Embedded Deep Learning : Challenges and Recent Advancements
Deep learning and convolutional neural nets have become state of the art techniques for solving many computer vision problems. The compute intensiveness and large memory requirement of these algorithms make it very challenging to realise them on low power embedded platforms. This talk focuses on these challenges, ongoing research on network compression techniques, sparse and binary networks. It spreads some light on our ongoing research in energy efficient acceleration of deep learning algorithms on heterogeneous embedded platforms including FPGA, GPU and NoC based systems. Also, it will touch upon our research in developing FPGA accelerators for binary neural nets.
Gopalakrishna Hegde is a research associate in the school of computer science and engineering at Nanyang Technological University, Singapore. He completed his Masters degree in Embedded Systems from Nanyang Technological University in 2015. He obtained bachelor degree in electronics and communication engineering and worked for 2 years as firmware development engineer in India. He is actively working on feasibility analysis and acceleration of deep learning algorithms on low power embedded platforms. His research interest include machine learning, deep learning, heterogeneous embedded architectures and parallel programming.