One of the fundamental goals of neuroscience is to understand how stimulus information is transformed into neuronal responses. To create a computational model that captures this transformation, it is necessary to collect neuronal responses across a diverse range of stimuli. However, due to the extensive range of possible stimuli, it is challenging to only select an optimal set of stimuli that can effectively elicit significant neural modulations. Moreover, there are practical limitations on the number of measurements and the duration of experiments that experimenters can perform at each time. As a result, experimenters rely heavily on their own prior knowledge to elicit the desired neural responses. However, this approach may not always result in an optimal experiment tailored to address their specific research questions.

To tackle this problem, I am currently developing a novel software tool that generates an optimal set of stimuli, tailored to the specific hypotheses of each researcher. Specifically, I am leveraging recent advancements in generative deep neural networks, such as GANs and stable diffusion models, to generate stimuli in a manner directly optimized to the brain response itself.  The goal is to produce a set of stimuli that would maximize the performance of a brain model, which, in turn, would most effectively modulate underlying neural responses.