arXiv Endorsement Request — cs.LG (Machine Learning)

Hello everyone,

My name is Ahmad Shajahan , a 2nd-year B.Tech Computer Science student at APJ Abdul Kalam Technological University, Kerala, India.

I have recently completed an original research paper titled:

“Thermodynamic Neural Computation (TNC): Novel Free-Energy Minimisation with Cache-Aware Memory-Mapped Annealing for Energy-Efficient AI.”

The paper proposes a new computational framework where gradient descent is replaced by thermodynamic free-energy minimisation , introducing 14 novel equations aimed at improving energy efficiency in AI training.

Key idea: Using thermodynamic principles and cache-aware annealing mechanisms to achieve significantly more energy-efficient optimization compared to standard SGD.

The paper is publicly available on Zenodo: https://doi.org/10.5281/zenodo.19029046

ORCID: https://orcid.org/0009-0001-3287-3055

I am preparing to submit this work to arXiv (cs.LG – Machine Learning) , but as a first-time submitter I require an endorsement from an existing arXiv author.

If anyone eligible to endorse in cs.LG is willing to review or endorse the submission, I would be very grateful.

I would be happy to share the full manuscript or discuss the work.

Thank you very much for your time and support.

— Ahmad Shajahan