Pith. sign in

REVIEW 4 major objections 5 minor 58 references

TaiBai: A fully programmable brain-inspired processor with topology-aware efficiency

T0 review · 4 major / 5 minor · reviewed 2026-08-05 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read TaiBai is a fully programmable many-core spiking-neural-network processor whose hierarchical topology encoding and instruction-set neuron cores let it run arbitrary network structures and heterogeneous neuron models, reporting 296–855x high

desk verdict Solid architecture paper with a genuinely useful topology-encoding scheme, but the headline energy-efficiency numbers come from an unvalidated simulator and should be read as plausibility estimates, not measured results. read the letter →

arxiv 2508.18961 v1 pith:KR3E5JMH submitted 2025-08-26 cs.AR

Qianpeng Li , Yu Song , Xin Liu , Wenna Song , Boshi Zhao , Zhichao Wang , Aoxin Chen , Tielin Zhang
show 1 more author
Liang Chen
This is my paper · ORCID
classification cs.AR
keywords brain-inspiredprocessorspikingneuralnetworkneuromorphichardwaretopologyencodingevent-drivenarchitectureon-chiplearningenergyefficiencycompilerstack
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

TaiBai is a many-core brain-inspired processor designed to show that full programmability and flexible network topology do not have to cost energy. The paper's central claim is that an event-driven design with a Turing-complete instruction set for neurons and a two-level fan-in/fan-out topology encoding can match a GPU's accuracy on ECG, speech, and brain-computer-interface tasks while consuming roughly a third of a watt—296 to 855 times less energy per frame. The underlying argument is that existing neuromorphic chips sacrifice generality (fixed neuron models, limited connectivity) to gain efficiency, and TaiBai's encoding schemes recover the efficiency without that sacrifice. If the simulated power numbers survive contact with real silicon, this points to a path for ultra-low-power, programmable spiking hardware.

What carries the argument

The load-bearing object is the two-level fan-in/fan-out topology table (a Directory Table of entries indexing an Information Table) with four specialized fan-in entry types. Type 0/1 handle sparse connections with different weight-decoding costs; Type 2 represents a fully connected layer with only four entries using incremental neuron addressing and a parallel sending mask; Type 3 encodes convolution by decoupling weight addressing into a global axon ID and local axon ID related by a polynomial (w_addr = global_axon * k^2 + local_axon), so one entry serves all channels. This table is what lets the chip handle arbitrary topologies without duplicating shared weights. The other half is the neur

What would settle it

Tape out the chip in 28nm and measure the three workloads (ECG, SHD speech, BCI) with the same GPU power protocol used in Sec. V.B.2; if the measured FPS/W ratio to the RTX 3090 is below 200x, the headline energy-efficiency claim is contradicted.

Watch

Extended reading notes

Core claim

The paper's central claim is that TaiBai achieves high energy efficiency and broad flexibility simultaneously through three co-designed mechanisms: a hierarchical two-level topology encoding (Directory Table + Information Table) that represents fully connected, sparse, convolutional, and skip connections with shared, non-replicated weights; a seven-stage 'reg-mem' neuron-core pipeline executing a Turing-complete instruction set so neuron dynamics, synaptic plasticity, and learning rules are software-defined; and a compiler stack with a behavior-level chip simulator that maps models onto 132 cortical column cores and produces the paper's power/efficiency estimates. The paper reports that on E

Load-bearing premise

The efficiency figures come from the authors' own behavior-level chip simulator paired with a 28nm synthesis power estimate, not from a fabricated chip, so the 296–855x advantages depend on that simulator predicting real silicon power and speed accurately.

Editorial extensions

If this is right

  • If the reported efficiency holds on fabricated silicon, sub-watt SNN inference for ECG, speech, and BCI at GPU-level accuracy becomes plausible, enabling wearable and edge deployments that current GPUs rule out.
  • The topology encoding means one chip can host networks with mixed connectivity (conv, FC, sparse, skip) without storing redundant weights, so larger models fit within on-chip SRAM.
  • Because neuron dynamics and learning rules are instructions, not fixed circuits, the same chip can serve as a research platform for iterating on new neuron models and plasticity rules instead of requiring a new chip per model.
  • The compiler stack with a simulator and expansion methods (fan-in PSUM expansion, fan-out decomposition, inter-chip proxy) offers a path for scaling to networks larger than the 264K neurons and up to 297M weights of a single chip.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The energy-efficiency ratios depend critically on spike sparsity (the three workloads have 1.2% to 33% firing rates); a workload with dense firing would likely shrink or eliminate the advantage, so the 296–855x range is not a general property of the chip.
  • The same two-level topology tables with incremental addressing could generalize to other sparse, weight-sharing workloads outside SNNs (e.g., graph neural networks, mixture-of-experts layers), though the paper only evaluates spiking networks.
  • A direct calibration study—comparing simulator power per core against measured FPGA dynamic power scaled to 28nm—would make the efficiency claims testable before tape-out; the paper does not report such a comparison.
  • The on-chip BCI learning uses accumulated spikes rather than full timestep-by-timestep backprop, which likely trades a small accuracy cost for large storage savings; an explicit comparison against full BPTT would quantify that trade.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 5 minor

