Development of a RISC-V based Platform for efficient Temporal Convolutional Network Inference in embedded Systems
- Subject:Low-Power Systeme
- Type:Masterarbeit
- Date:ab 05 / 2023
- Tutor:
Development of a RISC-V based Platform for efficient Temporal Convolutional Network Inference in embedded Systems
Context
The use of Internet-of-Things (IoT) devices is continuously increasing. At the same time, the complexity of the applications is also growing, which represents a major challenge for the hardware used. It not only has to process the data with the lowest possible latency, but at the same time it also has to act as energy-efficiently as possible to enable a long battery life. Consequently, it is necessary to match the system architecture very well to the application. In particular, the use of artificial intelligence (AI) in IoT devices requires efficient hardware/software co-design.
Tasks
- Familiarization with Temporal Convolutional Networks (TCN)
- State-of-the-Art Research
- Selection of a suitable baseline architecture
- Development of an efficient Hardware accelerator for TCNs
- Test and Evaluation of the accelerator performance
Requirement
- Knowledge in a hardware description language (e.g. VHDL or Verilog)
- Programming experience in C/C++
- Basic knowledge in ML and neural networks is an advantage