Skip to main content
eScholarship
Open Access Publications from the University of California

Scholarly Works (2 results)

  • Thesis
  • Peer Reviewed
UC Berkeley Electronic Theses and Dissertations (2016)

In recent years, the research community has discovered that deep neural networks (DNNs) and convolutional neural networks (CNNs) can yield higher accuracy than all previous solutions to a broad array of machine learning problems. To our knowledge, there is no single CNN/DNN architecture that solves all problems optimally. Instead, the “right” CNN/DNN architecture varies depending on the application at hand. CNN/DNNs comprise an enormous design space. Quantitatively, we find that a small region of the CNN design space contains 30 billion different CNN architectures.

In this dissertation, we develop a methodology that enables systematic exploration of the design space of CNNs. Our methodology is comprised of the following four themes.

1. Judiciously choosing benchmarks and metrics.

2. Rapidly training CNN models.

3. Defining and describing the CNN design space.

4. Exploring the design space of CNN architectures.

Taken together, these four themes comprise an effective methodology for discovering the “right” CNN architectures to meet the needs of practical applications.

    Cover page: Exploring the Design Space of Deep Convolutional Neural Networks at Large Scale
    • Article
    • Peer Reviewed
    Building Efficiency and Sustainability in the Tropics (SinBerBEST) (2014)
    Aircraft Electric Power Systems (EPS) route power from generators to vital avionic loads by configuring a set of electronic control switches denoted as contactors. In this paper, we address the problem of designing a hierarchical optimal control strategy for the EPS contactors in the presence of system faults. We first formalize the system connectivity, safety and performance requirements in terms of mathematical constraints. We then show that the EPS control problem can be formulated as a Mixed-Integer Linear Program (MILP) and efficiently solved to yield load shedding, source allocation, contactor switching and battery charging policies, while optimizing a number of performance metrics, such as the number of used generators and shed loads. This solution is then integrated into a hierarchical control scheme consisting of two layers of controllers. The high-level controller provides control optimality by solving the MILP within a receding horizon approach. The low-level controller handles system faults, by directly actuating the EPS contactors, and implements the solution from the high-level controller only if it is safe. Simulation results confirm the effectiveness of the proposed approach.
      Cover page: Optimal Load Management System for Aircraft Electric Power Distribution
      Top