pith. sign in

arxiv: 1907.10508 · v1 · pith:4VVEYPYFnew · submitted 2019-07-22 · 💻 cs.CY · cs.AI

A system of different layers of abstraction for artificial intelligence

Alexander Serb , Themistoklis Prodromakis This is my paper

Pith reviewed 2026-05-24 18:13 UTC · model grok-4.3

classification 💻 cs.CY cs.AI
keywords AI abstraction layerscomplexity-performance trade-offcontrol-complexity trade-offAI system structureconceptual map for AIAI implementations
0
0 comments X

The pith

AI implementations rest on five layers of abstraction linked by complexity-performance and control-complexity trade-offs.

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

The paper proposes a system of five layers of abstraction that underpin many AI implementations. Each layer is subject to a complexity-performance trade-off. Different layers are interlocked with one another through a control-complexity trade-off. This structure supplies a conceptual map for targeting innovation to achieve different functionality levels, assure safety, optimize under constraints, and identify social and economic opportunities.

Core claim

We propose a system of five levels/layers of abstraction that underpin many AI implementations. We further posit that each layer is subject to a complexity-performance trade-off whilst different layers are interlocked with one another in a control-complexity trade-off. This overview provides a conceptual map that can help to identify how and where innovation should be targeted in order to achieve different levels of functionality, assure them for safety, optimise performance under various operating constraints and map the opportunity space for social and economic exploitation.

What carries the argument

The five-layer abstraction system, with intra-layer complexity-performance trade-offs and inter-layer control-complexity trade-offs.

If this is right

  • Innovation can be targeted at particular layers to reach different levels of functionality.
  • Safety measures can be assigned to the appropriate abstraction layer.
  • Performance can be optimized under operating constraints by adjusting positions within the trade-offs.
  • Social and economic exploitation opportunities become identifiable through the layer map.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same layering might be tested on non-AI complex systems such as biological neural networks to check for similar trade-offs.
  • Quantitative metrics for complexity, performance, and control could be defined per layer and measured in deployed AI to validate the model.
  • The framework suggests that advances confined to one layer may be limited by the control trade-off with adjacent layers.

Load-bearing premise

The proposed five-layer decomposition and the two posited trade-offs accurately and generally describe the structure of AI implementations.

What would settle it

Examining a broad sample of existing AI systems and finding many that cannot be mapped onto the five layers or that fail to show the stated trade-offs would falsify the proposal.

read the original abstract

The field of artificial intelligence (AI) represents an enormous endeavour of humankind that is currently transforming our societies down to their very foundations. Its task, building truly intelligent systems, is underpinned by a vast array of subfields ranging from the development of new electronic components to mathematical formulations of highly abstract and complex reasoning. This breadth of subfields renders it often difficult to understand how they all fit together into a bigger picture and hides the multi-faceted, multi-layered conceptual structure that in a sense can be said to be what AI truly is. In this perspective we propose a system of five levels/layers of abstraction that underpin many AI implementations. We further posit that each layer is subject to a complexity-performance trade-off whilst different layers are interlocked with one another in a control-complexity trade-off. This overview provides a conceptual map that can help to identify how and where innovation should be targeted in order to achieve different levels of functionality, assure them for safety, optimise performance under various operating constraints and map the opportunity space for social and economic exploitation.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

0 major / 1 minor

Summary. The manuscript is a perspective piece that proposes a five-layer system of abstraction levels underpinning many AI implementations. It further posits per-layer complexity-performance trade-offs and inter-layer control-complexity trade-offs, presenting the framework as a conceptual map to guide targeting of innovation, safety assurance, performance optimization under constraints, and mapping of social/economic opportunities.

Significance. If adopted, the framing could help organize the breadth of AI subfields and highlight trade-off considerations relevant to functionality and safety. The paper's explicit positioning as a conceptual proposal rather than a derived or measured result is a strength that matches the perspective format and avoids unsupported empirical claims.

minor comments (1)
  1. [Abstract] Abstract: the five layers are referenced but neither enumerated nor briefly characterized; adding a short list or one-sentence description of each layer would improve immediate readability without altering the perspective nature of the work.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for the positive assessment of our perspective piece, the recognition of its conceptual nature, and the recommendation of minor revision. No specific major comments were provided in the report.

