MIT researchers have created a periodic table that connects over 20 classical machine learning algorithms. The framework sheds light on how scientists can combine strategies from different methods to improve existing AI models or develop new ones.
The researchers used their framework, called information contrastive learning (I-Con), to combine elements of two different algorithms and create a new image-classification algorithm that outperformed current state-of-the-art approaches by 8 percent. The periodic table categorizes algorithms based on the approximate relationships they learn between data points.
The researchers identified a unifying equation that underlies many classical AI algorithms, which they used to reframe popular methods and arrange them into a table. This allows researchers to design new algorithms without rediscovering existing ideas. The table also predicts where new algorithms should exist but have not been discovered yet.
According to Shaden Alshammari, an MIT graduate student and lead author of the paper, “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.” The research was presented at the International Conference on Learning Representations.
The researchers stumbled upon their unifying equation while studying clustering and contrastive learning algorithms. They found that these two disparate algorithms could be reframed using the same underlying equation, leading to new insights and discoveries. I-Con includes a wide range of algorithms, from classification to deep learning.
The periodic table provides a toolkit for machine learning scientists to think outside the box and combine ideas in innovative ways. According to Yair Weiss, a professor at the Hebrew University of Jerusalem, “Papers that unify and connect existing algorithms are extremely rare. I-Con provides an excellent example of such a unifying approach.”
Source: https://news.mit.edu/2025/machine-learning-periodic-table-could-fuel-ai-discovery-0423