Summary. The paper proposes TaiBai, a programmable many-core brain-inspired processor that combines an event-driven architecture, a hierarchical topology-encoding scheme, an instruction set for neuron/synapse dynamics and on-chip learning, and a compiler stack for mapping SNNs. The authors report functional FPGA demonstration, 28nm synthesis results, and a behavioral simulator used to estimate power and throughput. The central claim is that TaiBai achieves 'more than 200 times higher energy efficiency' than an NVIDIA RTX 3090 on three spiking applications (speech, ECG, BCI), with 296–855x efficiency in those cases, while three standard SNN benchmarks show only 6–20x efficiency gains.

Significance. If the efficiency numbers and the programmability claims hold, TaiBai would be a meaningful advance over existing neuromorphic processors, particularly because it targets heterogeneous neuron models, flexible topologies, and on-chip learning. The hierarchical topology representation (decomposed convolution, parallel sending, incremental addressing) is a credible engineering contribution, and the combination of an FPGA prototype with a compiler stack is a useful demonstration. However, the load-bearing quantitative claims rest on a behavioral simulator that is not validated against RTL, gate-level, or measured silicon power, and the abstract selects the most favorable efficiency range while the benchmark section reports a much smaller range for standard SNNs. The paper's strengths are the architectural ideas and the breadth of the prototype; its weakness is the unvalidated basis of the headline efficiency claim.

major comments (4)
  1. [Sec. V.B.2, Fig. 13(d), Fig. 15(c)] The central efficiency claim (abstract: 'more than 200 times higher energy efficiency'; Sec. V.C.4: 296–855x in Fig. 15(c)) is derived from the Python behavioral-level chip simulator, as stated in Sec. V.B.2: 'We use the chip simulator to obtain the running power consumption and running time of TaiBai.' GPU power is measured via pynvml, creating an asymmetric comparison. No calibration of the simulator’s power/timing model against RTL, gate-level synthesis, or FPGA measurements is reported, and the simulator is also used inside the compiler loop, making it both optimizer and evaluator. Since memory power (70.3%) and spike-sparsity savings are exactly the components most likely to be optimistic, the 296–855x ratios are not yet supported. The authors should either provide gate-level power validation (e.g., annotated RTL or post-synthesis simulation) or explicitly downgrade the claims to 's
  2. [Abstract and Sec. V.C.1 vs V.C.4] The abstract's 'more than 200 times' is based solely on the three custom, application-tuned networks in Fig. 15(c), whereas Sec. V.C.1 and Fig. 13(d) report only 6–20x power efficiency for three standard SNN benchmarks (e.g., 'PLF-NET and ResNet19... power efficiency improved by 6 to 20 times'). This selective reporting obscures the workload dependence. The manuscript should present both ranges in the abstract and emphasize that the large efficiency ratios are specific to small, highly sparse, application-specific models, not general SNN workloads.
  3. [Sec. IV.B and Sec. III.B] The paper repeatedly calls TaiBai 'fully programmable' and states that the instruction set is 'Turing-complete,' but no formal proof or compelling demonstration is given. The 'fully programmable' definition in Sec. V.C.2 is informal ('ability to implement any models'), and the concrete examples cover only a few neuron models and learning rules. In addition, Sec. IV.B acknowledges a hardware limit of 2K fan-ins per neuron, which requires special expansion for models like the dendritic DH-LIF (Sec. V.B.3). The programmability claim should be either formally supported (e.g., by showing that arbitrary neuron dynamics and learning rules can be encoded in the instruction set with bounded overhead) or qualified to 'programmable over the demonstrated model classes.'
  4. [Sec. V.A] The paper states that 'TaiBai is synthesized based on SMIC 28nm... and operates at 500MHz' and uses an FPGA prototype only for functional demonstration. It is not clear whether the reported 1.83W and 528 GSOPS are from synthesis or from the simulator. If the power and energy-efficiency numbers are pre-silicon estimates, the manuscript should clearly state this in every relevant table and figure caption (e.g., Table IV, Fig. 13, Fig. 15), and should not present them as measured chip results. Adding a disclaimer in the abstract or introduction would also prevent misinterpretation.
minor comments (5)
  1. [Sec. III.D] There are two subsections numbered '2)' (Sparse connection and Topology on hardware) and two numbered '4)' (Convolutional connection and Applications for TaiBai in Sec. V.C). Re-number for clarity.
  2. [Sec. IV.B] Typo: 'on-chip leanring' should be 'on-chip learning.' Also 'inculdes' in Fig. 15 caption and 'borad' in Sec. V.A.
  3. [Abstract] The phrase 'demonstrated its high potentiation' is unusual; likely 'potential' or 'potential for' is intended.
  4. [Sec. V.B.3] The description of the BCI model is dense; a small block diagram or explicit equations for the fusion (Hadamard product + matrix addition) would help the reader assess the claimed complexity and the 3.4x core reduction.
  5. [Fig. 13(c)] The power breakdown in Fig. 13(c) shows memory at 70.3%, but the text does not state whether this is from synthesis or the simulator. Clarify the source to help the reader judge the plausibility of the 1.83W figure.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the efficiency claim is simulator-based, not derived from its own inputs.