Circularity Check

0 steps flagged

No significant circularity

full rationale

The paper is a perspective piece that directly proposes a five-layer abstraction system and two trade-offs as a conceptual map for AI implementations. No equations, fitted parameters, or derivations appear in the provided text. The central claim is framed as a useful framing rather than a result forced by self-definition, self-citation chains, or renaming of prior results. No load-bearing steps reduce to the paper's own inputs by construction.

Axiom & Free-Parameter Ledger

0 free parameters · 3 axioms · 1 invented entities

The framework rests entirely on author-introduced domain assumptions about the existence and properties of the five layers and trade-offs, with no free parameters, external benchmarks, or independent evidence supplied.

axioms (3)
  • ad hoc to paper AI implementations can be decomposed into exactly five layers of abstraction.
    Introduced as the core proposal without derivation or citation to prior layered models that justify the count of five.
  • ad hoc to paper Each layer exhibits a complexity-performance trade-off.
    Posited as a general property without supporting analysis or examples in the abstract.
  • ad hoc to paper Layers are interlocked via a control-complexity trade-off.
    Posited without derivation or external grounding.
invented entities (1)
  • Five-layer abstraction system for AI no independent evidence
    purpose: To serve as a conceptual map for innovation, safety, and exploitation in AI.
    New organizing structure introduced by the paper.

pith-pipeline@v0.9.0 · 5707 in / 1236 out tokens · 32495 ms · 2026-05-24T18:13:59.230032+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

78 extracted references · 78 canonical work pages

  1. [1]

    <Internal signals> are interpreted as <internal data structures>, and many of those together give rise to <data output>

    with notable examples found in the SPA [8] and ACT-R [9]. At such high level the process of learning is largely behaviou ral and educational; learning processes target desired behaviours and equip the system with the high-level declarative knowledge required to successfully execute them. These obj ectives relate to the fields of reinforcement learning and...

    work page

  2. [2]

    BSIM4. 3.0 MOSFET Model,

    X. J. Xi et al., “BSIM4. 3.0 MOSFET Model,” 2003

    work page 2003

  3. [3]

    Deglitching circuit,

    Hajime Obinata, “Deglitching circuit,” 5,034,744, 08-Nov-1989

    work page 1989

  4. [4]

    Method and system for deglitching in a mobile multimedia processor,

    Timothy James Ramsdale, “Method and system for deglitching in a mobile multimedia processor,” 7,793,007, 19-Jan-2006

    work page 2006

  5. [5]

    Low-density parity-check codes,

    R. Gallager, “Low-density parity-check codes,” IEEE Trans. Inf. Theory, vol. 8, no. 1, pp. 21–28, Jan. 1962

    work page 1962

  6. [6]

    From Understanding Computation to Understanding Neural Circuitry,

    D. Marr and T. Poggio, “From Understanding Computation to Understanding Neural Circuitry,” May 1976

    work page 1976

  7. [7]

    Multi-agent systems : an introduction to distributed artificial intelligence,

    J. Ferber, “Multi-agent systems : an introduction to distributed artificial intelligence,” Harlow: Addison Wesley Longman, p. 509, 1999

    work page 1999

  8. [8]

    Toward a unified catalog of implemented cognitive architectures,

    A. V. Samsonovich, “Toward a unified catalog of implemented cognitive architectures,” Front. Artif. Intell. Appl., vol. 221, pp. 195–244, 2010

    work page 2010

  9. [9]

    A Large-Scale Model of the Functioning Brain,

    C. Eliasmith et al., “A Large-Scale Model of the Functioning Brain,” Science (80-. )., vol. 338, no. 6111, pp. 1202–1205, Nov. 2012

    work page 2012

  10. [10]

    ACT-R: A Theory of Higher Level Cognition and Its Relation to Visual Attention,

    J. R. Anderson, M. Matessa, and C. Lebiere, “ACT-R: A Theory of Higher Level Cognition and Its Relation to Visual Attention,” Human–Computer Interact., vol. 12, no. 4, pp. 439–462, Dec. 1997

    work page 1997

  11. [11]

