Brain-inspired computing exploiting carbon nanotube FETs and resistive RAM: Hyperdimensional computing case study

Wu, Tung-Yu; Li, Haitong; Huang, Pei‐Chen; Rahimi, Abbas; Rabaey, Jan M.; Wong, H.-S. Philip; Shulaker, Max M.; Mitra, Subhasish

doi:10.1109/isscc.2018.8310399

Cited by 103 publications

(44 citation statements)

References 5 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…It has been shown that HD computing degrades very gracefully in the presence of temporary and permanent faults compared to a baseline KNN classifier for the language recognition task: by injecting the intermittent hardware-induced errors in both classifiers, HD computing tolerates 8.8× higher probability of failure per individual memory cells [24]; considering the permanent hard errors, HD computing tolerates 60× higher probability of failures [7]. The robust operation under low SNR conditions and high variability perfectly matches with emerging nanotechnologies promising to deliver substantial energy savings [7], [5], [6].…”

Section: E Energy Efficiencymentioning

confidence: 68%

“…An architecture based on HD computing can be seen as an extremely wide dataflow processor with small instruction set of bit-level operations. Further, logic can be tightly integrated with the memory and all computations are fully distributed that can save energy [6], [7]. This forms a fundamental departure from the traditional von Neumann architectures where data has to be transported to the processing unit and back, creating the infamous memory wall.…”

Section: E Energy Efficiencymentioning

confidence: 99%

“…Advances in learning-based computing for IoT increase energy-efficiency towards TOPS/Watt [4], but further improvement requires a novel look at data representations, associated operations, circuits, and materials and substrates that enable them [5]. Monolithic 3D integrated nanotechnologies [6], [7] combined with novel brain-inspired computational paradigms that support fast learning and fault tolerance could lead the way [5].…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Efficient Biosignal Processing Using Hyperdimensional Computing: Network Templates for Combined Learning and Classification of ExG Signals

et al. 2019

Self Cite

View full text Add to dashboard Cite

Recognizing the very size of the brain's circuits, hyperdimensional (HD) computing can model neural activity patterns with points in a HD space, that is, with HD vectors. Key examined properties of HD computing include: a versatile set of arithmetic operations on HD vectors, generality, scalability, analyzability, one-shot learning, and energy efficiency. These make it a prime candidate for efficient biosignal processing where signals are noisy and nonstationary, training data sets are not huge, individual variability is significant, and energy efficiency constraints are tight. Purely based on native HD computing operators, we describe a combined method for multiclass learning and classification of various ExG biosignals such as electromyography (EMG), electroencephalography (EEG), and electrocorticography (ECoG). We develop a full set of HD network templates that comprehensively encode body potentials and brain neural activity recorded from different electrodes into a single HD vector without requiring domain expert knowledge or ad-hoc electrode selection process. Such encoded HD vector is processed as a single unit for fast one-shot learning, and robust classification. It can be interpreted to identify the most useful features as well. Compared to state-of-the-art counterparts, HD computing enables online, incremental, and fast learning as it demands less than a third as much training data as well as less preprocessing.

show abstract

Section: E Energy Efficiencymentioning

confidence: 68%

Section: E Energy Efficiencymentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Efficient Biosignal Processing Using Hyperdimensional Computing: Network Templates for Combined Learning and Classification of ExG Signals

et al. 2019

Self Cite

View full text Add to dashboard Cite

show abstract

“…These make it well-suited for efficient biosignal processing [17], e.g., 2× lower energy at iso-accuracy when compared to a highly-optimized SVM on an ARM Cortex M4 [15]. Larger energy saving is achieved by using emerging 3D nanoscale devices [18], [19]. HD computing has been applied to various learning and multiclass inference tasks, such as the language identification [22], [23], EMG gesture recognition [14], [15], EEG-based brain-machine interfaces [16], and in general ExG processing [17].…”

Section: B Background Of Hd Computingmentioning

confidence: 99%

Hyperdimensional Computing-based Multimodality Emotion Recognition with Physiological Signals

Chang

Rahimi

Benini

et al. 2019

2019 IEEE International Conference on Artificial Intelligence Circuits and Systems (AICAS)

Self Cite

View full text Add to dashboard Cite

To interact naturally and achieve mutual sympathy between humans and machines, emotion recognition is one of the most important function to realize advanced human-computer interaction devices. Due to the high correlation between emotion and involuntary physiological changes, physiological signals are a prime candidate for emotion analysis. However, due to the need of a huge amount of training data for a high-quality machine learning model, computational complexity becomes a major bottleneck. To overcome this issue, brain-inspired hyperdimensional (HD) computing, an energy-efficient and fast learning computational paradigm, has a high potential to achieve a balance between accuracy and the amount of necessary training data. We propose an HD Computingbased Multimodality Emotion Recognition (HDC-MER). HDC-MER maps real-valued features to binary HD vectors using a random nonlinear function, and further encodes them over time, and fuses across different modalities including GSR, ECG, and EEG. The experimental results show that, compared to the best method using the full training data, HDC-MER achieves higher classification accuracy for both valence (83.2% vs. 80.1%) and arousal (70.1% vs. 68.4%) using only 1/4 training data. HDC-MER also achieves at least 5% higher averaged accuracy compared to all the other methods in any point along the learning curve.

show abstract

“…Despite these benefits, the reported system level demonstrations using CNFET technology have been primarily limited to PMOS-only digital logic implementations such as 3D Nano System (which demonstrated sensing, computing, and memory integrated all on the same chip) [7], and a hyperdimensional computing case study exploiting CNFETs and RRAM [10]. In addition, all reported CNFET CMOS demonstrations have been limited to only individual devices, small-scale circuits, or digital logic.…”

Section: Background and Motivationmentioning

confidence: 99%

29.8 SHARC: Self-Healing Analog with RRAM and CNFETs

Amer

Hills

et al. 2019

2019 IEEE International Solid- State Circuits Conference - (ISSCC)

Self Cite

View full text Add to dashboard Cite

Next-generation applications require processing on massive amount of data in realtime, exceeding the capabilities of electronic systems today. This has spurred research in a wide-range of areas: from new devices to replace silicon-based fieldeffect transistors (FETs) to new circuit and system architectures with fine-grained and dense integration of logic and memory. However, isolated improvements in just one area is insufficient. Rather, enabling these next-generation applications will require combining benefits across all levels of the computing stack: leveraging new devices to realize new circuits and architectures. For instance, carbon nanotube (CNT) field-effect transistors (CNFETs) for logic and Resistive Random-Access Memory (RRAM) for memory are two promising emerging nanotechnologies for energy-efficient electronics. However, CNFETs suffer from inherent imperfections (such as of metallic CNTs, m-CNTs), which have prohibited realizing large-scale CNFET circuits in the past. This work proposes a circuit design technique that integrates and combines the benefits of both CNFETs with RRAM to realize threedimensional (3D) circuits that are immune to m-CNTs. Leveraging this technique, we show the first experimental demonstration of CNFET-based analog mixed-signal circuits.

show abstract

Brain-inspired computing exploiting carbon nanotube FETs and resistive RAM: Hyperdimensional computing case study

Cited by 103 publications

References 5 publications

Efficient Biosignal Processing Using Hyperdimensional Computing: Network Templates for Combined Learning and Classification of ExG Signals

Efficient Biosignal Processing Using Hyperdimensional Computing: Network Templates for Combined Learning and Classification of ExG Signals

Hyperdimensional Computing-based Multimodality Emotion Recognition with Physiological Signals

29.8 SHARC: Self-Healing Analog with RRAM and CNFETs

Contact Info

Product

Resources

About