New MIT Framework Unifies Machine Learning Algorithms for Future Discovery
CAMBRIDGE, MA, 23RD APRIL , 2025 – There is a periodic table of machine learning created by MIT researchers that connects over 20 classical algorithms deeply to each other, thus paving the way for future discoveries of AI models.
Inspired by the chemical periodic table, this framework describes algorithms categorically based on the types of relationships they approximate within datasets.
According to research led by graduate student Shaden Alshammari, the contribution of this work is to give a unified mathematical equation behind many algorithms of machine learning, from classification to deep learning.
This is a part of a larger framework known as Information Contrastive Learning (I-Con), as it allows scientists to assume that existing methodologies can be reformulated and/or combined, promoting intelligent algorithm design and innovation.
“It’s not just a metaphor,“We’re starting to see machine learning as a system with structure that is a space we can explore rather than just guess our way through.” says Alshammari.
“The work went gradually, but once we had identified the general structure of this equation, it was easier to add more methods to our framework,”
Alshammari is joined on the paper by Axel Feldmann, an MIT graduate student; John Hershey, a researcher at Google AI Perception; William Freeman, the Thomas and Gerd Perkins Professor of Electrical Engineering and Computer Science and a member of the Computer Science and Artificial Intelligence Laboratory (CSAIL); as well as senior author Mark Hamilton, an MIT graduate student and senior engineering manager at Microsoft.
“We almost got to this unifying equation by accident. Once Shaden discovered that it connects two methods, we just started dreaming up new methods to bring into this framework. Almost every single one we tried could be added in,” Hamilton says.
“We’ve shown that just one very elegant equation, rooted in the science of information, gives you rich algorithms spanning 100 years of research in machine learning. This opens up many new avenues for discovery”
By putting together clustering and contrastive learning through I-Con, the team designed a novel image-classification method that surpassed current best models by 8%.
In addition, they also presented evidence that a debiasing technique derived from contrastive learning raises the accuracy of clustering, thus exhibiting the pragmatic strength of the framework.
This new AI periodic table, just like the chemical periodic table, can speculate on the existence of undiscovered elements-declaring algorithmic gaps.
With the expansion or addition of rows and columns, I-Con illustrates a road map for the research community to the open and unexplored miles of the ML terrain.
Supported through NSF, the Air Force AI Accelerator, and Quanta Computer, the research will be presented at the International Conference on Learning Representations.
AI Just Took Over Your Payments—Thanks to Payarc and AWS