    Knowledge Representation: Logical, Philosophical, and Computational Foundations,

    S. C. Shapiro, “Knowledge Representation: Logical, Philosophical, and Computational Foundations,” Comput. Linguist., vol. 27, no. 2, pp. 286–294, 2001

    work page 2001

  12. [12]

    The Moral Machine experiment,

    E. Awad et al., “The Moral Machine experiment,” Nature, p. 1, Oct. 2018

    work page 2018

  13. [13]

    The binding problem Abbreviations,

    A. Treisman, “The binding problem Abbreviations,” Curr. Opin. Neurobiol., vol. 6, no. 2, pp. 171–78, 1996

    work page 1996

  14. [14]

    {G}love: {G}lobal {V}ectors for {W}ord {R}epresentation.,

    J. Pennington, R. Socher, and C. D. Manning, “{G}love: {G}lobal {V}ectors for {W}ord {R}epresentation.,” Proc. Conf. Empir. Methods Nat. Lang. Process., vol. 14, pp. 1532–1543, 2014

    work page 2014

  15. [15]

    Producing high-dimensional semantic spaces from lexical co- occurrence,

    K. Lund and C. Burgess, “Producing high-dimensional semantic spaces from lexical co- occurrence,” Behav. Res. Methods, Instruments, Comput., vol. 28, no. 2, pp. 203–208, Jun. 1996

    work page 1996

  16. [16]

    Fully Distributed Representation,

    P. Kanerva, “Fully Distributed Representation,” Proc. 1997 Real World Comput. Symp., no. c, pp. 358–365, 1997

    work page 1997

  17. [17]

    Hawkins and S

    J. Hawkins and S. Blakeslee, On intelligence. Henry Holt and Company, 2013

    work page 2013

  18. [18]

    Eliasmith, How to Build a Brain

    C. Eliasmith, How to Build a Brain. 2013

    work page 2013

  19. [19]

    Deep learning,

    Y. Lecun, Y. Bengio, and G. Hinton, “Deep learning,” Nature, vol. 521, no. 7553, pp. 436–444, 2015

    work page 2015

  20. [20]

    Contrast Gain Control in Auditory Cortex,

    N. C. Rabinowitz, B. D. B. Willmore, J. W. H. Schnupp, and A. J. King, “Contrast Gain Control in Auditory Cortex,” Neuron, vol. 70, no. 6, pp. 1178–1191, Jun. 2011

    work page 2011

  21. [21]

    Properties of the surround response mechanism of cat retinal ganglion cells and centre-surround interaction,

    C. Enroth-Cugell and L. H. Pinto, “Properties of the surround response mechanism of cat retinal ganglion cells and centre-surround interaction,” J. Physiol., vol. 220, no. 2, pp. 403– 439, Jan. 1972

    work page 1972

  22. [22]

    Dynamic organization of motor control within the olivocerebellar system,

    J. P. Welsh, E. J. Lang, I. Suglhara, and R. Llinás, “Dynamic organization of motor control within the olivocerebellar system,” Nature, vol. 374, no. 6521, pp. 453–457, Mar. 1995

    work page 1995

  23. [23]

    Design and Tool Flow of IBM’s TrueNorth,

    F. Akopyan and Filipp, “Design and Tool Flow of IBM’s TrueNorth,” in Proceedings of the 2016 on International Symposium on Physical Design - ISPD ’16, 2016, pp. 59–60

    work page 2016

  24. [24]

    Simple spike activity predicts occurrence of complex spikes in cerebellar Purkinje cells,

    R. C. Miall, J. G. Keating, M. Malkmus, and W. T. Thach, “Simple spike activity predicts occurrence of complex spikes in cerebellar Purkinje cells,” pp. 13–15, 1998

    work page 1998

  25. [25]

    A Current-Mode Hysteretic Winner-take-all Network, with Excitatory and Inhibitory Coupling,

    G. Indiveri, “A Current-Mode Hysteretic Winner-take-all Network, with Excitatory and Inhibitory Coupling,” Analog Integr. Circuits Signal Process., vol. 28, no. 3, pp. 279–291, Sep. 2001

    work page 2001

  26. [26]

    Going deeper with convolutions,

    C. Szegedy et al., “Going deeper with convolutions,” in 2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2015, pp. 1–9

    work page 2015

  27. [27]

