Profile

Bartlett W. Mel

Associate Professor of Biomedical Engineering
Member, Neuroscience Graduate Program

Bartlett W. Mel

Research Topics

  • Using computer models to study brain function at single cell and systems levels.
  • Role of active dendritic processing in the sensory and memory-related functions of pyramidal neurons.
  • Neuromorphic models of visual cortex; neurally-inspired approaches to image processing problems.

Research Images

Dendrites of the world (plus one axon and one glial cell).  Dendrites receive thousands (even up to hundreds of thousands) of synaptic inputs from other cells.  What do they compute?  Our lab uses detailed biophysical modeling studies to find out.
Dendrites of the world (plus one axon and one glial cell). Dendrites receive thousands (even up to hundreds of thousands) of synaptic inputs from other cells. What do they compute? Our lab uses detailed biophysical modeling studies to find out.
A simplifying abstraction for an individual pyramidal cell.  Based on modeling and experimental studies from out lab and other labs (especially our collaborator Jackie Schiller at the Technion in Haifa, Israel), this is our current working model of the individual pyramidal cell.  Thin basal and oblique branches act as separately thresholded nonlinear subunits whose outputs are summed at the cell body.  The apical tuft
A simplifying abstraction for an individual pyramidal cell. Based on modeling and experimental studies from out lab and other labs (especially our collaborator Jackie Schiller at the Technion in Haifa, Israel), this is our current working model of the individual pyramidal cell. Thin basal and oblique branches act as separately thresholded nonlinear subunits whose outputs are summed at the cell body. The apical tuft "network" helps set a gain factor that modulates the output firing rate (this part is mostly derived from the recent in vitro studies of Matthew Larkum and his colleagues).
Neuromorphic model of the contour extraction network within visual cortex.  Inputs (bottom row) are local edge detectors, output units (top row) are contour units that emphasize the shape-defining boundaries of objects in complex scenes.  Contour units are influenced by feedforward local edge inputs, by feedforward divisive inhibition as well as inhibitory feedback from other (incompatible) contour units, and feedback excitation mediating long-range contour shape and cross-scale interactions.  Results of applying the contour network to complex scenes are shown in the next figure.
Neuromorphic model of the contour extraction network within visual cortex. Inputs (bottom row) are local edge detectors, output units (top row) are contour units that emphasize the shape-defining boundaries of objects in complex scenes. Contour units are influenced by feedforward local edge inputs, by feedforward divisive inhibition as well as inhibitory feedback from other (incompatible) contour units, and feedback excitation mediating long-range contour shape and cross-scale interactions. Results of applying the contour network to complex scenes are shown in the next figure.
Seeing the line drawing.  Results of applying a neuromorphic contour-extraction network, inspired by the organization of visual cortex, to complex natural scenes.  Local edges are extracted (middle column), and fed to the contour network.  Result emphasize shape-defining contours, i.e. the scene's line drawing.
Seeing the line drawing. Results of applying a neuromorphic contour-extraction network, inspired by the organization of visual cortex, to complex natural scenes. Local edges are extracted (middle column), and fed to the contour network. Result emphasize shape-defining contours, i.e. the scene's line drawing.

Research Overview

My research interests lie in the areas of Computational Neuroscience and Neural Engineering. Most of the work in my lab involves the use of computer models to study brain function. Some of our goals are of a primarily scientific nature. For example, we use detailed biophysiical modeling studies to study synaptic integration in active dendritic trees, and explore how dendritic trees could contribute to the sensory and memory-related functions of the brain. To do this work, we use simulation packages such as NEURON and a variety of custom software developed by members of the lab.

Some of our work combines scientific and engineering goals. For example, we are interested in the massively parallel computations carried out in the visual cortex which allow us to recognize objects with a speed, accuracy, and robustness that are far beyond the technical state of the art. How does this amazing neural technology work? We have developed a number of models of this process, and have applied them to various types of visual recognition problems. In one of our ongoing projects, we are attempting to understand the mechanisms used by the brain to learn which features are best for recognizing objects and scenes. Our hope is to someday be able to construct high performance artificial vision systems which could be used to power intelligent machines.

Contact Information

Mailing Address University of Southern California
Hedco Neursocience Bldg, Mail Code 2520
Los Angeles, CA 90089
Office Location 103 Hedco Neuroscience, UPC
Office Phone (213) 740-0334
Lab Location B1 Hedco Neuroscience, UPC
Lab Phone (213) 740-3397
Fax (213) 740-1470
Office Location 103 Hedco Neuroscience, UPC

Websites

Education

  • B.S. in EECS, University of California, Berkeley, 1982.
  • Ph.D. in Computer Science, Univ. of Illinois, Urbana-Champaign, 1989.
  • Postdoctoral Fellow, California Institute of Technology, 1989-1994.

Selected Publications

View a complete PubMed searchView a complete Google Scholar search
  • Jain, R., Millin R., & Mel, B.W. (2015) Multimap formation in visual cortex.   J. Vision, 15(16):3. doi: 10.1167/15.16.3.

    PubMed Link
  • Behabadi, B.F. & Mel, B.W. (2014) Mechanisms underlying subunit independence in pyramidal neuron dendrites.  PNAS, U S A, 111(1):498-503. PubMed Link
  • Jadi, M., Behabadi, B.F., Poleg-Polsky, A., Schiller J. & Mel, B.W. (2014).  An augmented 2-layer model captures nonlinear analog spatial integration effects in pyramidal neuron dendrites. Proc. of the IEEE, (Special issue on Computational Neuroscience), 102(5):782-798 Link
  • Ramachandra, C. and Mel, B.W. (2013) Computing local edge probability in natural scenes from a population of oriented simple cells.  Journal of Vision, 13(14). PubMed Link
  • Behabadi, B.F., Polsky, A., Jadi, M., Schiller, J. & Mel, B.W. (2012). Location-dependent excitatory synaptic interactions in pyramidal neuron dendrites. PLoS Comput Biol 8: e1002599. doi:10.1371/journal.pcbi.1002599. PubMed Link
  • Wu, X. & Mel, B.W. (2009). Capacity enhancing synaptic learning rules in a medial temporal lobe online learning model. Neuron 62: 31-41. PubMed Link
  • Zhou, C. & Mel, B.W (2008).  Optimal cue combination and color edge detection in natural scenes. Journal of Vision, 8(4):4, 1-25, http://journalofvision.org/8/4/4/, doi:10.1167/8.4.4. PubMed Link
  • Chklovskii, D.B., Mel, B.W. & Svoboda, K. (2004) Cortical rewiring and information storage.  Nature, 7:782-788. PubMed Link
  • Polsky, A., Mel, B.W. & Schiller, J. (2004) Computational subunits in thin dendrites of pyramidal cells.  Nature Neurosci. 7: 621-627. PubMed Link
  • Poirazi, P., Brannon, T. & Mel, B.W. (2003) Pyramidal neuron as two-layer neural network.  Neuron 37: 989-999. PubMed Link
  • Poirazi, P. & Mel, B.W. (2001) Impact of active dendrites and structural plasticity on the storage capacity of neural tissue.  Neuron 29: 779-796. PubMed Link