Abstract
It is difficult to achieve real-time performance for a deep neural networks (DNN) in low-profile systems such as edge computing or Internet of Things (IoT) due to the large amount of computations that are required. A commonly used activation function is ReLU (Rectified Linear Unit) due to its simplicity and good performance. A key characteristic of the ReLU function is that it produces sparse vectors within the neural network. In this paper, we propose an optimized DNN accelerator for sparse vector multiplications. In particular, we propose a squeezer unit to detect zeros and then skip feeding that data to the processing elements. In addition, we design a dynamic scheduler to efficiently allocate multiple neuron computations to the processing elements. With our architecture, we achieve reduced hardware resources compared to prior work due to a 62% reduction in the number of required index bits. We also achieve 8.6% better processing speed than for a system without the acceleration.
Original language | English (US) |
---|---|
Title of host publication | Proceedings - 2023 Congress in Computer Science, Computer Engineering, and Applied Computing, CSCE 2023 |
Publisher | Institute of Electrical and Electronics Engineers Inc. |
Pages | 887-892 |
Number of pages | 6 |
ISBN (Electronic) | 9798350327595 |
DOIs | |
State | Published - 2023 |
Event | 2023 Congress in Computer Science, Computer Engineering, and Applied Computing, CSCE 2023 - Las Vegas, United States Duration: Jul 24 2023 → Jul 27 2023 |
Publication series
Name | Proceedings - 2023 Congress in Computer Science, Computer Engineering, and Applied Computing, CSCE 2023 |
---|
Conference
Conference | 2023 Congress in Computer Science, Computer Engineering, and Applied Computing, CSCE 2023 |
---|---|
Country/Territory | United States |
City | Las Vegas |
Period | 7/24/23 → 7/27/23 |
Bibliographical note
Publisher Copyright:© 2023 IEEE.
Keywords
- Deep neural networks
- DNN accelerator
- IoT
- low-cost
- sparse data multiplication