Frank Leone
From SensorimotorLab.com
Frank Leone is a graduate student, working on the topic of perception-action conversion using multivariate fMRI and computational modeling techniques. His main research interest boils down to the question "How can we explain complex human sensorimotor behavior in simple emergent, bottom-up processes?". Current projects include a computational model of spatial updating across complex movements, and fMRI studies on saccade cortical maps, gaze, hand, and foot movements, and grasping in different modalities. More information can be found below.
Research interests
I am interested in questions at the intersection of Cognitive Neuroscience, Philosophy and Artificial Intelligence. The most important ones:
How does our cognitive system efficiently convert perception to action? More specifically, how does it do so without any reference to higher order concepts like goals and actions, but purely as a distributed deterministic machine? This is particularly clear in my research on saccade reference frames (computer modeling), saccade maps, hand, foot, and eye movement, as well as grasping movements (all ongoing)
What are the Foundations of Neuroscience? What are the assumptions we all willingly, but not necessarily knowingly, make, and what are their implications? Two important points of critique from my side:
- A lot of research done in Neuroscience is too anthropomorphic, assuming the brains "thinks" like we humans think, which is far from a necessity. Ideally, research should be more bottom-up, letting the data define the categories. First steps in this direction are taken in articles like Kriegeskorte, Mur and Bandettini (2008) and Aflalo and Graziano (2006).
- Our brain is a statistical machine, why then does Neuroscience does not take into account world statistics? How much of our interesting/puzzling/contradicting results might not be explainable by checking for statistical relations of a stimulus with other concepts/stimuli, of a stimulus with an action, or of an action with another action? I don't think it is possible to define man-made categories and apply them directly to the brain, rather, the statistical properties of a stimulus should be used. For example, the Haxby et al (2001) results, finding information on different categories in a large part of the brain, might be well explainable using statistical relations among these stimuli.
These interests are particularly clear in two talks I have given a few times within the Donders Institute: "Towards an Holistic Approach to Neuroscience" and "Distributed Ontology, Localized Activity?".
How much information can be decoded from brain activation and what does this tell us? Is there more to understanding the brain than extracting information; is labeling and anthropomorphizing a necessity? This is especially a critical reflection on my own research; what does it tell us that I can find information on all kinds of things in a range of cortical locations? How does this relate to univariate analyses? These questions led me to the "Holistic Approach to Neuroscience".
How can we explain the brain in an emergent, bottom-up way? How can we develop an artificial brain or model with minimal assumptions, which can still explain higher order concepts? This is the question that led to the "Distributed Ontology, Localized Activity" story and is especially found in my computational modeling work.
Publications
- Medendorp W.P., Buchholz V.N., Van Der Werf J., Leone F.T.M.
- Parietofrontal circuits in goal-oriented behavior
- European Journal of Neuroscience", 33:2017-27, 2011
- Online version