Bayesian Optimisation of Large-scale Photonic Reservoir Computers

被引:13
作者
Antonik, Piotr [1 ,2 ]
Marsal, Nicolas [1 ,2 ]
Brunner, Daniel [3 ,4 ]
Rontani, Damien [1 ,2 ]
机构
[1] Cent Supelec, LMOPS EA 4423 Lab, 2 Rue Edouard Belin, F-57070 Metz, France
[2] Univ Lorraine, 2 Rue Edouard Belin, F-57070 Metz, France
[3] CNRS, Opt Dept, FEMTO ST Inst, 15B Ave Montboucons, F-25030 Besancon, France
[4] Univ Bourgogne Franche Comte, 15B Ave Montboucons, F-25030 Besancon, France
关键词
Bayesian optimisation; Photonic reservoir computing; Large-scale networks; Hyper-parameter optimisation; FRAMEWORK; SYSTEMS;
D O I
10.1007/s12559-020-09732-6
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
Reservoir computing is a growing paradigm for simplified training of recurrent neural networks, with a high potential for hardware implementations. Numerous experiments in optics and electronics yield comparable performance with digital state-of-the-art algorithms. Many of the most recent works in the field focus on large-scale photonic systems, with tens of thousands of physical nodes and arbitrary interconnections. While this trend significantly expands the potential applications of photonic reservoir computing, it also complicates the optimisation of the high number of hyper-parameters of the system. In this work, we propose the use of Bayesian optimisation for efficient exploration of the hyper-parameter space in a minimum number of iteration. We test this approach on a previously reported large-scale experimental system, compare it with the commonly used grid search, and report notable improvements in performance and the number of experimental iterations required to optimise the hyper-parameters. Bayesian optimisation thus has the potential to become the standard method for tuning the hyper-parameters in photonic reservoir computing.
引用
收藏
页码:1452 / 1460
页数:9
相关论文
共 42 条
[1]  
Akrout A., 2016, ARXIV161208606
[2]  
[Anonymous], 2002, A tutorial on principal components analysis
[3]  
[Anonymous], 2016, ARXIV161105193
[4]  
[Anonymous], 2006, 2006 07 FORECASTING
[5]   Human action recognition with a large-scale brain-inspired photonic computer [J].
Antonik, Piotr ;
Marsal, Nicolas ;
Brunner, Daniel ;
Rontani, Damien .
NATURE MACHINE INTELLIGENCE, 2019, 1 (11) :530-537
[6]   Large-Scale Spatiotemporal Photonic Reservoir Computer for Image Classification [J].
Antonik, Piotr ;
Marsal, Nicolas ;
Rontani, Damien .
IEEE JOURNAL OF SELECTED TOPICS IN QUANTUM ELECTRONICS, 2020, 26 (01)
[7]   Online Training of an Opto-Electronic Reservoir Computer Applied to Real-Time Channel Equalization [J].
Antonik, Piotr ;
Duport, Francois ;
Hermans, Michiel ;
Smerieri, Anteo ;
Haelterman, Marc ;
Massar, Serge .
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS, 2017, 28 (11) :2686-2698
[8]   Online Training for High-Performance Analogue Readout Layers in Photonic Reservoir Computers [J].
Antonik, Piotr ;
Haelterman, Marc ;
Massar, Serge .
COGNITIVE COMPUTATION, 2017, 9 (03) :297-306
[9]   Information processing using a single dynamical node as complex system [J].
Appeltant, L. ;
Soriano, M. C. ;
Van der Sande, G. ;
Danckaert, J. ;
Massar, S. ;
Dambre, J. ;
Schrauwen, B. ;
Mirasso, C. R. ;
Fischer, I. .
NATURE COMMUNICATIONS, 2011, 2
[10]  
Bahi HE, 2015, ROBUST SYSTEM PRINTE