MUMBAI, India, Jan. 23 -- Intellectual Property India has published a patent application (202641001645 A) filed by Muthayammal Engineering College, Namakkal, Tamil Nadu, on Jan. 7, for 'apparatus for self-assembling modular neural architectures for lifelong continual learning.'
Inventor(s) include Dr. G. Kavitha; Ms. A. Mehala; Mr. Pudi Praveen Kumar; Mr. Puramsetti Leelaram Kumar; Mr. Pusukuri Mohana Venkata Kishor; and Mr. Rajala Veera Divya Teja.
The application for the patent was published on Jan. 23, under issue no. 04/2026.
According to the abstract released by the Intellectual Property India: "The invention presents a solution and a system of mitigating catastrophic memory in autonomous AI systems. The device features an expansionable modular memory of dynamically expanding memory, where neural primitives are stored. These modules are selectively activated and combined into the optimal architecture of the task, when activated by a task-shift detection mechanism, to a routing meta-controller. The device can fix the knowledge acquired on the previous tasks through a gradient masking unit to allow the system to amplify its ability of acquiring new knowledge in architecture. This self-assembly design enables transfer of knowledge forward and provides to the non-stationary environment with computational efficiency hence it is suitable to edge computing and long-term autonomous robotics."
Disclaimer: Curated by HT Syndication.
