MReC: A Multilayer Photonic Reservoir Computing Architecture

Photonic reservoir computing is a promising paradigm for large-scale classification and prediction problems. However, its single-layer nature is a bottleneck for higher performance and accuracy. This paper proposes MReC, a novel multi-layer photonic reservoir computing architecture made-up of photonic components such as Mach-Zehnder interferometer and optical fiber spool. The proposed design involves a deep pipelined architecture and has been synthesized using commercial photonic CAD tools. System-level simulation of large-scale tasks like spoken digit recognition, channel equalization, and time-series prediction on MReC demonstrates an accuracy improvement up to 50% compared to the state-of-the-art photonic architecture in the literature. Further results indicate a power reduction up to 34% and at least 132x speedup when compared with the best reported results in the literature at a cost of 12% area overhead.

[1]  Claudio Gallicchio,et al.  Echo State Property of Deep Reservoir Computing Networks , 2017, Cognitive Computation.

[2]  Claudio Gallicchio,et al.  Deep reservoir computing: A critical experimental analysis , 2017, Neurocomputing.

[3]  Jason Cong,et al.  FPGA-based accelerator for long short-term memory recurrent neural networks , 2017, 2017 22nd Asia and South Pacific Design Automation Conference (ASP-DAC).

[4]  Yang Yi,et al.  A deep learning based approach for analog hardware implementation of delayed feedback reservoir computing system , 2018, 2018 19th International Symposium on Quality Electronic Design (ISQED).

[5]  Serge Massar,et al.  High performance photonic reservoir computer based on a coherently driven passive cavity , 2015, ArXiv.

[6]  Tureiti Keith,et al.  A General-Purpose GPU Reservoir Computer , 2013 .

[7]  Shih-Chii Liu,et al.  Phased LSTM: Accelerating Recurrent Network Training for Long or Event-based Sequences , 2016, NIPS.

[8]  Daniel Brunner,et al.  Parallel photonic information processing at gigabyte per second data rates using transient states , 2013, Nature Communications.

[9]  Trevor Darrell,et al.  Caffe: Convolutional Architecture for Fast Feature Embedding , 2014, ACM Multimedia.

[10]  Wim Bogaerts,et al.  The IPKISS photonic design framework , 2016, 2016 Optical Fiber Communications Conference and Exhibition (OFC).

[11]  Benjamin Schrauwen,et al.  Compact hardware liquid state machines on FPGA for real-time speech recognition , 2008, Neural Networks.

[12]  Benjamin Schrauwen,et al.  An overview of reservoir computing: theory, applications and implementations , 2007, ESANN.

[13]  A. Weigend,et al.  Time Series Prediction: Forecasting the Future and Understanding the Past , 1994 .

[14]  H. John,et al.  Why future supercomputing requires optics , 2010 .

[15]  Razvan Pascanu,et al.  How to Construct Deep Recurrent Neural Networks , 2013, ICLR.

[16]  Chen Sun,et al.  DSENT - A Tool Connecting Emerging Photonics with Electronics for Opto-Electronic Networks-on-Chip Modeling , 2012, 2012 IEEE/ACM Sixth International Symposium on Networks-on-Chip.

[17]  Serge Massar,et al.  Fully analogue photonic reservoir computer , 2016, Scientific Reports.