Unsupervised learning of event-based image recordings using spike-timing-dependent plasticity

Iyer, Laxmi; Basu, Arindam

doi:10.1109/ijcnn.2017.7966074

Cited by 20 publications

(28 citation statements)

References 19 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Without any device variability, the mean recognition rate is 92.4 % ( Fig. 4(a)), close to the 93.68 % rate from the literature with a slightly different STDP and 400 output neurons [19].…”

Section: When One Needs a Longer Term Input Memorysupporting

confidence: 83%

See 1 more Smart Citation

Hardware Spiking Neural Networks: Slow Tasks Resilient Learning with Longer Term-Memory Bits

Lewden

Vincent

Meyer

et al. 2019

2019 IEEE Biomedical Circuits and Systems Conference (BioCAS)

View full text Add to dashboard Cite

When aiming at efficient and low-power processing of event-based data, hardware implementations of spiking neural networks that co-integrate analog silicon neurons with memristive synaptic crossbar arrays are a promising framework. Fully analog systems however commonly make it difficult to learn patterns with real-world timescales, which are typically beyond the millisecond, due to intrinsic constraints of the underlying technologies. In this work, we propose to alleviate this issue by supplementing each presynaptic unit with a single memory bit, which allows to implement a hardware-friendly Spike Timing-Dependent Plasticity. By simulation means on the N-MNIST dataset, we illustrate the potential of this concept and show its robustness to postsynaptic neuron variability. We also discuss how to circumvent challenges raised by initial weight distribution. These results could facilitate the emergence of embedded smart systems directly fed by event-based sensors.

show abstract

“…Without any device variability, the mean recognition rate is 92.4 % ( Fig. 4(a)), close to the 93.68 % rate from the literature with a slightly different STDP and 400 output neurons [19].…”

Section: When One Needs a Longer Term Input Memorysupporting

confidence: 83%

“…This makes the hardware implementation simpler: the timer needed for each presynaptic neuron is replaced with logic blocks. Similarly to other works in the field, we focus on the subset of digits 5, 6 and 9 to reduce simulation time [19]. One epoch thus corresponds to 17,288 (2859) training (test) samples.…”

Section: When One Needs a Longer Term Input Memorymentioning

confidence: 99%

Hardware Spiking Neural Networks: Slow Tasks Resilient Learning with Longer Term-Memory Bits

Lewden

Vincent

Meyer

et al. 2019

2019 IEEE Biomedical Circuits and Systems Conference (BioCAS)

View full text Add to dashboard Cite

show abstract

“…TABLE I shows that the recognition performance of our approach is higher than that of Zhao's method [10], BOE [7] and HFirst [12]. In addition, compared with Iyer & Basu's unsupervised model [22] on NMNIST, which achieves the accuracy of 80.63%, our approach can give higher accuracy of 89.70%. 4) On AER Posture dataset: In this dataset, we randomly select 80% of human actions for training and the others for testing.…”

Section: Experiments Settingsmentioning

confidence: 84%

“…Diehl et al [19] proposed a SNN for image recognition that employs STDP learning to process the Poisson-distributed spike-trains with firing rates proportional to the intensity of the image pixel. Iyer et al [22] applied the Diehl's model [19] on native AER data. Experiments on the N-MNIST dataset [23] show that the method provides an effective unsupervised application on AER event streams.…”

Section: Introductionmentioning

confidence: 99%

Unsupervised AER Object Recognition Based on Multiscale Spatio-Temporal Features and Spiking Neurons

Liu

Pan

Ruan

et al. 2020

IEEE Trans. Neural Netw. Learning Syst.

