Research approach
My research in NeuroAI lies at the intersection of Artificial Intelligence (AI) and Computational Neuroscience. I develop biologically inspired AI models to uncover how neural circuits generate cognition and behavior, and how their disruption leads to behavioral deficits. My general approach follows four steps:
- Define tasks that capture, and where possible isolate, the behavioral aspect of cognition I aim to explain
- Develop biologically inspired neural network models trained to solve these tasks
- Validate the models at both the behavioral and neural levels against human or animal data
- Reverse-engineer the trained models to derive mechanistic insights that are difficult to obtain from data alone
Research Focus
I study the cognitive abilities we use to solve problems. The example below shows several of them in action.
Imagine being asked to calculate 15 plus 27 in your head. To solve this addition, you learned an algorithm in school: a way to plan your behavior as a sequence of actions. First, you focus your Attention on the units, 5 and 7. Then, using your Semantic Memory, you remember that 5 plus 7 equals 12. You then focus on the tens, 1 and 2, while keeping the carry and the intermediate result in Working Memory. Finally, you combine the intermediate results to obtain 42. The ability to follow these steps in the right order can be thought of as Algorithmic Memory.
Computational Models Zoo
I have developed models for each of the cognitive components described above. Here is a summary of these models.
