Resolving superposition in AI for interpretability and cross-modal alignment in patient-neuronal images
For interpretability and cross-modal analysis in biological AI, this work provides a method to resolve superposition and align imaging with transcriptomics, though it is incremental in applying known techniques (SAEs, GW optimal transport) to a new domain.
The paper addresses superposition in neural networks hindering interpretability and corrupting latent space geometry. Using sparse autoencoders on over 100,000 patient-derived neuron images, they recover geometric fidelity and align image representations with scRNA-seq data via Gromov-Wasserstein optimal transport, reconstructing pathology pathways without reference spatial transcriptomics.
Artificial intelligence is transforming our capability to solve biological challenges. In dimensionality bottleneck regimes exacerbated by high-dimensional biological data, Neural networks force distinct concepts into the lower dimensions known as superposition. Although this superposition is widely known to hinder interpretability, its impact on corrupting the geometry of latent spaces remains critically overlooked. Here, we utilized sparse autoencoders (SAEs) trained on over 100,000 multiplexed images of patient-derived Parkinson's disease and healthy neurons to resolve superposition. This approach bypasses the mathematical non-uniqueness of feature attribution by shifting to interpretable latent representation analysis. We theoretically and empirically demonstrate that superposition contaminates representational metric spaces, and thereby SAEs successfully recover geometric fidelity. By treating these geometrically purified representations as single-cell state vectors, we adapted single-cell RNA sequencing (scRNA-seq) data analysis methodologies directly to the image domain. Finally, we introduce GW-map, utilizing Gromov-Wasserstein optimal transport to align these image representations with authentic scRNA-seq data \emph{de novo}. This coupling reconstructs hierarchical neuronal pathology pathways such as Calcium-AIS scaffold, without reference spatial transcriptomics, establishing a scalable foundation for spatial biology. Code is available at https://github.com/jijihihi/Bio_superposition