A Prospect of Achieving Artificial Neural Networks through FPGA

With the beginning of the 21st century, industries and researchers witnessed the end of the Moore’s Law. The stagnation of the size of transistor and other related limitations have compelled researchers to look for alternative approaches in order to maintain the momentum for design of computational intensive, high performing reliable systems focusing applications involving cloud computing, artificial intelligence, big data and IoT. The fuzzy Artificial Neural Network (ANN) based design approach have assisted a lot, in optimizing the systems manifold, for instance optimizing the system design targeting the pattern recognition problem in a large dataset. The capabilities and advantages associated with ANN have made it possible to address varied set of problems. Until now it was being implemented through software. Recent studies have showcased different limitations associated with software level ANN implementations. To address these issues hardware implementation of ANN based algorithms have been proposed and found more efficient vis-a-vis software based implementations. With the advances in technology, hardware available today is more efficient and capable. Devices based on GPUs, FPGAs and ASICs are used for the same, however FPGA shows more potentials as it is closer to wafer processing, providing options for high degree optimizations in power, speed and area; which seems to be perfect for today’s dynamic dataset requirements. This paper focuses on the use of FPGAs for ANN inspired hardware level implementations. Authors have made a comprehensive study of the amalgamations of the technologies. Authors then propound the implementation of processing blocks, activations functions, weight managements and applications on the FPGA augmenting with external data interface.

[1]  Joo-Young Kim,et al.  A Scalable High-Bandwidth Architecture for Lossless Compression on FPGAs , 2015, 2015 IEEE 23rd Annual International Symposium on Field-Programmable Custom Computing Machines.

[2]  Steven A. Guccione,et al.  Neural network implementation using reconfigurable architectures , 1994 .

[3]  Jason Cong,et al.  Optimizing FPGA-based Accelerator Design for Deep Convolutional Neural Networks , 2015, FPGA.

[4]  Yuki Kobayashi,et al.  A Convolutional Neural Network Fully Implemented on FPGA for Embedded Platforms , 2017, 2017 New Generation of CAS (NGCAS).

[5]  Seul Jung,et al.  Hardware Implementation of a Real-Time Neural Network Controller With a DSP and an FPGA for Nonlinear Systems , 2007, IEEE Transactions on Industrial Electronics.

[6]  W. E. Blanz,et al.  GANGLION-a fast field-programmable gate array implementation of a connectionist classifier , 1992 .

[7]  Jagath C. Rajapakse,et al.  Neural networks in FPGAs , 2002, Proceedings of the 9th International Conference on Neural Information Processing, 2002. ICONIP '02..

[8]  G.E. Moore,et al.  Cramming More Components Onto Integrated Circuits , 1998, Proceedings of the IEEE.

[9]  John P. Hayes,et al.  Survey of Stochastic Computing , 2013, TECS.

[10]  Hui Li,et al.  A Stochastic-Based FPGA Controller for an Induction Motor Drive With Integrated Neural Network Algorithms , 2008, IEEE Transactions on Industrial Electronics.

[11]  Hari Angepat,et al.  An FPGA-based In-Line Accelerator for Memcached , 2014, IEEE Computer Architecture Letters.

[12]  Marc Chaumont,et al.  Speeding-up a convolutional neural network by connecting an SVM network , 2016, 2016 IEEE International Conference on Image Processing (ICIP).

[13]  Brad Hutchings,et al.  RRANN: a hardware implementation of the backpropagation algorithm using reconfigurable FPGAs , 1994, Proceedings of 1994 IEEE International Conference on Neural Networks (ICNN'94).

[14]  Nouma Izeboudjen,et al.  Digital Implementation of Artificial Neural Networks: From VHDL Description to EPGA Implementation , 1999, IWANN.

[15]  James R. Larus,et al.  A reconfigurable fabric for accelerating large-scale datacenter services , 2014, 2014 ACM/IEEE 41st International Symposium on Computer Architecture (ISCA).