    Deep Speech: Scaling up end-to-end speech recognition,

    A. Hannun et al., “Deep Speech: Scaling up end-to-end speech recognition,” Dec. 2014

    work page 2014

  28. [28]

    ImageNet Classification with Deep Convolutional Neural Networks,

    A. Krizhevsky, I. Sutskever, and G. E. Hinton, “ImageNet Classification with Deep Convolutional Neural Networks,” Adv. Neural Inf. Process. Syst., pp. 1–9, 2012

    work page 2012

  29. [29]

    Reservoir computing approaches to recurrent neural network training,

    M. Lukosevicius and H. Jaeger, “Reservoir computing approaches to recurrent neural network training,” Comput. Sci. Rev., vol. 3, no. 3, pp. 127–149, Aug. 2009

    work page 2009

  30. [30]

    Deep Boltzmann Machines

    R. Salakhutdinov and G. Hinton, “Deep Boltzmann Machines.”

    work page

  31. [31]

    LSTM: A Search Space Odyssey,

    K. Greff, R. K. Srivastava, J. Koutnik, B. R. Steunebrink, and J. Schmidhuber, “LSTM: A Search Space Odyssey,” IEEE Trans. Neural Networks Learn. Syst., vol. 28, no. 10, pp. 2222–2232, Oct. 2017

    work page 2017

  32. [32]

    A Hebbian/Anti-Hebbian Neural Network for Linear Subspace Learning: A Derivation from Multidimensional Scaling of Streaming Data,

    C. Pehlevan, T. Hu, and D. B. Chklovskii, “A Hebbian/Anti-Hebbian Neural Network for Linear Subspace Learning: A Derivation from Multidimensional Scaling of Streaming Data,” Neural Comput., vol. 27, no. 7, pp. 1461–1495, Jul. 2015

    work page 2015

  33. [33]

    Back propagation implementation on the adaptive solutions CNAPS neurocomputer chip,

    H. McCartor, “Back propagation implementation on the adaptive solutions CNAPS neurocomputer chip,” Proc. Conf. Adv. Neural Inf. Process. Syst., pp. 1028–1031, 1990

    work page 1990

  34. [34]

    K-winners-take-all circuit with O(N) complexity.,

    K. Urahama and T. Nagao, “K-winners-take-all circuit with O(N) complexity.,” IEEE Trans. Neural Netw., vol. 6, no. 3, pp. 776–8, Jan. 1995

    work page 1995

  35. [35]

    An Artificial Neuron Model with a Periodic Activation Function,

    M. Nakagawa, “An Artificial Neuron Model with a Periodic Activation Function,” J. Phys. Soc. Japan, vol. 64, no. 3, pp. 1023–1031, Mar. 1995

    work page 1995

  36. [36]

    Simple sigmoid-like activation function suitable for digital hardware implementation,

    H. K. Kwan, “Simple sigmoid-like activation function suitable for digital hardware implementation,” IET Electron. Lett., vol. 28, no. 15, pp. 1379–1380, 1992

    work page 1992

  37. [37]

    Understanding the difficulty of training deep feedforward neural networks

    X. Glorot and Y. Bengio, “Understanding the difficulty of training deep feedforward neural networks.”

    work page

  38. [38]

    A Review of Spiking Neuron Models,

    R. Coelho and M. Gerais, “A Review of Spiking Neuron Models,” no. November, 2005

    work page 2005

  39. [39]

    Simple Model of Spiking Neurons,

    E. M. Izhikevich, “Simple Model of Spiking Neurons,” vol. 14, no. 6, pp. 1569–1572, 2003

    work page 2003

  40. [40]

    Homeostatic plasticity in neuronal networks: the more things change, the more they stay the same.,

    G. G. Turrigiano, “Homeostatic plasticity in neuronal networks: the more things change, the more they stay the same.,” Trends Neurosci., vol. 22, no. 5, pp. 221–227, 1999

    work page 1999

  41. [41]

    Learning Spike Time Codes Through Morphological Learning With Binary Synapses,

    S. Roy, P. P. San, S. Hussain, L. W. Wei, and A. Basu, “Learning Spike Time Codes Through Morphological Learning With Binary Synapses,” IEEE Trans. Neural Networks Learn. Syst., vol. 27, no. 7, pp. 1572–1577, Jul. 2016

    work page 2016

  42. [42]

