Project description
Accelerating deep neural networks: static random-access memory and in-memory computing
Deep neural networks (DNNs) are artificial neural networks with depth or multiple hidden layers between input and output. Resembling the human nervous system, DNNs can learn more complex patterns and solve more complex problems – but this typically comes at a higher energy cost. Hardware accelerators speed up AI applications relative to software running on conventional central processing units. With the support of the Marie Skłodowska-Curie Actions programme, the ACROBAT project will design and implement an improved energy-optimising DNN accelerator. It will include a special static random-access memory with in-memory computing, meaning the computations will occur entirely in computer memory. The project will employ data path optimisation techniques to achieve an optimal energy-accuracy balance.
Objective
Deep Neural Networks (DNNs) are the fundamental component in most artificial intelligence applications. With the increasing number of applications based on artificial intelligence, the performance and energy efficiency of architectures running these algorithms have become crucial, especially for battery-powered platforms. In this work, I propose an energy optimizing memory design framework with a special SRAM/in-memory-computing structure. It also utilizes datapath optimization techniques like quantization and pruning with a fine-level assignment. Compared to other hardware accelerator studies for DNN processing, in this work, I will show that this special memory design, together with the architectural datapath optimization techniques, will have a much better capability of finding the Pareto optimal point in the energy-accuracy trade-off and increase the profitability of the final design.
Programme(s)
- HORIZON.1.2 - Marie Skłodowska-Curie Actions (MSCA) Main Programme
Funding Scheme
HORIZON-TMA-MSCA-PF-GF - HORIZON TMA MSCA Postdoctoral Fellowships - Global FellowshipsCoordinator
06800 Bilkent Ankara
Türkiye