Sayanta Roychowdhury
Papers
1
Total Citations
5
H-Index
1
About
Sayanta Roychowdhury is a researcher at the forefront of safety-critical artificial intelligence, specializing in the robustness and fault tolerance of deep neural networks. His work addresses a pressing challenge: ensuring that convolutional neural networks (CNNs) remain reliable when deployed in real-world, high-stakes environments like automated driving and human-robot interaction. Roychowdhury’s major contribution lies in developing rigorous methods to detect and mitigate hardware-induced errors, particularly soft errors that can corrupt model computations. In his highly cited work, "Towards a Safety Case for Hardware Fault Tolerance in Convolutional Neural Networks Using Activation Range Supervision" (2021), he introduced a novel activation range supervision technique that enables CNNs to self-monitor and maintain accuracy even when underlying hardware is compromised. This research is foundational for building verifiable safety cases in autonomous systems. With growing recognition for bridging the gap between AI performance and dependability, Roychowdhury’s work is shaping how engineers design trustworthy neural networks for the physical world—a critical step toward safe, widespread deployment of intelligent machines.
Research Focus
Key Achievements
Top Papers
- 1