    Neural learning rules for the vestibulo-ocular reflex.,

    J. L. Raymond and S. G. Lisberger, “Neural learning rules for the vestibulo-ocular reflex.,” J. Neurosci., vol. 18, no. 21, pp. 9112–9129, 1998

    work page 1998

  43. [43]

    Anti-hebbian spike-timing-dependent plasticity and adaptive sensory processing.,

    P. D. Roberts and T. K. Leen, “Anti-hebbian spike-timing-dependent plasticity and adaptive sensory processing.,” Front. Comput. Neurosci., vol. 4, no. December, p. 156, Jan. 2010

    work page 2010

  44. [44]

    The tempotron: a neuron that learns spike timing-based decisions.,

    R. Gütig and H. Sompolinsky, “The tempotron: a neuron that learns spike timing-based decisions.,” Nat. Neurosci., vol. 9, no. 3, Mar. 2006

    work page 2006

  45. [45]

    Synaptic Modifications in Cultured Hippocampal Neurons: Dependence on Spike Timing, Synaptic Strength, and Postsynaptic Cell Type,

    G. Bi and M. Poo, “Synaptic Modifications in Cultured Hippocampal Neurons: Dependence on Spike Timing, Synaptic Strength, and Postsynaptic Cell Type,” J. Neurosci., vol. 18, no. 24, pp. 10464–10472, Dec. 1998

    work page 1998

  46. [46]

    Spike Timing-Dependent Plasticity of Neural Circuits,

    Y. Dan and M. Poo, “Spike Timing-Dependent Plasticity of Neural Circuits,” Neuron, vol. 44, no. 1, pp. 23–30, 2004

    work page 2004

  47. [47]

    Parallel distributed processing: explorations in the microstructure of cognition. Volume 1. Foundations

    D. E. Rumelhart and J. L. Mcclelland, “Parallel distributed processing: explorations in the microstructure of cognition. Volume 1. Foundations.” MIT Press,Cambridge, MA, 01-Jan- 1986

    work page 1986

  48. [48]

    Bioinspired Programming of Memory Devices for Implementing an Inference Engine,

    D. Querlioz, O. Bichler, A. F. Vincent, and C. Gamrat, “Bioinspired Programming of Memory Devices for Implementing an Inference Engine,” Proc. IEEE, vol. 103, no. 8, pp. 1398–1416, Aug. 2015

    work page 2015

  49. [49]

    A stochastic version of the delta rule,

    S. J. Hanson, “A stochastic version of the delta rule,” Phys. D Nonlinear Phenom., vol. 42, no. 1–3, pp. 265–272, Jun. 1990

    work page 1990

  50. [50]

    Increased rates of convergence through learning rate adaptation,

    R. A. Jacobs, “Increased rates of convergence through learning rate adaptation,” Neural Networks, vol. 1, no. 4, pp. 295–307, Jan. 1988

    work page 1988

  51. [51]

    Computational principles of biological memory,

    M. K. Benna and S. Fusi, “Computational principles of biological memory,” p. 21, Jul. 2015

    work page 2015

  52. [52]

    Cascade models of synaptically stored memories.,

    S. Fusi, P. J. Drew, and L. F. Abbott, “Cascade models of synaptically stored memories.,” Neuron, vol. 45, no. 4, pp. 599–611, Feb. 2005

    work page 2005

  53. [53]

    Matching properties of MOS transistors,

    M. J. M. Pelgrom, A. C. J. Duinmaijer, and A. P. G. Welbers, “Matching properties of MOS transistors,” IEEE J. Solid-State Circuits, vol. 24, no. 5, pp. 1433–1439, Oct. 1989

    work page 1989

  54. [54]

    CMOS SRAM scaling limits under optimum stability constraints,

    A. Makosiej, O. Thomas, A. Amara, and A. Vladimirescu, “CMOS SRAM scaling limits under optimum stability constraints,” in 2013 IEEE International Symposium on Circuits and Systems (ISCAS2013), 2013, pp. 1460–1463

    work page 2013

  55. [55]

