Probabilistic Interpolation Recoder for Energy-Error-Product Efficient DBNs With p-Bit Devices

Pourmeidani, Hossein; Sheikhfaal, Shadi; Zand, Ramin; DeMara, Ronald F.

doi:10.1109/tetc.2020.2965079

Cited by 16 publications

(6 citation statements)

References 31 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In this paper, we utilize a DBNs circuit-level implementation employing p-bit as activation functions and memristive crossbars as weighted connections which is called Probabilistic Inference Network-Simulator (PIN-Sim) [16,17]. As depicted in Figure 3, PIN-Sim is a hierarchical simulation framework composed of seven principal modules.…”

Section: Probabilistic Inference Network-simulatormentioning

confidence: 99%

High-Accuracy Deep Belief Network: Fuzzy Neural Networks Using MRAM-Based Stochastic Neurons

Pourmeidani¹,

DeMara

2021

IEEE J. Explor. Solid-State Comput. Devices Circuits

Self Cite

View full text Add to dashboard Cite

Two-terminal magnetoresistive random access memory (MRAM) devices provide a recent approach to intrinsically-realizing stochastic neuronal behavior in cognitive architectures such as restricted Boltzmann machines (RBMs) for deep belief networks (DBNs). MRAM-based DBNs have achieved substantial energy and area improvements in comparison with the prior DBN hardware implementations. However, MRAM-based DBNs suffer from low accuracy in image classification. In order to resolve this problem, we present a new DBN-fuzzy system based on the combination of MRAM-based DBNs and fuzzy systems in the interest of improving the accuracy of MRAM-based neural networks. First, the MRAM-based DBN is employed to identify the top recognition results with the highest probability. Second, the fuzzy system is utilized to obtain the top-1 recognition results. We assess the accuracy of neural networks on MNIST dataset, finding that the top-1 accuracy of the DBN-fuzzy neural network is improved from 64% to 82% in comparison to the MRAM-based DBNs for a 784 × 10 network.

show abstract

Section: Probabilistic Inference Network-simulatormentioning

confidence: 99%

High-Accuracy Deep Belief Network: Fuzzy Neural Networks Using MRAM-Based Stochastic Neurons

Pourmeidani¹,

DeMara

2021

IEEE J. Explor. Solid-State Comput. Devices Circuits

Self Cite

View full text Add to dashboard Cite

show abstract

“…With the rapid development of the learning method, it is possible to use learning techniques to represent the complicated nonlinear relationship of the outputs of the generator with its parameters [9], [10]. Recently, machine learning approaches have been applied for parameter calibration.…”

Section: Introductionmentioning

confidence: 99%

A Parallel Multi-Modal LSTM for Power System Model Calibration

Khazeiynasab¹

2022

Preprint

View full text Add to dashboard Cite

<div>Renewable energies, smart loads, energy storage, and new market behavior are adding new sources of uncertainty to power systems. Therefore, planning in real-time and developing high-quality models is crucial to adapt to uncertainties. Model validation based on actual measurements is necessary for obtaining accurate representations of power systems dynamics with system uncertainties. This paper presents a new measurement-based method to calibrate the parameters of a synchronous generator by deep learning method based on the long-short term memory (LSTM) network. First, critical parameters are determined regarding the active/reactive behavior of the generator. Then, a parallel multimodal LSTM (PM-LSTM) is designed with flexible input time steps to capture important features of temporal patterns from time-series measurements. The extracted features are then fed into a dense layer to capture the joint representation</div><div>of inputs. The simulations conducted for a hydro generator under different events show that the proposed method can estimate the model parameters accurately and efficiently.</div>

show abstract

“…In recent years, several studies theoretically and experimentally have investigated the usage of thermally-unstable MTJs with near-zero energy barrier based on superparamagnetic materials to realize probabilistic neuromorphic paradigms using functional spintronic devices [2]- [4]. For instance, [6][7] focus on leveraging p-bit devices in deep belief network (DBN) architectures. In this paper, we seek to examine the effects of process variation (PV) on the energy barrier of the p-bit devices and their consequent impacts on the accuracy and energy consumption of a representative neuromorphic DBN architecture.…”

Section: Introductionmentioning

confidence: 99%

Electrically-Tunable Stochasticity for Spin-based Neuromorphic Circuits: Self-Adjusting to Variation

Pourmeidani

Debashis

Chen

et al. 2020

2020 IEEE 63rd International Midwest Symposium on Circuits and Systems (MWSCAS)

Self Cite

View full text Add to dashboard Cite

Energy-efficient methods are addressed for leveraging low energy barrier nanomagnetic devices within neuromorphic architectures. Using a Magnetoresistive Random Access Memory (MRAM) probabilistic device (p-bit) as the basis of neuronal structures in Deep Belief Networks (DBNs), the impact of reducing the Magnetic Tunnel Junction's (MTJ's) energy barrier is assessed and optimized for the resulting stochasticity present in the learning system. This can mitigate the process variation sensitivity of stochastic DBNs which encounter a sharp drop-off when energy barriers exceed nearzero kT. As evaluated for the MNIST dataset for energy barriers at near-zero kT to 2.0 kT in increments of 0.5 kT, it is shown that the stability factor changes by 5 orders of magnitude. The selfcompensating circuit developed herein provides a compact, and low complexity approach to mitigating process variation impacts towards practical implementation and fabrication.

show abstract

Probabilistic Interpolation Recoder for Energy-Error-Product Efficient DBNs With p-Bit Devices

Cited by 16 publications

References 31 publications

High-Accuracy Deep Belief Network: Fuzzy Neural Networks Using MRAM-Based Stochastic Neurons

High-Accuracy Deep Belief Network: Fuzzy Neural Networks Using MRAM-Based Stochastic Neurons

A Parallel Multi-Modal LSTM for Power System Model Calibration

Electrically-Tunable Stochasticity for Spin-based Neuromorphic Circuits: Self-Adjusting to Variation

Contact Info

Product

Resources

About