full rationale

I walked the claimed derivation chain for each headline result. (1) The 296-855x energy-efficiency numbers are not obtained by an equation that already contains the result; they are outputs of the authors' behavioral-level chip simulator. The paper states "We use the chip simulator to obtain the running power consumption and running time of TaiBai" and separately measures GPU power via the pynvml API. A simulator used for both mapping optimization and evaluation is an external-validity risk, but it is not a circular reduction: the simulator is not fitted to the 296-855x figure, and no equation in the paper defines efficiency as equal to an input assumption. (2) The topology-representation storage savings (286-947x in Fig. 14) are computed by directly comparing the proposed table entries against explicit fully-connected unfolded baselines; this is a concrete counting comparison, not a self-justifying definition. (3) The programmability claim rests on the described instruction set, the seven-stage datapath, and the FPGA functional demonstration, not on a self-citation or imported uniqueness theorem. (4) The only apparent self-citation, the DHSNN model from [15] (Zheng et al.) used in the speech-recognition application, is an input workload for demonstrating flexibility, not the architectural justification; even if it is the authors' prior work, it is not load-bearing for the chip's claimed capabilities. I also note a citation mismatch in the text ("the DHSNN with the dendrite model proposed by Deng [15]" while [15] is Zheng et al.) and that the abstract selects the 296-855x application-tuned range rather than the 6-20x range of Fig. 13(d); both are accuracy-of-reporting concerns, not circularity. No step could be exhibited in which a prediction equals its input by construction.

Assumptions & free parameters 1 free parameters · 4 assumptions · 0 invented entities

TaiBai introduces no new physical entities. The central efficiency claim rests on the simulator's power model and the fairness of the GPU baseline; the programmability claim rests on an unproven Turing-completeness assertion. The only listed free parameter is the fan-in hardware limit, a design choice that shapes application deployment.

free parameters (1)
  • Per-neuron fan-in limit of 2,048
    Hardware design constraint chosen in Sec. IV.B; forces fan-in expansion for the DH-LIF neuron model with 2,800 fan-ins, affecting core count and latency in the speech task.
assumptions (4)
  • domain assumption The behavioral simulator (plus 28nm synthesis power estimates) accurately predicts the power, latency, and resource usage of a fabricated TaiBai chip.
    All TaiBai efficiency numbers in Sec. V are simulator-based; no fabricated chip exists. The FPGA prototype is used for functional demonstration, not power validation.
  • domain assumption The TaiBai instruction set is Turing-complete, so any neuron/synapse model and any on-chip learning rule can be expressed.
    Asserted in Sec. III.B ('Turing-complete instruction set') without formal proof; only LIF, ALIF, DH-LIF, and a custom BCI model are demonstrated.
  • standard math Standard LIF neuron dynamics and STBP surrogate-gradient training are correctly implemented.
    Background stated in Sec. II.A; these are common modeling assumptions in SNN literature.
  • domain assumption The GPU power measurement via pynvml (average over 15 runs) is a fair baseline for the energy efficiency comparison.
    GPU power is measured on real hardware while TaiBai power is simulated; the asymmetry may inflate the reported efficiency ratio.

how reviews work

0 comments
Cite this review

Pith. "Pith review of TaiBai: A fully programmable brain-inspired processor with topology-aware efficiency." pith.science (2026). https://pith.science/paper/KR3E5JMH

@misc{pith2026250818961,
  author       = {Pith},
  title        = {Pith review of: TaiBai: A fully programmable brain-inspired processor with topology-aware efficiency},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/KR3E5JMH}},
  note         = {Machine review of arXiv:2508.18961}
}
read the original abstract

