown the reversibility of this effect upon reverse biasing of the module. The importance of sodium in the failure mechanism is clear, however, little is known regarding the factors that control its diffusion into the wafer, making it nearly impossible to predict the performance of a given module and engineer it to be better. Sodium migration from module glass into silicon cells and the resulting module degradation is a clear example of how defect kinetics can determine overall module performance and long-term reliability. To the detriment of the industry and its bankability, no quantitative models yet exist to predict defect-assisted module degradation, limiting the progress in improving reliability. In particular, the understanding of defect behavior under high electric fields, under stresses imparted by encapsulation or temperature, and under real operating conditions over long periods of time is a crucial gap in the current state-of-the-art. In this work we developed a Defect-Device-Degradation model to predict defect behavior and its impact on device performance over the module operational lifetime using experimentally-determined defect parameterizations. The validated model will provide a platform for manufacturing process optimization across input materials and architectures to avoid deleterious defects upstream and enable enhanced module robustness.

36 MATERIALS SCIENCE↗
DOE OSTI2016

Final Scientific Report

Project 1: We have created a 3D optical research and design software platform for simulation and optimization, geared toward asymmetric, folded optical systems and new, enabling freeform surfaces. The software, Eikonal+, targets both institutional researchers and leading optical surface fabricators. With a modular design and the source code available to the development team at the University of Rochester, custom modules can be created for specific research interests and is accelerating the work on freeform optics currently being carried out at the Institute of Optics. With a research-based optical design environment, the fabrication, assembly, and testing industries can anticipate, innovate, and retool for the future of optical systems. Targeted proposals for science and innovation in freeform optics spanning design to fabrication, assembly, and testing can proceed with a level of technical transparency that has been unachievable in this field since the 1960’s, when optics design code was commercialized and became unavailable to the research community for competitive reasons. Project 2: The University of Rochester Laboratory for Laser Energetics (LLE) with personnel from Flint Creek Resources (FCR) collaborated to develop technologies for the reclamation and reuse of cerium oxide based slurries intended for the polishing of optical components. The pilot process was evaluated and modifications were made to improve the collection of spent glass polish, to improve the efficiency and capacity of the recycling equipment, and to expand the customer base. A portable, self-contained system was developed and fabricated to recycle glass polishing compounds where the spent materials are produced.

36 MATERIALS SCIENCE↗
DOE OSTI2020

Enabling Highly Efficient Capsule Networks Processing Through A PIM-Based Architecture Design

In recent years, the CNNs have achieved great successes in the image processing tasks, e.g., image recognition and object detection. Unfortunately, traditional CNN's classication is found to be easily misled by increasingly complex image features due to the usage of pooling operations, hence unable to preserve accurate position and pose information of the objects. To address this challenge, a novel neural network structure called Capsule Network has been proposed, which introduces equivariance through capsules to signicantly enhance the learning ability for image segmentation and object detection. Due to its requirement of performing a high volume of matrix operations, CapsNets have been generally accelerated on modern GPU platforms that provide highly optimized software library for common deep learning tasks. However, based on our performance characterization on modern GPUs, CapsNets exhibit low effciency due to the special program and execution features of their routing procedure, including massive unshareable intermediate variables and intensive syn- chronizations, which are very dicult to optimize at software level. To address these challenges, we propose a hybrid computing architecture design named PIM-CapsNet. It preserves GPU's on-chip computing capability for accelerating CNN types of layers in CapsNet, while pipelining with an off-chip in-memory acceleration solution that effectively tackles routing procedure's ineffciency by leveraging the processing-in-memory capability of today's 3D stacked memory. Using routing procedure's inherent parallellization feature, our design enables hierarchical improvements on CapsNet inference effciency through minimizing data movement and maximizing parallel processing in memory. Evaluation results demonstrate that our proposed design can achieve substantial improvement on both performance and energy savings for CapsNet inference, with almost zero accuracy loss. The results also suggest good performance scalability in optimizing the routing procedure with increasing network size.

Zhang, Xingyao↗