View full text Add to dashboard Cite

This paper proposes an unsupervised address event representation (AER) object recognition approach. The proposed approach consists of a novel multiscale spatio-temporal feature (MuST) representation of input AER events and a spiking neural network (SNN) using spike-timing-dependent plasticity (STDP) for object recognition with MuST. MuST extracts the features contained in both the spatial and temporal information of AER event flow, and meanwhile forms an informative and compact feature spike representation. We show not only how MuST exploits spikes to convey information more effectively, but also how it benefits the recognition using SNN. The recognition process is performed in an unsupervised manner, which does not need to specify the desired status of every single neuron of SNN, and thus can be flexibly applied in real-world recognition tasks. The experiments are performed on five AER datasets including a new one named GESTURE-DVS. Extensive experimental results show the effectiveness and advantages of this proposed approach.

show abstract

“…Some studies group events into frames or time slices imposing fixed timescales [6][7][8][9][10][11][12][13][14][15]. Other studies process the data stream on an event-by-event basis [16][17][18][19][20][21][22][23][24][25][26][27][28][29], with the timing of an event determining its relevance to an internal model. Some event-based studies accumulate event activities and then decay them as time passes [30][31][32][33][34].…”

Section: Processing the Temporal And Spatial Information In Event-basmentioning

confidence: 99%

Inspired by nature: timescale-free and grid-free event-based computing with spiking neural networks

Gibson¹

View full text Add to dashboard Cite

Conception and design (50%) Analysis and interpretation (70%) Drafting and production (70%) Author Scott Heath Conception and design (2%) Analysis and interpretation (10%) Drafting and production (2%) Author Robert P. Quinn Conception and design (2%) Analysis and interpretation (0%) Drafting and production (2%) Author Alexia H. Lee Conception and design (2%) Analysis and interpretation (0%) Drafting and production (2%) Author Joshua T. Arnold Conception and design (3%) Analysis and interpretation (10%) Drafting and production (2%) Author Tharun S. Sonti Conception and design (3%) Analysis and interpretation (10%) Drafting and production (2%) Author Andrew Whalley Conception and design (2%) Analysis and interpretation (0%) Drafting and production (2%) Author George P. Shannon Conception and design (2%) Analysis and interpretation (0%) Drafting and production (2%) Author Brian T. Song Conception and design (2%) Analysis and interpretation (0%) Drafting and production (2%) Author James A. Henderson Conception and design (2%) Analysis and interpretation (0%) Drafting and production (2%) Author Janet Wiles Conception and design (30%) Analysis and interpretation (0%) Drafting and production (12%) 7 [2] Tingting (Amy) Gibson, James A. Henderson, and Janet Wiles, "Predicting temporal sequences using an event-based spiking neural network incorporating learnable delays," in Neural Networks (IJCNN), 2014 International Joint Conference on. IEEE, 2014, pp. 3213-3220.-incorporated as Chapter 4 Contributor Statement of contribution Author Tingting (Amy) Gibson (Candidate) Conception and design (70%) Analysis and interpretation (80%) Drafting and production (80%) Author James A. Henderson Conception and design (0%) Analysis and interpretation (0%) Drafting and production (10%) Author Janet Wiles Conception and design (30%) Analysis and interpretation (20%) Drafting and production (10%) List of Abbreviations DVS Dynamic vision sensor ANN Artificial neural network SNN Spiking neural network AER Address-event representation iRat Intelligent Rat animat Technology MNIST Modified National Institute of Standards and Technology CIS Complex and Intelligent Systems (CIS is now part of the Co-Innovation group, CIN) D D D KL Kullback-Leibler divergence 1. Introduction to timescale-free and grid-free events with bio-inspired spiking neural networks 1. Introduction to timescale-free and grid-free events with bio-inspired spiking neural networks

show abstract

Unsupervised learning of event-based image recordings using spike-timing-dependent plasticity

Cited by 20 publications

References 19 publications

Hardware Spiking Neural Networks: Slow Tasks Resilient Learning with Longer Term-Memory Bits

Hardware Spiking Neural Networks: Slow Tasks Resilient Learning with Longer Term-Memory Bits

Unsupervised AER Object Recognition Based on Multiscale Spatio-Temporal Features and Spiking Neurons

Inspired by nature: timescale-free and grid-free event-based computing with spiking neural networks

Contact Info

Product

Resources

About