Jean-Marc Fellous, Ph.D.
Associate Professor, Psychology
Research Scientist, ARL Division of Neural Systems, Memory and Aging
Faculty Affiliate, Evelyn F. McKnight Brain Institute

Dr. Fellous' research focuses on the way large networks of neurons interact in the face of background noise and unreliable synaptic transmission.

His background is in Computer Science and Artificial Intelligence. His previous work included automated face recognition using image processing techniques based on the biological features of neurons of the early stages of visual processing. This early work was followed up by a series of studies on the nature of facial information such as sex, age or emotional expressions.

Higher cognitive functions such as face perception rely on the computations of large networks of neurons spanning many areas of the nervous system. Interestingly, in some cases, these large network computations can be understood at the level of single cells. For example, so called 'face cells' in the temporal lobe, dozen of synapses away from the eye, will fire only when the picture of a specific individual is presented. As of today, such a selectivity no matter what the details of the inputs are (e.g. face orientation, size, makeup, facial hair) cannot be achieved by any known artificial system, computerized or otherwise. Yet, it is achieved effortlessly by face cells in the human and monkey brains. A similar selectivity has been observed in 'place cells' in the hippocampus, a structure involved in short term memory. These cells are selective to a particular spatial location in a given environment, and are again dozens of synapses away from the basic sensory apparatus.

Dr. Fellous current research focuses on how large networks of neurons transfer and process information in the face of large amounts of neuronal and synaptic noise to yield such a reliable output. Members of his laboratory use a combination of experimental in vitro and in vivo techniques together with sophisticated computer simulations to understand the basic neural processing principles that are required to yield such effective and selective computations.

• In vitro, the emphasis is on the reliability of spike timing. How 'trustworthy' is the action potential emitted by a single neuron? How does this reliability change in the presence of neuromodulators?
• In vivo, the emphasis is on the detection and analyses of precise spike patterns within a population of simultaneously recorded neurons. Recordings are conducted from 'place cells' and from neuromodulatory centers while rats are performing various rewarded spatial tasks.
• Computational studies integrate the data and insights obtained in vitro and in vivo. Biophysical computational models of large networks of neurons with noise and stochastic synapses are built and studied. What are the key parameters for reliable firing? What physiologically plausible or pathological conditions are conducive to changes in reliability?

Other research interests of the laboratory include the neural basis of emotion, memory reconsolidation, and the computational roles of neuromodulation in the young and aged.

Dr. Fellous teaches courses in Neural Data Analyses, Computational Neuroscience, and Physiological Psychology. He supervises and advises undergraduate, graduate and postdoctoral students in Psychology, Applied Mathematics, Neuroscience and Physiological Science