Shadmehr Laboratory
Laboratory for Computational Motor Control · Johns Hopkins University
Ali Ghazizadeh

Ali Ghazizadeh

MS (2004)

Research

Ali joined the lab as a student in the BME MS program in 2003. Ali demonstrated that by using theory of optimal state estimation, it was possible to account for generalization patterns that take place during motor learning. He was also instrumental in devising the two-state model of motor learning. The idea in this model of learning is that when the brain experiences a prediction error, there are two distinct processes that learn from that error: one process that has strong sensitivity to error, learning significantly, but suffering from poor retention. And another process that also learns from that error, but more slowly, having low sensitivity, but good retention. This two-state model of learning is now the basis of many ideas and experiments in motor control. After completing his MS in BME, he enrolled as a PhD student at UC Berkeley, and then he was a postdoctoral fellow with Okihide Hikosaka at the National Institutes of Health. He subsequently was appointed as Assistant Professor at Sharif University of Technology.

Ideas developed by Ali

  • Developing state-space models of motor memory. Ali's work helped develop computational models that describe motor learning as the trial-by-trial updating of an internal state. This framework provided a quantitative way to infer otherwise hidden processes of learning and memory from changes in behavior.
  • Discovering multiple timescales in motor learning. Ali contributed to the development of the two-state model of motor adaptation, in which behavior reflects the interaction of a fast learning process that also forgets rapidly and a slower process that learns gradually but retains its memory. This framework helped explain important features of motor learning, including rapid adaptation, savings, and spontaneous recovery.
  • Showing that generalization depends on the history of prior actions. Ali's work helped demonstrate that what is learned from an error does not generalize simply according to the current movement. Instead, generalization depends on the movements that preceded the error, showing that the representation underlying motor learning is shaped by the recent history of actions.
  • Selected publications

    Smith MA, Ghazizadeh A, Shadmehr R. Interacting adaptive processes with different timescales underlie short-term motor learning. PLoS Biology. 2006;4:e179.

    Krakauer JW, Mazzoni P, Ghazizadeh A, Ravindran R, Shadmehr R. Generalization of motor learning depends on history of prior action. PLoS Biology. 2006;4:e316.

    Ghazizadeh A. State-space models of online acquisition in motor memory. MS thesis, Johns Hopkins University. 2005.

    Gallery

    ← Return to lab members