Revealing the core dimensions underlying representations in brains, behavior and AI
For researchers studying representations in neuroscience, psychology, and AI, SRF provides a general-purpose tool to extract interpretable dimensions from similarity data, overcoming limitations of existing methods.
The authors introduce Similarity-Based Representation Factorization (SRF), a method to recover interpretable low-dimensional embeddings from similarity matrices. Across simulations and multiple neural, behavioral, and AI datasets, SRF recovers dimensions that match task-specific models, predict behavioral properties, and improve hypothesis testing.
The study of representations is widespread across fields, including neuroscience, psychology, and artificial intelligence. While representations are often studied and compared through similarities between stimuli, current methods provide only limited access to the dimensions that shape these representations and are often limited in interpretability. To overcome these challenges, here we introduce Similarity-Based Representation Factorization (SRF), a general computational method for recovering low-dimensional, non-negative, interpretable embeddings from similarity matrices derived from measured data. Across simulations and many neural, behavioral, and computational datasets, SRF recovers interpretable dimensions from diverse forms of representational data, even for very sparsely sampled, incomplete data. The dimensions derived from these datasets match those obtained by task-specific models, predict independent behavioral properties, improve exploratory analysis, and offer higher power for confirmatory hypothesis testing than comparing similarity matrices. Together, these results establish SRF as a general-purpose method with broad applications for uncovering, understanding, and leveraging the dimensions underlying representations.