    Nanoionics-based resistive switching memories.,

    R. Waser and M. Aono, “Nanoionics-based resistive switching memories.,” Nat. Mater., vol. 6, no. 11, pp. 833–40, Nov. 2007

    work page 2007

  56. [56]

    Neuromorphic silicon neuron circuits.,

    G. Indiveri et al., “Neuromorphic silicon neuron circuits.,” Front. Neurosci., vol. 5, no. May, p. 73, Jan. 2011

    work page 2011

  57. [57]

    Spin-transfer torque magnetic memory as a stochastic memristive synapse for neuromorphic systems.,

    A. F. Vincent et al., “Spin-transfer torque magnetic memory as a stochastic memristive synapse for neuromorphic systems.,” IEEE Trans. Biomed. Circuits Syst., vol. 9, no. 2, pp. 166– 74, Apr. 2015

    work page 2015

  58. [58]

    Experimental Demonstration and Tolerancing of a Large-Scale Neural Network (165 000 Synapses) Using Phase-Change Memory as the Synaptic Weight Element,

    G. W. Burr et al., “Experimental Demonstration and Tolerancing of a Large-Scale Neural Network (165 000 Synapses) Using Phase-Change Memory as the Synaptic Weight Element,” IEEE Trans. Electron Devices, vol. 62, no. 11, pp. 3498–3507, Nov. 2015

    work page 2015

  59. [59]

    Supervised Learning Using Spike-Timing-Dependent Plasticity of Memristive Synapses.,

    Y. Nishitani, Y. Kaneko, and M. Ueda, “Supervised Learning Using Spike-Timing-Dependent Plasticity of Memristive Synapses.,” IEEE Trans. neural networks Learn. Syst., vol. 26, no. 12, pp. 2999–3008, Dec. 2015

    work page 2015

  60. [60]

    Nanoscale memristor device as synapse in neuromorphic systems,

    S. H. Jo, T. Chang, I. Ebong, B. B. Bhadviya, P. Mazumder, and W. Lu, “Nanoscale memristor device as synapse in neuromorphic systems,” Nano Lett., vol. 10, no. 4, pp. 1297–1301, Apr. 2010

    work page 2010

  61. [61]

    Analog memristive synapse in spiking networks implementing unsupervised learning,

    E. Covi, S. Brivio, A. Serb, T. Prodromakis, M. Fanciulli, and S. Spiga, “Analog memristive synapse in spiking networks implementing unsupervised learning,” Front. Neurosci., vol. 10, no. OCT, 2016

    work page 2016

  62. [62]

    Multibit memory operation of metal-oxide bi-layer memristors,

    S. Stathopoulos et al., “Multibit memory operation of metal-oxide bi-layer memristors,” Sci. Rep., vol. 7, no. 1, p. 17532, Dec. 2017

    work page 2017

  63. [63]

    High Density Crossbar Arrays with Sub- 15 nm Single Cells via Liftoff Process Only,

    A. Khiat et al., “High Density Crossbar Arrays with Sub- 15 nm Single Cells via Liftoff Process Only,” Sci. Rep., vol. 6, p. 32614, Sep. 2016

    work page 2016

  64. [64]

    A non-volatile organic electrochemical device as a low-voltage artificial synapse for neuromorphic computing,

    Y. Van De Burgt et al., “A non-volatile organic electrochemical device as a low-voltage artificial synapse for neuromorphic computing,” Nat. Mater., vol. 16, no. 4, pp. 414–418, Apr. 2017

    work page 2017

  65. [65]

    Unsupervised learning in probabilistic neural networks with multi-state metal-oxide memristive synapses,

    A. Serb, J. Bill, A. Khiat, R. Berdan, R. Legenstein, and T. Prodromakis, “Unsupervised learning in probabilistic neural networks with multi-state metal-oxide memristive synapses,” Nat. Commun., vol. 7, p. 12611, Sep. 2016

    work page 2016

  66. [66]

    Single transistor learning synapse with long term storage,

    P. Hasler, C. Diorio, B. A. Minch, and C. Mead, “Single transistor learning synapse with long term storage,” in Proceedings of ISCAS’95 - International Symposium on Circuits and Systems, 1995, vol. 3, pp. 1660–1663

    work page 1995

  67. [67]

