Towards artificial general intelligence with hybrid Tianjic chip architecture

By Jing Pei, Lei Deng, Sen Song, Mingguo Zhao, Youhui Zhang, Shuang Wu, Guanrui Wang, Zhe Zou, Zhenzhi Wu, Wei He and 14 more. Nature, 2019

1,085 citations recorded by the source corpus.

Subfields: Cellular and Molecular Neuroscience, Cognitive Neuroscience, Electrical and Electronic Engineering

View Paper: https://www.semanticscholar.org/paper/56559fcd8b588357051e0d3c46631bdb435b963b

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Cellular and Molecular Neuroscience

1. J. Joshua Yang
2. Yonghong Tian
3. Qiangfei Xia
4. Zhongrui Wang
5. Wei Dang Lu
6. Hao Jiang
7. Lei Deng
8. Guoqi Li

In Cognitive Neuroscience

1. Yonghong Tian
2. Guoqi Li
3. Liuliang Yuan
4. Man Yao
5. Lei Deng
6. Yuan Xie
7. Wei Dang Lu
8. Mohammed Bennamoun

In Electrical and Electronic Engineering

1. Yonghong Tian
2. Liuliang Yuan
3. Man Yao
4. Guoqi Li
5. J. Joshua Yang
6. Qiangfei Xia
7. Lei Deng
8. Zhongrui Wang

Citations in the corpus

• Opportunities for neuromorphic computing algorithms and applications (2022, background): “One of the neuromorphic platforms targeting more general computations for wider classes of applications is the Tianjic chip, a platform that supports both neuromorphic spiking neural networks and the traditional artificial neural networks for different categories of problems 22 .”
• Model Compression and Hardware Acceleration for Neural Networks: A Comprehensive Survey (2020, background): “The complete architecture is similar to the spatial architectures in [264]–[267] (see Fig.”
• Spike-driven Transformer (2023, methodology): “When implementing SNNs on neuromorphic chips, such as TrueNorth [2], Loihi [3], and Tianjic [4], only a small fraction of spiking neurons at any moment being active and the rest being idle.”
• Training Spiking Neural Networks Using Lessons From Deep Learning (2021, methodology): “From the engineer’s perspective, neuron models derived from experimental results have allowed us to design extremely energy-efficient networks when running on hardware tailored to SNNs [219], [220], [221], [222], [223], [224], [225].”
• Efficient Acceleration of Deep Learning Inference on Resource-Constrained Edge Devices: A Review (2023, background): “Tianjic [324] and Simba [325] are two spatial architectures that utilized the concept of distributed weights [326] and weight exchange among PEs [319].”
• Hebbian Learning based Orthogonal Projection for Continual Learning of Spiking Neural Networks (2024, methodology): “Considering neuromorphic computing systems, this may be supported by some hybrid hardware (Pei et al., 2019).”
• Neuromorphic computing at scale (2025, neutral): “Although there are several examples of this in neuromorphic systems 3,32,39,49 , such as skip-zero and zero-weight approaches, there is an opportunity to scale these mechanisms.”
• Hardware implementation of memristor-based artificial neural networks (2024, neutral): “Also, the platforms developed by IBM (TrueNorth 33 ), and Tsingshua 34 are well known examples of the research effort of both the industry and the academia in this fi eld.”
• Survey of Machine Learning Accelerators (2020, background): “Another recent notable research accelerator is the Tianjic research chip [19], developed by a team at Tsinghua University, which demonstrated the capability of choosing either a spiking neural network (SNN) layer or non-spiking artificial neural network (ANN) layer to instantiate each layer of a DNN…”
• A compute-in-memory chip based on resistive random-access memory (2022, neutral): “Extended Data Table 2 shows that such scaling will enable NeuRRAM to deliver higher energy and area efficiency than today’s state-of-the-art edge inference accelerators 58–61 .”
• Three-dimensional memristor circuits as complex neural networks (2020, background): “In the future, the fully connected layers could also be implemented in memristor arrays to form a highly efficient hybrid 3D/2D system for activation-driven analogue computing (speed and energy estimation are shown in Supplementary Note 5 and Supplementary Tables 1 and 2, and a preliminary demonstra”
• Attention Spiking Neural Networks (2022, background): “Loihi [10], and Tianjic [11], we are not very far from neuromorphic processors becoming a part of everyday life.”