[2506.03088] Modelling the Effects of Hearing Loss on Neural Coding in the Auditory Midbrain with Variational Conditioning

View PDF of the article “Modelling the Effects of Hearing Loss on Neural coding in the Auditory Midbrain with Contrasting Conditioning,” by Lloyd Bellat and co-authors

View PDF HTML (beta)

a summary:The mapping from sound to the neural activity that underlies hearing is highly nonlinear. The first few stages of this cochlear mapping have been modeled successfully, with hand-built biophysical models and, more recently, with DNN models trained on datasets simulated by biophysical models. Modeling the auditory brain has been challenging because central auditory processing is too complex to build models manually, and datasets to train DNN models directly were not available. Recent work has taken advantage of large-scale, high-resolution neural recordings from the auditory midbrain to build a DNN model of natural hearing with great success. But this model assumes that auditory processing is the same in all brains, and therefore cannot capture the diverse effects of hearing loss.

We propose a novel contrastive conditional model for learning the encoding of hearing loss space directly from recordings of neural activity in the auditory midbrain of intact and noise-exposed animals. Limiting hearing loss to only 6 free parameters per animal, our model accurately predicts 62% of the explainable variance in neural responses from normal-hearing animals and 68% from hearing-impaired animals, within a few percentage points of state-of-the-art animal-specific models. We demonstrate that the model can be used to simulate realistic activity from out-of-sample animals by fitting only the learned conditioning parameters with Bayesian optimization, achieving an entropy loss within 2% of the optimal in 15–30 iterations. Including more animals in the training data slightly improved performance on unseen animals. This model will enable the future development of hearing loss compensation models with parameters trained to directly restore normal neural coding in hearing-impaired brains, which can be quickly fitted to a new user by optimizing the human loop.