Dice Question Streamline Icon: https://streamlinehq.com

Relating probabilistic models to neural activity and mapping models

Identify how probabilities are represented in neural activity and develop an appropriate mapping model for connecting probabilistic inference models to neural data when the inference model is itself probabilistic.

Information Square Streamline Icon: https://streamlinehq.com

Background

The paper motivates both discriminative and generative perspectives in probabilistic terms and emphasizes the need to relate these probabilistic constructs to neural measurements via mapping models.

Two specific unresolved issues are highlighted: the neural code for probabilities and the correct mapping-model framework when inference models output probability distributions.

References

There are at least two unresolved questions in relating probabilistic models to neural activity: First, how do probabilities relate to neural activity? Second, what is the right mapping model when the model of inference is probabilistic?

— How does the primate brain combine generative and discriminative computations in vision? (2401.06005 - Peters et al., 11 Jan 2024) in Box 5: Selected questions and challenges