Brain-inspired computing has emerged as a promising paradigm to overcome the energy-efficiency limitations of conventional intelligent systems by emulating the brain's partitioned architecture and event-driven sparse computation. However, existing brain-inspired chips often suffer from rigid network topology constraints and limited neuronal programmability, hindering their adaptability. To address these challenges, we present TaiBai, an event-driven, programmable many-core brain-inspired processor that leverages temporal and spatial spike sparsity to minimize bandwidth and computational overhead. TaiBai chip contains three key features: First, a brain-inspired hierarchical topology encoding scheme is designed to flexibly support arbitrary network architectures while slashing storage overhead for large-scale networks; Second, a multi-granularity instruction set enables programmability of brain-like spiking neuron or synapses with various dynamics and on-chip learning rules; Third, a co-designed compiler stack optimizes task mapping and resource allocation. After evaluating across various tasks, such as speech recognition, ECG classification, and cross-day brain-computer interface decoding, we found spiking neural networks embedded on the TaiBai chip could achieve more than 200 times higher energy efficiency than a standard NVIDIA RTX 3090 GPU at a comparable accuracy. These results demonstrated its high potentiation as a scalable, programmable, and ultra-efficient solution for both multi-scale brain simulation and brain-inspired computation.

Figures

Figures reproduced from arXiv: 2508.18961 by the authors.

Figure 1
Figure 1. The chip is inspired by (a) the key characteristics of brain structure to enhance (b) multiple programmable capabilit [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Overview of TaiBai. (a) The TaiBai chip, (b) cortical column core and (c) neuron core. [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. The architecture of neuron core. neuron cores (NCs) for model computation. Inspired by biological neurons the NC based on the brain-inspired instruction set can flexibly implement various models ( [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (12 more)
Figure 4
Figure 4. Figure 4: The block diagram of scheduler. 1) Scheduler: The scheduler in CC is primarily used to schedule spike events and NCs, thereby achieving efficient parallel computation under sparse event conditions. When the scheduler receives a spike event packet, it determines the cor…
Figure 5
Figure 5. Figure 5: Sparsely connected representation scheme, and the [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]
Figure 6
Figure 6. Figure 6: Topology representation scheme with parallel sending [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]
Figure 7
Figure 7. Figure 7: Topology representation scheme with parallel sending [PITH_FULL_IMAGE:figures/full_fig_p007_7.png]
Figure 8
Figure 8. Figure 8: The representation scheme for skip connection. [PITH_FULL_IMAGE:figures/full_fig_p007_8.png]
Figure 9
Figure 9. Figure 9: Data structure and model representation in TaiBai. The (a) structure and (b) calculation code of LIF neuron model wit [PITH_FULL_IMAGE:figures/full_fig_p008_9.png]
Figure 10
Figure 10. Figure 10: Workflow of the chip. B. Model Extensibility We build data structure and model representation in TaiBai for neuron models and learning algorithms extensibility [PITH_FULL_IMAGE:figures/full_fig_p008_10.png]
Figure 11
Figure 11. Figure 11: Fan-in capability expansion. In fan-out expansion, the original neuron needs to be decomposed into multiple neurons with local destination areas. These neurons have the same firing time, and the sum of all destination areas is the original destination area, which can …
Figure 12
Figure 12. Figure 12: Compiler stack. The process of deploying (a) a workload built on multiple frameworks with multiple network topologie [PITH_FULL_IMAGE:figures/full_fig_p010_12.png]
Figure 13
Figure 13. Figure 13: Chip evaluation and performance. (a) Chip layout. (b) Layout of CC and router. (c) Power breakdown of TaiBai. (d) [PITH_FULL_IMAGE:figures/full_fig_p012_13.png]
Figure 14
Figure 14. Figure 14: Efficiency of network topology representation on [PITH_FULL_IMAGE:figures/full_fig_p013_14.png]
Figure 15
Figure 15. Figure 15: Applications performance of (a) accuracy, (b) power, [PITH_FULL_IMAGE:figures/full_fig_p014_15.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

58 extracted references · 57 canonical work pages

  1. [1]

    Scheduler: The scheduler in CC is primarily used to schedule spike events and NCs, thereby achieving efficient parallel computation under sparse event conditions. When the scheduler receives a spike event packet, it determines the corresponding axon ID and target neuron ID through the fan- in topology table, and initiates the corresponding NC for computat...

    work page

  2. [2]

    Both tables (Fig

    Topology on hardware : The network topology is divided into fan-in and fan- out from the perspective of the connection between neuron inputs and outputs, and they are stored in the fan-in topology table and fan- out topology table, respectively. Both tables (Fig. 4) are organized as 2 -level tables with sparse connectivity representation and indexing capa...

    work page

  3. [3]

    Type 0 : IE is the IDs of the neurons that are connected to the upstream neuron, and NC decodes the sparse weight address through the global axon ID (Fig

