An Optimal Scale for Edge Detection

Many problems in early vision are ill posed. Edge detection is a typical example. This paper applies regularization techniques to the problem of edge detection. We derive an optimal filter for edge detection with a size controlled by the regularization parameter $\\ lambda $ and compare it to the Gaussian filter. A formula relating the signal-to-noise ratio to the parameter $\\lambda $ is derived from regularization analysis for the case of small values of $\\lambda$. We also discuss the method of Generalized Cross Validation for obtaining the optimal filter scale. Finally, we use our framework to explain two perceptual phenomena: coarsely quantized images becoming recognizable by either blurring or adding noise.

Neural Network Exploration Using Optimal Experiment Design

We consider the question "How should one act when the only goal is to learn as much as possible?" Building on the theoretical results of Fedorov [1972] and MacKay [1992], we apply techniques from Optimal Experiment Design (OED) to guide the query/action selection of a neural network learner. We demonstrate that these techniques allow the learner to minimize its generalization error by exploring its domain efficiently and completely. We conclude that, while not a panacea, OED-based query/action has much to offer, especially in domains where its high computational costs can be tolerated.

Near-Optimal Distributed Failure Circumscription

Small failures should only disrupt a small part of a network. One way to do this is by marking the surrounding area as untrustworthy --- circumscribing the failure. This can be done with a distributed algorithm using hierarchical clustering and neighbor relations, and the resulting circumscription is near-optimal for convex failures.

Permutation Tests for Classification

We introduce and explore an approach to estimating statistical significance of classification accuracy, which is particularly useful in scientific applications of machine learning where high dimensionality of the data and the small number of training examples render most standard convergence bounds too loose to yield a meaningful guarantee of the generalization ability of the classifier. Instead, we estimate statistical significance of the observed classification accuracy, or the likelihood of observing such accuracy by chance due to spurious correlations of the high-dimensional data patterns with the class labels in the given training set. We adopt permutation testing, a non-parametric technique previously developed in classical statistics for hypothesis testing in the generative setting (i.e., comparing two probability distributions). We demonstrate the method on real examples from neuroimaging studies and DNA microarray analysis and suggest a theoretical analysis of the procedure that relates the asymptotic behavior of the test to the existing convergence bounds.

Optimal Approximations of the Frequency Moments

We give a one-pass, O~(m^{1-2/k})-space algorithm for estimating the k-th frequency moment of a data stream for any real k>2. Together with known lower bounds, this resolves the main problem left open by Alon, Matias, Szegedy, STOC'96. Our algorithm enables deletions as well as insertions of stream elements.

Generating Circuit Tests by Exploiting Designed Behavior

This thesis describes two programs for generating tests for digital circuits that exploit several kinds of expert knowledge not used by previous approaches. First, many test generation problems can be solved efficiently using operation relations, a novel representation of circuit behavior that connects internal component operations with directly executable circuit operations. Operation relations can be computed efficiently by searching traces of simulated circuit behavior. Second, experts write test programs rather than test vectors because programs are more readable and compact. Test programs can be constructed automatically by merging program fragments using expert-supplied goal-refinement rules and domain-independent planning techniques.