Generalized and Invariant Single-Neuron In-Vivo Activity Representation Learning

In neuroscience, models that learn representations of single-neuron in-vivo activity are essential for understanding the functional identities of individual neurons. The primary goal of these models—spanning Transformer-based, contrastive, and variational autoencoder frameworks, is not to predict neural activity, but to distill it into a stable, low-dimensional embedding that captures a neuron’s intrinsic features. These learned identity embeddings should be invariant to changing experimental conditions while reflecting the neuron’s molecular type and anatomical location, thus enabling downstream tasks like in-vivo cell type prediction. However, current models suffer from limited generalizability due to batch effects: non-biological variations arising from differences in experimental design, animal subjects, or recording platforms. These batch effects cause overfitting, reducing model robustness and utility. Crucially, previous work has not rigorously evaluated model performance on unseen, or "out-of-domain," animals and stimuli, creating a significant gap in the field. To solve this, we first introduce a comprehensive benchmark protocol that explicitly evaluates generalization to unseen batches. Second, we propose a model-agnostic adversarial training strategy where a discriminator network forces the primary model to learn embeddings that are invariant to batch information. Our approach is compatible with all major single-neuron representation models and significantly improves their robustness. This work highlights the critical need for generalization in such models and offers an effective solution, paving the way for the creation of unified neural atlases from in-vivo activity.