Gregory P. Amis, Ph.D.

Machine learning expert and algorithm R&D team lead, with doctoral training in autonomous learning systems and production software engineering experience. My interests are in solving real-world problems in predictive analysis while advancing the state of the art.

• Resume (pdf)

  Requires Adobe Acrobat Reader.

  

• Main dissertation project

  I developed a neural network for pattern classification that learns from multiple data sources with disparate input dimensionalities, creating a unified model of the underlying class patterns. The system learns on-line, does not require parameter adjustment for robust performance, and is resilient to changing deployment contexts. In short, it is a fully autonomous learning system.

  This work was done in collaboration with my PhD adviser, Gail A. Carpenter, in the Technology Lab at Boston University's Department of Cognitive & Neural Systems. A first generation of this model has been accepted for publication to the journal Neural Networks.

  • Tech report (pdf)

    Amis, G. & Carpenter, G. A. (2009). Self-Supervised ARTMAP. Neural Networks, doi:10.1016/j.neunet. 2009.07.026. (In press.)

  • Demo

    Animations illustrating the self-supervised learning paradigm and self-supervised ARTMAP performance. (Requires Adobe Flash plug-in.)

  • Code

    Zip archive of Java implementation with MATLAB scripts.

  • Data

    Zip archive of a MATLAB .mat file for Boston remote sensing benchmark.

• Other research projects
  • Default ARTMAP 2

    A successor to default ARTMAP (Carpenter, 2003) for robust pattern learning and classification across problem domains.

    • Publication (pdf)

      Amis, G. & Carpenter, G. A. (2007). Default ARTMAP 2. Proceedings of the International Joint Conference on Neural Networks (IJCNN'07), Orlando, FL, 777-782.

    • Demo

      PowerPoint show illustrating how default ARTMAP 2 differs from default ARTMAP. (Requires Microsoft PowerPoint 2000 or newer, or the free viewer.)

    • Code

      Zip archive of MATLAB implementation, including test cases, examples, and demos.

• Writing samples
  • Distributed ARTMAP

    A summary of the distributed ARTMAP model, from
    Amis, G. P., Carpenter, G. A., Ersoy, B., & Grossberg, S. (2009). Cortical learning of recognition categories: a resolution of the exemplar vs. prototype debate. Technical Report, CAS/CNS TR-2009-002, Boston, MA: Boston University.

  • "Traveling Gaussian Waves" Simulation Assignment

    One of the myriad simulation assignments I completed in my first three years in the PhD program at CNS. It is a mathematical and computational analysis of one theory of how the human visual cortex might process motion, assigned by
    Prof. Ennio Mingolla in his introductory class on neural models of vision.