    Sparse connection: For sparse connection, two types of fan-in IE are provided to meet different requirements for sparse weight decoding speed and topology representation resource overhead. Type 0 : IE is the IDs of the neurons that are connected to the upstream neuron, and NC decodes the sparse weight address through the global axon ID (Fig. 5 (a)). This ...

    work page

  4. [4]

    In this connection pattern, a spike event will cause the computation of all destination neurons

    Full connection : For full connection, each upstream neuron has the same destination neuron, and the weight address of the destination neuron is only related to the upstream neuron ID. In this connection pattern, a spike event will cause the computation of all destination neurons. We proposed a type 2 IE, on the one hand, the mechanism of incremental addr...

    work page

  5. [5]

    However, decoding the shared filter weights corresponding to the events requires a complex algorithm with high computational overhead [41]

    Convolutional connection: For convolutional connection, upstream neurons at the same position in different channels have the same destination neurons, and the destination neurons of the same channel share a set of filters. However, decoding the shared filter weights corresponding to the events requires a complex algorithm with high computational overhead ...

    work page

  6. [6]

    The traditional method adds a set of relay neurons to cache spikes for synchronization, but this will bring a very large resource consumption (Fig

    Skip connections: Skip connections disrupt the layer -wise timestep synchronization of the brain -inspired chip pipeline, resulting in network layers needing to receive multiple sets of data processed at different timesteps. The traditional method adds a set of relay neurons to cache spikes for synchronization, but this will bring a very large resource co...

    work page

  7. [7]

    Implementation of the compiler stack : We developed an assembler for the TaiBai instruction set based on flex and bison, and developed the mapper using Python for optimizing operator intermediate representation, network partition, core placement optimization, and resource optimization. Genetic algorithms or simulated annealing algorithms are used to optim...

    work page

  8. [8]

    For GPU power consumption, we record the static power during an extended period with no workload

    Benchmarks: We use the chip simulator to obtain the running power consumption and running time of TaiBai. For GPU power consumption, we record the static power during an extended period with no workload. Then we record the power while the model is running and calculate the average power. All power consumption is obtained by the pynvml API provided by NVID...

    work page

Show all 58 references
  1. [9]

    The models of these applications cover a variety of neuron models, network topologies and on -chip learning algorithms

    Algorithms and datasets for multiple applications : We demonstrated three complex tasks on the chip prototype system, namely ECG signal recognition, speech recognition and BCI decoding. The models of these applications cover a variety of neuron models, network topologies and o...

  2. [10]

    The chip can support a total of 264K neurons and 6.95M (sparse mode) ~ 297M (convolutional multiplexing mode) synaptic weights

    Performance of the TaiBai chip: Table III lists the basic characteristics and parameters. The chip can support a total of 264K neurons and 6.95M (sparse mode) ~ 297M (convolutional multiplexing mode) synaptic weights. The 1 Fig. 13. Chip evaluation and performance. (a) Chip la...

  3. [11]

    third - factor

    Comparison with the state -of-the-art brain-inspired chips: As shown in Table IV, we divide brain-inspired chips into two categories based on programmability: those on the left support specific models, and those on the right have different degrees of programmability. Programma...

  4. [12]

    Incremental addressing of neurons in the fully connected layer: All neurons are addressed by only 4 entries, regardless of the number of neurons in the fully connected layer

    Efficiency of the network topology representation : The network topology representation scheme we proposed has three main contributions. Incremental addressing of neurons in the fully connected layer: All neurons are addressed by only 4 entries, regardless of the number of neu...

  5. [13]

    These applications feature different input data types, network topologies, neuron dynamic, and synapse dynamic models

    Applications for TaiBai : To demonstrate the multi -scale flexible programmability and on -chip learning capabilities of TaiBai, we selected three typical SNN applications. These applications feature different input data types, network topologies, neuron dynamic, and synapse d...

  6. [14]

    Introducing the human brain project,

    H. Markram, K. Meier, T. Lippert, et al., “Introducing the human brain project,” Procedia Computer Science, vol. 7, pp. 39 --42, 2011, proceedings of the 2nd European Future Technologies Conference and Exhibition 2011 (FET 11). [Online]. Available: https://www.sciencedirect.co...

