Electrochemical random access memory (ECRAM) is an emerging three-terminal nonvolatile memory (NVM) with highly controllable channel conductance which is promising for use as an analog memory (or synapse) in analog in-memory computing (IMC) systems. Energy-efficient analog IMC computing is particularly desirable for power-constrained, high-radiation environments such as satellites. However, little is known about the suitability of ECRAM for use in a total ionizing dose (TID) environment. This work investigates the effect of Co-60 gamma radiation on the channel conductance and noise—two properties critical for analog IMC systems—of a TaOx-based ECRAM up to 17.3 Mrad(SiO2) for both low- and high-channel-conductance state devices. A transient increase in conductance is observed in response to radiation which consists of two elements: an immediate increase in conductivity due to photocurrent and a secondary increase in conductivity, which has a slower rise and saturation and can persist for hours after exposure. This secondary, persistent photoconductivity is attributed to charging caused by hole trapping. These transient effects would not likely occur in a space environment due to the low dose rate compared with this experiment. No permanent change is found in the low conductance state (LCS) following exposure and the minor shift in the high conductance change would be less significant than the regular retention decay in this state. A permanent increase in the random telegraph noise is observed, possibly due to increased traps created in the channel. This work demonstrates that TaOx-based ECRAM is suitable for use in spaceborne analog IMC systems that are subject to significant TID.
Next-generation space remote sensing systems may be equipped with imaging arrays that sense data at a rate that outstrips the processing capability of any computing hardware that can operate within a satellite’s power budget. This project developed novel convolutional and recurrent neural networks to detect and estimate point-like events amid clutter, and investigated their efficient and accurate implementation on analog in-memory computing systems that are 10-1000× more energy-efficient than digital processors. This project leveraged two memory devices at different levels of technological maturity: a large-scale analog computing prototype using commercial SONOS charge-trap memory, and electrochemical memory (ECRAM) with intrinsic radiation hardness. We experimentally demonstrated end-to-end analog processing of our neural networks on SONOS and characterized the radiation response of both SONOS and ECRAM. We advanced the state-of-the-art in ECRAM precision and reliability, and developed co-design methods to enable accurate long-term operation of SONOS analog accelerators in space radiation environments.
This poster describes progress in training a variety of neural network architectures on a typical detection task. The poster will be presented at the Sandia MLDL Workshop
Neuromorphic computing systems have been used for the processing of spatiotemporal video-like data, requiring the use of recurrent networks, while attempting to minimize power consumption by utilizing binary activation functions. However, previous work on binary activation networks has primarily focused on training of feed-forward networks due to difficulties in training recurrent binary networks. Spiking neural networks however have been successfully trained in recurrent networks, despite the fact that they operate with binary communication. Intrigued by this discrepancy, we design a generalized leaky-integrate and fire neuron which can be deconstructed to a binary activation unit, allowing us to investigate the minimal dynamics from a spiking network that are required to allow binary activation networks to be trained. We find that a subthreshold integrative membrane potential is the only requirement to allow an otherwise standard binary activation unit to be trained in a recurrent network. Investigating further the trained networks, we find that these stateful binary networks learn a soft reset mechanism by recurrent weights, allowing them to approximate the explicit reset of spiking networks.
Hydrogen diffusion in metals and alloys plays an important role in the discovery of new materials for fuel cell and energy storage technology. While analytic models use hand-selected features that have clear physical ties to hydrogen diffusion, they often lack accuracy when making quantitative predictions. Machine learning models are capable of making accurate predictions, but their inner workings are obscured, rendering it unclear which physical features are truly important. To develop interpretable machine learning models to predict the activation energies of hydrogen diffusion in metals and random binary alloys, we create a database for physical and chemical properties of the species and use it to fit six machine learning models. Our models achieve root-mean-squared errors between 98-119 meV on the testing data and accurately predict that elemental Ru has a large activation energy, while elemental Cr and Fe have small activation energies. By analyzing the feature importances of these fitted models, we identify relevant physical properties for predicting hydrogen diffusivity. While metrics for measuring the individual feature importances for machine learning models exist, correlations between the features lead to disagreement between models and limit the conclusions that can be drawn. Instead grouped feature importance, formed by combining the features via their correlations, agree across the six models and reveal that the two groups containing the packing factor and electronic specific heat are particularly significant for predicting hydrogen diffusion in metals and random binary alloys. This framework allows us to interpret machine learning models and enables rapid screening of new materials with the desired rates of hydrogen diffusion.
The ability to rapidly screen material performance in the vast space of high entropy alloys is of critical importance to efficiently identify optimal hydride candidates for various use cases. Given the prohibitive complexity of first principles simulations and large-scale sampling required to rigorously predict hydrogen equilibrium in these systems, we turn to compositional machine learning models as the most feasible approach to screen on the order of tens of thousands of candidate equimolar high entropy alloys (HEAs). Critically, we show that machine learning models can predict hydride thermodynamics and capacities with reasonable accuracy (e.g. a mean absolute error in desorption enthalpy prediction of ∼5 kJ molH2−1) and that explainability analyses capture the competing trade-offs that arise from feature interdependence. We can therefore elucidate the multi-dimensional Pareto optimal set of materials, i.e., where two or more competing objective properties can't be simultaneously improved by another material. This provides rapid and efficient down-selection of the highest priority candidates for more time-consuming density functional theory investigations and experimental validation. Various targets were selected from the predicted Pareto front (with saturation capacities approaching two hydrogen per metal and desorption enthalpy less than 60 kJ molH2−1) and were experimentally synthesized, characterized, and tested amongst an international collaboration group to validate the proposed novel hydrides. Additional top-predicted candidates are suggested to the community for future synthesis efforts, and we conclude with an outlook on improving the current approach for the next generation of computational HEA hydride discovery efforts.