    Hardware-Level Bayesian Inference,

    L. T.-T. and T. P. A. Serb, E. Manino, I. Messaris, “Hardware-Level Bayesian Inference,” in 31st Conference on Neural Information Processing Systems (NIPS), 2017

    work page 2017

  68. [68]

    Always-on vision processing unit for mobile applications,

    B. Barry et al., “Always-on vision processing unit for mobile applications,” IEEE Micro, vol. 35, no. 2, pp. 56–66, 2015

    work page 2015

  69. [69]

    NVIDIA tensor core programmability, performance & precision,

    S. Markidis, S. W. Der Chien, E. Laure, I. B. Peng, and J. S. Vetter, “NVIDIA tensor core programmability, performance & precision,” Proc. - 2018 IEEE 32nd Int. Parallel Distrib. Process. Symp. Work. IPDPSW 2018, pp. 522–531, 2018

    work page 2018

  70. [70]

    Loihi: A Neuromorphic Manycore Processor with On-Chip Learning,

    M. Davies et al., “Loihi: A Neuromorphic Manycore Processor with On-Chip Learning,” IEEE Micro, vol. 38, no. 1, pp. 82–99, 2018

    work page 2018

  71. [71]

    The SpiNNaker Project,

    S. B. Furber, F. Galluppi, S. Temple, and L. A. Plana, “The SpiNNaker Project,” Proc. IEEE, vol. 102, no. 5, pp. 652–665, May 2014

    work page 2014

  72. [72]

    A Scalable Multicore Architecture With Heterogeneous Memory Structures for Dynamic Neuromorphic Asynchronous Processors (DYNAPs),

    S. Moradi, N. Qiao, F. Stefanini, and G. Indiveri, “A Scalable Multicore Architecture With Heterogeneous Memory Structures for Dynamic Neuromorphic Asynchronous Processors (DYNAPs),” IEEE Trans. Biomed. Circuits Syst., no. August, 2017

    work page 2017

  73. [73]

    Training and operation of an integrated neuromorphic network based on metal-oxide memristors,

    M. Prezioso, F. Merrikh-Bayat, B. D. Hoskins, G. C. Adam, K. K. Likharev, and D. B. Strukov, “Training and operation of an integrated neuromorphic network based on metal-oxide memristors,” Nature, vol. 521, no. 7550, pp. 61–64, May 2015

    work page 2015

  74. [74]

    Single pairing spike-timing dependent plasticity in BiFeO3 memristors with a time window of 25 ms to 125 μs.,

    N. Du et al., “Single pairing spike-timing dependent plasticity in BiFeO3 memristors with a time window of 25 ms to 125 μs.,” Front. Neurosci., vol. 9, p. 227, Jan. 2015

    work page 2015

  75. [75]

    STDP and STDP variations with memristors for spiking neuromorphic learning systems,

    T. Serrano-Gotarredona, T. Masquelier, T. Prodromakis, G. Indiveri, and B. Linares-Barranco, “STDP and STDP variations with memristors for spiking neuromorphic learning systems,” Front. Neurosci., vol. 7, 2013

    work page 2013

  76. [76]

    Equivalent-accuracy accelerated neural-network training using analogue memory,

    S. Ambrogio et al., “Equivalent-accuracy accelerated neural-network training using analogue memory,” Nature, vol. 558, no. 7708, pp. 60–67, Jun. 2018

    work page 2018

  77. [77]

    NVM neuromorphic core with 64k-cell (256-by-256) phase change memory synaptic array with on-chip neuron circuits for continuous in-situ learning,

    S. Kim et al., “NVM neuromorphic core with 64k-cell (256-by-256) phase change memory synaptic array with on-chip neuron circuits for continuous in-situ learning,” in 2015 IEEE International Electron Devices Meeting (IEDM), 2015, p. 17.1.1-17.1.4

    work page 2015

  78. [78]

    TrueNorth: Design and Tool Flow of a 65 mW 1 Million Neuron Programmable Neurosynaptic Chip,

    F. Akopyan et al., “TrueNorth: Design and Tool Flow of a 65 mW 1 Million Neuron Programmable Neurosynaptic Chip,” IEEE Trans. Comput. Des. Integr. Circuits Syst., vol. 34, no. 10, pp. 1537–1557, Oct. 2015

    work page 2015