    2011

  7. [15]

    Brain-inspired computing needs a master plan,

    A. Mehonic and A. J. Kenyon, “Brain-inspired computing needs a master plan,” Nature, vol. 604, no. 7905, pp. 255--260, 2022. 1

    2022

  8. [16]

    Neuro-inspired computing chips,

    W. Zhang, B. Gao, J. Tang, et al. , “Neuro-inspired computing chips, ” Nature electronics, vol. 3, no. 7, pp. 371--382, 2020

    2020

  9. [17]

    Truenorth: Design and tool flow of a 65 mw 1 million neuron programmable neurosynaptic chip,

    F. Akopyan, J. Sawada, A. Cassidy, et al., “Truenorth: Design and tool flow of a 65 mw 1 million neuron programmable neurosynaptic chip, ” IEEE transactions on computer -aided design of integrated circuits and systems, vol. 34, no. 10, pp. 1537--1557, 2015

    2015

  10. [18]

    Spinnaker: A 1 -w 18 -core system-on-chip for massively -parallel neural network simulation,

    E. Painkras, L. A. Plana, J. Garside, et al. , “Spinnaker: A 1 -w 18 -core system-on-chip for massively -parallel neural network simulation, ” IEEE Journal of Solid-State Circuits, vol. 48, no. 8, pp. 1943--1953, 2013

    1943

  11. [19]

    Loihi: A neuromorphic manycore processor with on-chip learning,

    M. Davies, N. Srinivasa, T. -H. Lin, et al. , “Loihi: A neuromorphic manycore processor with on-chip learning,” Ieee Micro, vol. 38, no. 1, pp. 82--99, 2018

    2018

  12. [20]

    Towards artificial general intelligence with hybrid tianjic chip architecture,

    J. Pei, L. Deng, S. Song, et al., “Towards artificial general intelligence with hybrid tianjic chip architecture,” Nature, vol. 572, no. 7767, pp. 106- -111, 2019

    2019

  13. [21]

    Influence of dendritic structure on firing pattern in model neocortical neurons,

    Z. F. Mainen and T. J. Sejnowski, “Influence of dendritic structure on firing pattern in model neocortical neurons, ” Nature, vol. 382, no. 6589, pp. 363--366, 1996

    1996

  14. [22]

    Canonical microcircuits for predictive coding,

    A. M. Bastos, W. M. Usrey, R. A. Adams, et al., “Canonical microcircuits for predictive coding,” Neuron, vol. 76, no. 4, pp. 695--711, 2012

    2012

  15. [23]

    The neocortical circuit: themes and variations,

    K. D. Harris and G. M. Shepherd, “The neocortical circuit: themes and variations,” Nature neuroscience, vol. 18, no. 2, pp. 170--181, 2015

    2015

  16. [24]

    Attractor and integrator networks in the brain,

    M. Khona and I. R. Fiete, “Attractor and integrator networks in the brain,” Nature Reviews Neuroscience, vol.~23, no.~12, pp. 744--766, 2022

    2022

  17. [25]

    Cellular and molecular features of axon collaterals and dendrites,

    A. Acebes and A. Ferr ú s, “Cellular and molecular features of axon collaterals and dendrites, ” Trends in neurosciences , vol. 23, no. 11, pp. 557--565, 2000

    2000

  18. [26]

    The topographic brain: from neural connectivity to cognition,

    J.-P. Thivierge and G. F. Marcus, “The topographic brain: from neural connectivity to cognition, ” Trends in neurosciences , vol. 30, no. 6, pp. 251--259, 2007

    2007

  19. [27]

    Theoretical neuroanatomy and the connectivity of the cerebral cortex,

    O. Sporns, G. Tononi, and G. Edelman, “Theoretical neuroanatomy and the connectivity of the cerebral cortex, ” Behavioural brain research, vol. 135, no. 1-2, pp. 69--74, 2002

    2002

  20. [28]

    Temporal dendritic heterogeneity incorporated with spiking neural networks for learning multi -timescale dynamics,

    H. Zheng, Z. Zheng, R. Hu, et al. , “Temporal dendritic heterogeneity incorporated with spiking neural networks for learning multi -timescale dynamics,” Nature Communications, vol. 15, no. 1, p. 277, 2024

    2024

  21. [29]

    A reservoir of timescales emerges in recurrent circuits with heterogeneous neural assemblies,

    M. Stern, N. Istrate, and L. Mazzucato, “A reservoir of timescales emerges in recurrent circuits with heterogeneous neural assemblies,” Elife, vol. 12, p. e86552, 2023

    2023

  22. [30]

    Adapting to time: why nature evolved a diverse set of neurons,

    K. G. Habashy, B. D. Evans, D. F. Goodman, and J. S. Bowers, “Adapting to time: why nature evolved a diverse set of neurons,” arXiv e- prints, pp. arXiv --2404, 2024. [Online]. Available: https://doi.org/10.48550/arXiv.2404.14325

    2024 doi arXiv

  23. [31]

    Neural heterogeneity promotes robust learning,

    N. Perez-Nieves, V. C. Leung, P. L. Dragotti, and D. F. Goodman, “Neural heterogeneity promotes robust learning,” Nature communications, vol. 12, no. 1, p. 5791, 2021

