This is the lab page of David T Blake, PhD. The lab investigates questions concerning brain electrophysiology and behavior. The most recent focus has been on basal forebrain deep brain stimulation, its physiological and behavioral effects, and the potential to use such deep brain stimulation to treat neurological disorders. Honestly, I think it is going to cure Alzheimer's, and by that I mean its application in patients will prevent cognitive decline from altering their activities of daily living. Human trials of our methods are ongoing in some centers. If you would like to apply our methods to your human trials, contact me. |
Main questions for the lab |
How is learning expressed in areas of the cerebral cortex that are wired differently e.g., sensory vs association cortex, i and how is learning modulated by acetylcholine, dopamine, and noradrenaline? |
Weeks of application of deep brain stimulation can dramatically improve executive function even in fully functioning young adults. Can it do the same in Alzheimer's patients? |
What do we need to know about learning rules in sensory cortex to build machine learning models that perform human-equivalent sensory discrimination? |
What are the cell biological pathways involved in instantiating those learning rules? |
What are the machine learning concepts embodied by those learning rules? What are their limitations? Mathematically, how can we optimize them? And how can we combine them with other neural network concepts to build a biologically based machine learning network that performs human equivalent sensory discrimination? Or discrimination that is better than human equivalent? |
In 1996, I went to the lab of Michael Merzenich. There, with R. Christopher deCharms, we succeeded in developing one of the earliest high yield sensory cortex implants. The following experiments, done also with fabrizio strata, Anne Churchland, and Richard Kempter, were among the first to resample, with high spatial resolution, the same cortical locations throughout a period of learning. The discoveries that followed were unexpected. A robust nonselective response enhancement occurred within 24 hours of the time that an animal learned to associate a new sensory stimulus with reward. Over longer periods of time, the representations of stimuli associated with omission of reward, or task distractors, became selectively suppressed. Surprisingly, any tendency for the representations of targets to be enhanced, relative to the representations of control stimuli, were weak or absent. The following figure, taken from Blake et al 2002, shows the rapidity and strength of the nonselective response enhancement. Cortical responses double in 24 hours after learning. |