February 18, 2025
Functionally Invariant Path Algorithm: Empowering Neural Networks to Retain Knowledge and Adapt to New Data

Functionally Invariant Path Algorithm: Empowering Neural Networks to Retain Knowledge and Adapt to New Data

Neural networks have demonstrated impressive capabilities in learning specific tasks, such as recognizing handwritten digits. However, these models often face the challenge of catastrophic forgetting when presented with new tasks. Neural networks, including those used in self-driving cars, must be completely reprogrammed to learn new assignments, leading to the loss of previously acquired knowledge. In contrast, biological brains, such as those of humans and animals, exhibit remarkable flexibility in learning new skills without forgetting the old.

Caltech researchers, inspired by the adaptability of biological brains, have developed a novel algorithm called Functionally Invariant Path (FIP) to enable Neural Networks  to learn from new data without losing their existing knowledge. This groundbreaking research, led by Assistant Professor of Computational Biology Matt Thomson and former graduate student Guru Raghavan, has the potential to revolutionize various applications, from enhancing online shopping recommendations to fine-tuning self-driving cars.

The FIP algorithm was developed in the Thomson lab, in collaboration with the lab of Carlos Lois, Research Professor of Biology, who studies how birds can rewire their brains to learn to sing again after a brain injury. Humans also possess this ability, as those who have experienced brain damage from a stroke can often regain everyday functions.
Thomson and Raghavan’s research journey began with understanding the fundamental science of how brains learn flexibly. “We asked ourselves, how do we give this capability to artificial neural networks?” says Thomson. The team employed a mathematical technique called differential geometry to create the FIP algorithm, which allows a neural network to be modified while preserving previously encoded information.

In 2022, with the guidance of Julie Schoenfeld, Caltech Entrepreneur In Residence, Raghavan and Thomson founded Yurts to further develop the FIP algorithm and deploy machine learning systems at scale to address a wide range of problems. Raghavan, along with industry professionals Ben Van Roo and Jason Schnitzer, co-founded Yurts.

Raghavan is the study’s first author. In addition to Raghavan and Thomson, Caltech co-authors include graduate students Surya Narayanan Hari and Shichen Rex Liu, and collaborator Dhruvil Satani. Bahey Tharwat of Alexandria University in Egypt is also a co-author. Thomson is an affiliated faculty member with the Tianqiao and Chrissy Chen Institute for Neuroscience at Caltech.

*Note:
1. Source: Coherent Market Insights, Public sources, Desk research
2. We have leveraged AI tools to mine information and compile it

About Author - Ravina Pandya

Ravina Pandya,  a content writer, has a strong foothold in the market research industry. She specializes in writing well-researched articles from different industries, including food and beverages, information and technology, healthcare, chemicals and materials, etc. With an MBA in E-commerce, she has expertise in SEO-optimized content that resonates with industry professionals.  LinkedIn Profile

View all posts by About Author - Ravina Pandya →