    2021

  24. [32]

    Accurate and efficient time-domain classification with adaptive spiking recurrent neural networks,

    B. Yin, F. Corradi, and S. M. Bohté, “Accurate and efficient time-domain classification with adaptive spiking recurrent neural networks, ” Nature Machine Intelligence, vol. 3, no. 10, pp. 905--913, 2021

    2021

  25. [33]

    Competitive hebbian learning through spike-timing-dependent synaptic plasticity,

    S. Song, K. D. Miller, and L. F. Abbott, “Competitive hebbian learning through spike-timing-dependent synaptic plasticity,” Nature neuroscience, vol. 3, no. 9, pp. 919--926, 2000

    2000

  26. [34]

    Spatio-temporal backpropagation for training high -performance spiking neural networks,

    Y. Wu, L. Deng, G. Li, J. Zhu, and L. Shi, “Spatio-temporal backpropagation for training high -performance spiking neural networks,” Frontiers in neuroscience, vol. 12, p. 331, 2018

    2018

  27. [35]

    Efficient neuromorphic signal processing with loihi 2,

    G. Orchard, E. P. Frady, D. B. D. Rubin, et al., “Efficient neuromorphic signal processing with loihi 2, ” in 2021 IEEE Workshop on Signal Processing Systems (SiPS). IEEE, 2021, pp. 254--259

    2021

  28. [36]

    Neurogrid: A mixed-analog- digital multichip system for large -scale neural simulations,

    B. V. Benjamin, P. Gao, E. McQuinn, et al., “Neurogrid: A mixed-analog- digital multichip system for large -scale neural simulations,” Proceedings of the IEEE, vol. 102, no. 5, pp. 699--716, 2014

    2014

  29. [37]

    Flexon: A flexible digital neuron for efficient spiking neural network simulations,

    D. Lee, G. Lee, D. Kwon, et al. , “Flexon: A flexible digital neuron for efficient spiking neural network simulations, ” in 2018 ACM/IEEE 45th Annual International Symposium on Computer Architecture (ISCA). IEEE, 2018, pp. 275--288

    2018

  30. [38]

    Flexlearn: fast and highly efficient brain simulations using flexible on -chip learning,

    E. Baek, H. Lee, Y. Kim, and J. Kim, “Flexlearn: fast and highly efficient brain simulations using flexible on -chip learning,” in Proceedings of the 52nd Annual IEEE/ACM International Symposium on Microarchitecture , 2019, pp. 304--318

    2019

  31. [39]

    Paicore: a 1.9-million-neuron 5.181- tsops/w digital neuromorphic processor with unified snn -ann and on-chip learning paradigm,

    Y. Zhong, Y. Kuang, K. Liu, et al., “Paicore: a 1.9-million-neuron 5.181- tsops/w digital neuromorphic processor with unified snn -ann and on-chip learning paradigm,” IEEE Journal of Solid-State Circuits, 2024

    2024

  32. [40]

    Darwin3: a large- scale neuromorphic chip with a novel isa and on -chip learning,

    D. Ma, X. Jin, S. Sun, et al., “Darwin3: a large- scale neuromorphic chip with a novel isa and on -chip learning,” National Science Review, vol. 11, no. 5, p. nwae102, 2024

    2024

  33. [41]

    Brain-inspired computing: A systematic survey and future trends,

    G. Li, L. Deng, H. Tang, et al., “Brain-inspired computing: A systematic survey and future trends,” Proceedings of the IEEE, 2024

    2024

  34. [42]

    Spiking neural network integrated circuits: A review of trends and future directions,

    A. Basu, L. Deng, C. Frenkel, and X. Zhang, “Spiking neural network integrated circuits: A review of trends and future directions, ” in 2022 IEEE Custom Integrated Circuits Conference (CICC). IEEE, 2022, pp. 1-- 8

    2022

  35. [43]

    Network group partition and core placement optimization for neuromorphic multi -core and multi -chip systems,

    Y. Yang, Q. Fan, T. Yan, J. Pei, and G. Li, “Network group partition and core placement optimization for neuromorphic multi -core and multi -chip systems,” IEEE Transactions on Emerging Topics in Computational Intelligence, 2024

    2024

  36. [44]

    Policy gradient-based core placement optimization for multichip many -core systems,

    W. Myung, D. Lee, C. Song, G. Wang, and C. Ma, “Policy gradient-based core placement optimization for multichip many -core systems, ” IEEE Transactions on Neural Networks and Learning Systems , vol. 34, no. 8, pp. 4529--4543, 2021

    2021

  37. [45]

    A hybrid-mode on-chip router for the large-scale fpga- based neuromorphic platform,

