News
MIT Develops Search Engine for Optimal Design of Deep Neural Network Accelerators
MIT researchers have created a search engine called SecureLoop that can efficiently identify the best designs for deep neural network accelerators, while ensuring data security is preserved. Deep neural network accelerators are specialized hardware components used in computationally intensive machine-learning applications. They help move and process large amounts of data. The challenge lies in choosing the most suitable design, especially when cryptographic operations need to be added to protect the data from hackers.
SecureLoop considers how encryption and authentication measures will impact the performance and energy usage of the accelerator chip. With this tool, engineers can obtain the optimal design for an accelerator tailored to their specific neural network and machine-learning task. Unlike conventional scheduling techniques that don’t consider security, SecureLoop can improve performance while keeping data protected.
This search engine is particularly useful for demanding AI applications like autonomous driving or medical image classification. It allows users to enhance the speed and performance of these applications while ensuring the safety of sensitive user data.
SecureLoop was developed by lead author Kyungmi Lee, an electrical engineering and computer science graduate student, along with researchers Joel Emer, Mengjia Yan, and Anantha Chandrakasan. The research findings will be presented at the IEEE/ACM International Symposium on Microarchitecture.
The team behind SecureLoop discovered that adding cryptographic operations to accelerators can significantly distort the design space of energy-efficient accelerators. By addressing this issue, they aim to create an efficient solution that optimizes both speed and energy efficiency for deep neural network accelerators.
Source: MIT News
Source article: MIT News. “Engineers efficiently search for optimal deep neural network designs.” ScienceDaily. ScienceDaily, 2 November 2021.
The post MIT Develops Search Engine for Optimal Design of Deep Neural Network Accelerators appeared first on Fagen Wasanni Technologies.