When constructing AI, is easier higher? New analysis challenges assumptions

January 6, 2025

Editors' notes

This text has been reviewed based on Science X's editorial course of and insurance policies. Editors have highlighted the next attributes whereas making certain the content material's credibility:

fact-checked

peer-reviewed publication

trusted supply

proofread

When constructing AI, is easier higher? New analysis challenges assumptions

When making an attempt to resolve issues, synthetic intelligence usually makes use of neural networks to course of information and make choices in a manner that mimics the human mind.

In his newest analysis, Binghamton College Assistant Professor Sadamori Kojaku challenges a elementary assumption in AI circles—that extra advanced neural networks are at all times higher.

The paper, printed in Nature Communications, exhibits that straightforward neural networks can discover communities in advanced networks with theoretical optimality, questioning the frequent view that extra advanced fashions outperform easier ones.

"What we discovered was that the coaching issues, not the programming structure itself," stated Kojaku, who joined the college of the Thomas J. Watson School of Engineering and Utilized Science's College of Programs Science and Industrial Engineering in Fall 2023.

"There are lots of methods to show a neural community, however we discovered that one finest instructing methodology is contrastive studying, the place you current actual information and pretend information so the neural community is skilled to distinguish the 2. This easy coaching achieves optimum efficiency."

Understanding how AIs work is key to establishing belief when it makes choices in vital areas comparable to well being care or electrical grids.

Proper now, the precise route that AIs use to get their conclusions is inside what programmers name a "black field." Information enter results in a end result, however the pathway between these factors could be mysterious.

"Our work unboxes the neural networks after which tries to interpret the way it works to offer a assure that this neural community works optimally for this particular job," Kojaku stated. "That is our first work that tries to hammer on the black field."

Additionally contributing to the paper are Professors Filippo Radicchi, Yong-Yeol Ahn and Santo Fortunato from Indiana College, the place Kojaku served as a postdoctoral fellow after incomes his Ph.D. at Hokkaido College in Japan and earlier than coming to Binghamton.

Extra data: Sadamori Kojaku et al, Community neighborhood detection by way of neural embeddings, Nature Communications (2024). DOI: 10.1038/s41467-024-52355-w

Journal data: Nature Communications Offered by Binghamton College Quotation: When constructing AI, is easier higher? New analysis challenges assumptions (2025, January 6) retrieved 6 January 2025 from https://techxplore.com/information/2025-01-ai-simpler-assumptions.html This doc is topic to copyright. Other than any honest dealing for the aim of personal research or analysis, no half could also be reproduced with out the written permission. The content material is offered for data functions solely.

Discover additional

New methodology for evaluating neural networks exposes how synthetic intelligence works 0 shares