    C. Ding, Y. Huan, H. Jia, et al., “A hybrid-mode on-chip router for the large-scale fpga- based neuromorphic platform, ” IEEE Transactions on Circuits and Systems I: Regular Papers , vol. 69, no. 5, pp. 1990 --2001, 2022

    1990

  38. [46]

    Braindrop: A mixed -signal neuromorphic architecture with a dynamical systems -based programming model,

    A. Neckar, S. Fok, B. V. Benjamin, et al. , “Braindrop: A mixed -signal neuromorphic architecture with a dynamical systems -based programming model,” Proceedings of the IEEE, vol. 107, no. 1, pp. 144--164, 2018

    2018

  39. [47]

    A reconfigurable on-line learning spiking neuromorphic processor comprising 256 neurons and 128k synapses,

    N. Qiao, H. Mostafa, F. Corradi, et al., “A reconfigurable on-line learning spiking neuromorphic processor comprising 256 neurons and 128k synapses,” Frontiers in neuroscience, vol. 9, p. 141, 2015

    2015

  40. [48]

    Edge learning using a fully integrated neuro-inspired memristor chip,

    W. Zhang, P. Yao, B. Gao, et al., “Edge learning using a fully integrated neuro-inspired memristor chip, ” Science, vol. 381, no. 6663, pp. 1205 -- 1211, 2023

    2023

  41. [49]

    Hybrid 2d --cmos microchips for memristive applications,

    K. Zhu, S. Pazos, F. Aguirre, et al., “Hybrid 2d --cmos microchips for memristive applications,” Nature, vol. 618, no. 7963, pp. 57--62, 2023

    2023

  42. [50]

    Two sparsities are better than one: unlocking the performance benefits of sparse --sparse networks,

    K. Hunter, L. Spracklen, and S. Ahmad, “Two sparsities are better than one: unlocking the performance benefits of sparse --sparse networks, ” Neuromorphic Computing and Engineering, vol. 2, no. 3, p. 034004, 2022

    2022

  43. [51]

    Cerebron: A reconfigurable architecture for spatiotemporal sparse spiking neural networks,

    Q. Chen, C. Gao, and Y. Fu, “Cerebron: A reconfigurable architecture for spatiotemporal sparse spiking neural networks, ” IEEE Transactions on Very Large Scale Integration (VLSI) Systems , vol. 30, no. 10, pp. 1425 -- 1437, 2022

    2022

  44. [52]

    A set of composite, non - redundant eeg measures of nrem sleep based on the power law scaling of the fourier spectrum,

    R. Bódizs, O. Szalárdy, C. Horváth, et al. , “A set of composite, non - redundant eeg measures of nrem sleep based on the power law scaling of the fourier spectrum,” Scientific reports, vol. 11, no. 1, p. 2041, 2021

  45. [53]

    Mosaic: in-memory computing and routing for small -world spike-based neuromorphic systems,

    T. Dalgaty, F. Moro, Y. Demiraǔ, et al., “Mosaic: in-memory computing and routing for small -world spike-based neuromorphic systems, ” Nature Communications, vol. 15, no. 1, p. 142, 2024

    2024

  46. [54]

    Efficient hardware acceleration of sparsely active convolutional spiking neural networks,

    J. Sommer, M. A. Özkan, O. Keszocze, and J. Teich, “Efficient hardware acceleration of sparsely active convolutional spiking neural networks, ” IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems, vol. 41, no. 11, pp. 3767--3778, 2022

    2022

  47. [55]

    Deep residual learning for image recognition,

    K. He, X. Zhang, S. Ren, and J. Sun, “Deep residual learning for image recognition,” in Proceedings of the IEEE conference on computer vision and pattern recognition, 2016, pp. 770--778

    2016

  48. [56]

    Very deep convolutional networks for large-scale image recognition,

    K. Simonyan and A. Zisserman, “Very deep convolutional networks for large-scale image recognition,” arXiv preprint, 2014. [Online]. Available: https://doi.org/10.48550/arXiv.1409.1556

    2014 doi arXiv

  49. [57]

    A database for evaluation of algorithms for measurement of qt and other waveform intervals in the ecg,

    P. Laguna, R. G. Mark, A. Goldberg, and G. B. Moody, “A database for evaluation of algorithms for measurement of qt and other waveform intervals in the ecg, ” in Computers in cardiology 1997 . IEEE, 1997, pp. 673--676

    1997

  50. [58]

    The heidelberg spiking data sets for the systematic evaluation of spiking neural networks,

    B. Cramer, Y. Stradmann, J. Schemmel, and F. Zenke, “The heidelberg spiking data sets for the systematic evaluation of spiking neural networks,” IEEE Transactions on Neural Networks and Learning Systems, vol. 33, no. 7, pp. 2744--2757, 2020

    2020

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.