International Business Machines Corporation

Arizona State University

Columbia University

Intel Corporation

Other

Jae-sun Seo

Leland Chang

Robert K Montoye

Dharmendra S Modha

Daniel Joseph Friedman

Jose A Tierno

Bernard V Brezzo

Shihui Yin

John Vernon Arthur

Paul A Merolla

Bipin Rajendran

Steven Kyle Esser

Yong Liu

Mingoo Seok

Shimeng Yu

Zhewei Jiang

Gregory Kengho Chen

Albert M Young

Sarma Vrudhula

Minkyu Kim

Yu Cao

Sang Joon Kim

Ram K Krishnamurthy

Robert J Drost

Raghavan Kumar

Ronald Ho

Himanshu Kaul

Chisung Bae

Robert David Hopkins

Deepak Kadetotad

Thomas Chen

Chaitali Chakrabarti

Jieping Ye

Sai Kiran Cherupally

Bo Zhang

Sairam Arunachalam

A programmable in-memory computing (IMC) accelerator for low-precision deep neural network inference, also referred to as PIMCA, is provided. Embodiments of the PIMCA integrate a large number of capacitive-coupling-based IMC static random-access memory (SRAM) macros and demonstrate large-scale integration of IMC SRAM macros. For example, a 28 nm prototype integrates 108 capacitive-coupling-based IMC SRAM macros of a total size of 3.4 megabytes (Mb), demonstrating one of the largest IMC hardware to date. In addition, a custom instruction set architecture (ISA) is developed featuring IMC and single-instruction-multiple-data (SIMD) functional units with hardware loop to support a range of deep neural network (DNN) layer types. The 28 nm prototype chip achieves a peak throughput of 4.9 tera operations per second (TOPS) and system-level peak energy-efficiency of 437 TOPS per watt (TOPS/W) at 40 megahertz (MHz) with a 1 volt (V) supply.

Programmable in-memory computing accelerator for low-precision deep neural network inference

In some embodiments, an in-memory-computing SRAM macro based on capacitive-coupling computing (C3) (which is referred to herein as 'C3SRAM') is provided. In some embodiments, a C3SRAM macro can support array-level fully parallel computation, multi-bit outputs, and configurable multi-bit inputs. The macro can include circuits embedded in bitcells and peripherals to perform hardware acceleration for neural networks with binarized weights and activations in some embodiments. In some embodiments, the macro utilizes analog-mixed-signal capacitive-coupling computing to evaluate the main computations of binary neural networks, binary-multiply-and-accumulate operations. Without needing to access the stored weights by individual row, the macro can assert all of its rows simultaneously and form an analog voltage at the read bitline node through capacitive voltage division, in some embodiments. With one analog-to-digital converter (ADC) per column, the macro cab realize fully parallel vector-matrix multiplication in a single cycle in accordance with some embodiments.

Circuits and methods for in-memory computing

Techniques are described for efficiently reducing the amount of total computation in convolutional neural networks (CNNs) without affecting the output result or classification accuracy. Computation redundancy in CNNs is reduced by exploiting the computing nature of the convolution and subsequent pooling (e.g., sub-sampling) operations. In some implementations, the input features may be divided into a group of precision values and the operation(s) may be cascaded. A maximum may be identified (e.g., by 90% probability) using a small number of bits in the input features, and the full-precision convolution may then be performed on the maximum input. Accordingly, the total number of bits used to perform the convolution is reduced without affecting the output features or the final classification accuracy.

Cascaded computing for convolutional neural networks

A static random-access memory (SRAM) system includes SRAM cells configured to perform exclusive NOR operations between a stored binary weight value and a provided binary input value. In some embodiments, SRAM cells are configured to perform exclusive NOR operations between a stored binary weight value and a provided ternary input value. The SRAM cells are suitable for the efficient implementation of emerging deep neural network technologies such as binary neural networks and XNOR neural networks.

Static random-access memory for deep neural networks

A smart hardware security engine using biometric features and hardware-specific features is provided. The smart security engine can combine one or more entropy sources, including individually distinguishable biometric features, and hardware-specific features to perform secret key generation for user registration and authentication. Such hybrid signatures may be distinct from person-to-person (e.g., due to the biometric features) and from device-to-device (e.g., due to the hardware-specific features) while varying over time. Thus, embodiments described herein can be used for personal device authentication as well as secret random key generation, significantly reducing the scope of an attack.

Smart hardware security engine using biometric features and hardware-specific features

A resistive random-access memory (RRAM) system includes an RRAM cell. The RRAM cell includes a first select line and a second select line, a word line, a bit line, a first resistive memory device, a first switching device, a second resistive memory device, a second switching device, and a comparator. The first resistive memory device is coupled between a first access node and the bit line. The first switching device is coupled between the first select line and the first access node. The second resistive memory device is coupled between a second access node and the bit line. The second switching device is coupled between the second select line and the second access node. The comparator includes a first input coupled to the bit line, a second input, and an output.

Resistive random-access memory for embedded computation

Embodiments of the invention relate to a time-division multiplexed neurosynaptic module with implicit memory addressing for implementing a neural network. One embodiment comprises maintaining neuron attributes for multiple neurons and maintaining incoming firing events for different time steps. For each time step, incoming firing events for said time step are integrated in a time-division multiplexing manner. Incoming firing events are integrated based on the neuron attributes maintained. For each time step, the neuron attributes maintained are updated in parallel based on the integrated incoming firing events for said time step.

Time-division multiplexed neurosynaptic module with implicit memory addressing for implementing a neural network

An information processing system, which includes a control system and an artificial neural network, is disclosed. The artificial neural network includes a group of neurons and a group of synapses, which includes a first portion and a second portion. The control system selects one of a group of operating modes. The group of neurons processes information. The group of synapses provide connectivity to each of the group of neurons. During a first operating mode of the group of operating modes, the first portion of the group of synapses is enabled and the second portion of the group of synapses is enabled. During a second operating mode of the group of operating modes, the first portion of the group of synapses is enabled and the second portion of the group of synapses is disabled.

Static and dynamic precision adaptation for hardware learning and classification

Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for obtaining an electrocardiographic (ECG) signal of a user; obtaining a feature vector of the ECG signal of the user with neural network based feature extraction. Comparing the feature vector of the ECG signal with a stored feature vector of a registered user. Authenticating the user in response to determining that a similarity of the ECG feature vector of the ECG signal and the stored ECG feature vector of the registered user exceeds a pre-defined threshold value.

Electrocardiographic biometric authentication

A reconfigurable neural network circuit is provided. The reconfigurable neural network circuit comprises an electronic synapse array including multiple synapses interconnecting a plurality of digital electronic neurons. Each neuron comprises an integrator that integrates input spikes and generates a signal when the integrated inputs exceed a threshold. The circuit further comprises a control module for reconfiguring the synapse array. The control module comprises a global final state machine that controls timing for operation of the circuit, and a priority encoder that allows spiking neurons to sequentially access the synapse array.

Reconfigurable and customizable general-purpose circuits for neural networks

Resistive random-access memory for exclusive NOR (XNOR) neural networks

Growing community of inventors

Tempe, AZ, United States of America

Jae-sun Seo