Abstract
Liquid State Machine (LSM) is a spiking variant of recurrent neural networks with promising results for speech, video and other temporal datasets classification. LSM employ a network of fixed and randomly connected neurons, called a reservoir. Parameter selection for building the best performing reservoir is a difficult task given the vast parameter space. A memory metric extracted from a state-space approximation of the LSM has been proposed in the past and empirically shown to be best-in-class for performance prediction. However, the working principle of this memory metric has not been studied. We first show equivalence of LSM simulated on MATLAB to those run on Intel’s neuromorphic chip Loihi. This enables us to perform in-depth statistical analysis of the memory metric on Loihi: effect of weight scaling and effect of time averaging window. Analysis of state space matrices generated with a reasonably sized averaging window reveal that the diagonal elements are sufficient to capture network dynamics. This strengthens the relevance of the first order decay constant based memory metric which correlates well with the classification performance.
This is a preview of subscription content, log in via an institution to check access.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
References
de Azambuja, R., Klein, F.B., Adams, S.V., Stoelen, M.F., Cangelosi, A.: Short-term plasticity in a liquid state machine biomimetic robot arm controller. In: 2017 International Joint Conference on Neural Networks (IJCNN), pp. 3399–3408. IEEE (2017)
Chrol-Cannon, J., Jin, Y.: On the correlation between reservoir metrics and performance for time series classification under the influence of synaptic plasticity. PLoS ONE 9(7), e101792 (2014)
Das, A., et al.: Unsupervised heart-rate estimation in wearables with liquid states and a probabilistic readout. Neural Netw. 99, 134–147 (2018)
Davies, M., et al.: Loihi: a neuromorphic manycore processor with on-chip learning. IEEE Micro 38(1), 82–99 (2018)
Gibbons, T.E.: Unifying quality metrics for reservoir networks. In: The 2010 International Joint Conference on Neural Networks (IJCNN), pp. 1–7. IEEE (2010)
Gorad, A., Saraswat, V., Ganguly, U.: Predicting performance using approximate state space model for liquid state machines. In: 2019 International Joint Conference on Neural Networks (IJCNN), pp. 1–8. IEEE (2019)
Ju, H., Xu, J.X., Chong, E., VanDongen, A.M.: Effects of synaptic connectivity on liquid state machine performance. Neural Netw. 38, 39–51 (2013)
Legenstein, R., Maass, W.: Edge of chaos and prediction of computational performance for neural circuit models. Neural Netw. 20(3), 323–334 (2007)
Lyon, R.: A computational model of filtering, detection, and compression in the cochlea. In: IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 1982, vol. 7, pp. 1282–1285. IEEE (1982)
Maass, W., Natschläger, T., Markram, H.: Real-time computing without stable states: a new framework for neural computation based on perturbations. Neural Comput. 14(11), 2531–2560 (2002)
Norton, D., Ventura, D.: Improving liquid state machines through iterative refinement of the reservoir. Neurocomputing 73(16–18), 2893–2904 (2010)
Ponghiran, W., Srinivasan, G., Roy, K.: Reinforcement learning with low-complexity liquid state machines. Front. Neurosci. 13, 883 (2019)
Rosselló, J.L., Alomar, M.L., Morro, A., Oliver, A., Canals, V.: High-density liquid-state machine circuitry for time-series forecasting. Int. J. Neural Syst. 26(05), 1550036 (2016)
Soures, N., Kudithipudi, D.: Deep liquid state machines with neural plasticity for video activity recognition. Front. Neurosci. 13, 686 (2019)
Verstraeten, D., Schrauwen, B., Stroobandt, D., Van Campenhout, J.: Isolated word recognition with the liquid state machine: a case study. Inf. Process. Lett. 95(6), 521–528 (2005)
Zhang, Y., Li, P., Jin, Y., Choe, Y.: A digital liquid state machine with biologically inspired learning and its application to speech recognition. IEEE Trans. Neural Netw. Learn. Syst. 26(11), 2635–2649 (2015)
Acknowledgements
We thank Intel Neuromorphic Research Community (INRC) for providing us remote access to Loihi. We also thank Apoorv Kishore and Ajinkya Gorad for providing valuable insights and helping with the MATLAB simulations.
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2022 The Author(s), under exclusive license to Springer Nature Switzerland AG
About this paper
Cite this paper
Patel, R., Saraswat, V., Ganguly, U. (2022). Liquid State Machine on Loihi: Memory Metric for Performance Prediction. In: Pimenidis, E., Angelov, P., Jayne, C., Papaleonidas, A., Aydin, M. (eds) Artificial Neural Networks and Machine Learning – ICANN 2022. ICANN 2022. Lecture Notes in Computer Science, vol 13531. Springer, Cham. https://doi.org/10.1007/978-3-031-15934-3_57
Download citation
DOI: https://doi.org/10.1007/978-3-031-15934-3_57
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-031-15933-6
Online ISBN: 978-3-031-15934-3
eBook Packages: Computer ScienceComputer Science (R0)