The domain wall-magnetic tunnel junction (DW-MTJ) is a versatile device that can simultaneously store data and perform computations. These three-terminal devices are promising for digital logic due to their nonvolatility, low-energy operation, and radiation hardness. Here, we augment the DW-MTJ logic gate with voltage-controlled magnetic anisotropy (VCMA) to improve the reliability of logical concatenation in the presence of realistic process variations. VCMA creates potential wells that allow for reliable and repeatable localization of domain walls (DWs). The DW-MTJ logic gate supports different fanouts, allowing for multiple inputs and outputs for a single device without affecting the area. We simulate a systolic array of DW-MTJ multiply-accumulate (MAC) units with 4-bit and 8-bit precision, which uses the nonvolatility of DW-MTJ logic gates to enable fine-grained pipelining and high parallelism. The DW-MTJ systolic array provides comparable throughput and efficiency to state-of-the-art CMOS systolic arrays while being radiation-hard. These results improve the feasibility of using DW-based processors, especially for extreme-environment applications such as space.
The ASC program seeks to use machine learning to improve efficiencies in its stockpile stewardship mission. Moreover, there is a growing market for technologies dedicated to accelerating AI workloads. Many of these emerging architectures promise to provide savings in energy efficiency, area, and latency when compared to traditional CPUs for these types of applications — neuromorphic analog and digital technologies provide both low-power and configurable acceleration of challenging artificial intelligence (AI) algorithms. If designed into a heterogeneous system with other accelerators and conventional compute nodes, these technologies have the potential to augment the capabilities of traditional High Performance Computing (HPC) platforms [5]. This expanded computation space requires not only a new approach to physics simulation, but the ability to evaluate and analyze next-generation architectures specialized for AI/ML workloads in both traditional HPC and embedded ND applications. Developing this capability will enable ASC to understand how this hardware performs in both HPC and ND environments, improve our ability to port our applications, guide the development of computing hardware, and inform vendor interactions, leading them toward solutions that address ASC’s unique requirements.
This project evaluated the use of emerging spintronic memory devices for robust and efficient variational inference schemes. Variational inference (VI) schemes, which constrain the distribution for each weight to be a Gaussian distribution with a mean and standard deviation, are a tractable method for calculating posterior distributions of weights in a Bayesian neural network such that this neural network can also be trained using the powerful backpropagation algorithm. Our project focuses on domain-wall magnetic tunnel junctions (DW-MTJs), a powerful multi-functional spintronic synapse design that can achieve low power switching while also opening the pathway towards repeatable, analog operation using fabricated notches. Our initial efforts to employ DW-MTJs as an all-in-one stochastic synapse with both a mean and standard deviation didn’t end up meeting the quality metrics for hardware-friendly VI. In the future, new device stacks and methods for expressive anisotropy modification may make this idea still possible. However, as a fall back that immediately satisfies our requirements, we invented and detailed how the combination of a DW-MTJ synapse encoding the mean and a probabilistic Bayes-MTJ device, programmed via a ferroelectric or ionically modifiable layer, can robustly and expressively implement VI. This design includes a physics-informed small circuit model, that was scaled up to perform and demonstrate rigorous uncertainty quantification applications, up to and including small convolutional networks on a grayscale image classification task, and larger (Residual) networks implementing multi-channel image classification. Lastly, as these results and ideas all depend upon the idea of an inference application where weights (spintronic memory states) remain non-volatile, the retention of these synapses for the notched case was further interrogated. These investigations revealed and emphasized the importance of both notch geometry and anisotropy modification in order to further enhance the endurance of written spintronic states. In the near future, these results will be mapped to effective predictions for room temperature and elevated operation DW-MTJ memory retention, and experimentally verified when devices become available.
Analog computing has been widely proposed to improve the energy efficiency of multiple important workloads including neural network operations, and other linear algebra kernels. To properly evaluate analog computing and explore more complex workloads such as systems consisting of multiple analog data paths, system level simulations are required. Moreover, prior work on system architectures for analog computing often rely on custom simulators creating signficant additional design effort and complicating comparisons between different systems. To remedy these issues, this report describes the design and implementation of a flexible tile-based analog accelerator element for the Structural Simulation Toolkit (SST). The element focuses on heavily on the tile controller—an often neglected aspect of prior work—that is sufficiently versatile to simulate a wide range of different tile operations including neural network layers, signal processing kernels, and generic linear algebra operations without major constraints. The tile model also interoperates with existing SST memory and network models to reduce the overall development load and enable future simulation of heterogeneous systems with both conventional digital logic and analog compute tiles. Finally, both the tile and array models are designed to easily support future extensions as new analog operations and applications that can benefit from analog computing are developed.
