Learning Object-Independent Modes of Variation with Feature Flow Fields

We present a unifying framework in which "object-independent" modes of variation are learned from continuous-time data such as video sequences. These modes of variation can be used as "generators" to produce a manifold of images of a new object from a single example of that object. We develop the framework in the context of a well-known example: analyzing the modes of spatial deformations of a scene under camera movement. Our method learns a close approximation to the standard affine deformations that are expected from the geometry of the situation, and does so in a completely unsupervised (i.e. ignorant of the geometry of the situation) fashion. We stress that it is learning a "parameterization", not just the parameter values, of the data. We then demonstrate how we have used the same framework to derive a novel data-driven model of joint color change in images due to common lighting variations. The model is superior to previous models of color change in describing non-linear color changes due to lighting.

(Semi-)Predictive Discretization During Model Selection

In this paper, we present an approach to discretizing multivariate continuous data while learning the structure of a graphical model. We derive the joint scoring function from the principle of predictive accuracy, which inherently ensures the optimal trade-off between goodness of fit and model complexity (including the number of discretization levels). Using the so-called finest grid implied by the data, our scoring function depends only on the number of data points in the various discretization levels. Not only can it be computed efficiently, but it is also independent of the metric used in the continuous space. Our experiments with gene expression data show that discretization plays a crucial role regarding the resulting network structure.

On the Behavior of the Homogeneous Self-Dual Model for Conic Convex Optimization

There is a natural norm associated with a starting point of the homogeneous self-dual (HSD) embedding model for conic convex optimization. In this norm two measures of the HSD model’s behavior are precisely controlled independent of the problem instance: (i) the sizes of ε-optimal solutions, and (ii) the maximum distance of ε-optimal solutions to the boundary of the cone of the HSD variables. This norm is also useful in developing a stopping-rule theory for HSD-based interior-point methods such as SeDuMi. Under mild assumptions, we show that a standard stopping rule implicitly involves the sum of the sizes of the ε-optimal primal and dual solutions, as well as the size of the initial primal and dual infeasibility residuals. This theory suggests possible criteria for developing starting points for the homogeneous self-dual model that might improve the resulting solution time in practice

ACTORS: A Model of Concurrent Computation in Distributed Systems

A foundational model of concurrency is developed in this thesis. We examine issues in the design of parallel systems and show why the actor model is suitable for exploiting large-scale parallelism. Concurrency in actors is constrained only by the availability of hardware resources and by the logical dependence inherent in the computation. Unlike dataflow and functional programming, however, actors are dynamically reconfigurable and can model shared resources with changing local state. Concurrency is spawned in actors using asynchronous message-passing, pipelining, and the dynamic creation of actors. This thesis deals with some central issues in distributed computing. Specifically, problems of divergence and deadlock are addressed. For example, actors permit dynamic deadlock detection and removal. The problem of divergence is contained because independent transactions can execute concurrently and potentially infinite processes are nevertheless available for interaction.

A Detailed Look at Scale and Translation Invariance in a Hierarchical Neural Model of Visual Object Recognition

The HMAX model has recently been proposed by Riesenhuber & Poggio as a hierarchical model of position- and size-invariant object recognition in visual cortex. It has also turned out to model successfully a number of other properties of the ventral visual stream (the visual pathway thought to be crucial for object recognition in cortex), and particularly of (view-tuned) neurons in macaque inferotemporal cortex, the brain area at the top of the ventral stream. The original modeling study only used ``paperclip'' stimuli, as in the corresponding physiology experiment, and did not explore systematically how model units' invariance properties depended on model parameters. In this study, we aimed at a deeper understanding of the inner workings of HMAX and its performance for various parameter settings and ``natural'' stimulus classes. We examined HMAX responses for different stimulus sizes and positions systematically and found a dependence of model units' responses on stimulus position for which a quantitative description is offered. Interestingly, we find that scale invariance properties of hierarchical neural models are not independent of stimulus class, as opposed to translation invariance, even though both are affine transformations within the image plane.

Core-Shell Assisted Bimetallic Assembly of Pt and Ru Nanoparticles by DNA Hybridization

We have discovered that the current protocols to assemble Au nanoparticles based on DNA hybridization do not work well with the small metal nanoparticles (e.g. 5 nm Au, 3.6 nm Pt and 3.2 nm Ru particles). Further investigations revealed the presence of strong interaction between the oligonucleotide backbone and the surface of the small metal nanoparticles. The oligonucleotides in this case are recumbent on the particle surface and are therefore not optimally oriented for hybridization. The nonspecific adsorption of oligonucleotides on small metal nanoparticles must be overcome before DNA hybridization can be accepted as a general assembly method. Two methods have been suggested as possible solutions to this problem. One is based on the use of stabilizer molecules which compete with the oligonucleotides for adsorption on the metal nanoparticle surface. Unfortunately, the reported success of this approach in small Au nanoparticles (using K₂BSPP) and Au films (using 6-mercapto-1-hexanol) could not be extended to the assembly of Pt and Ru nanoparticles by DNA hybridization. The second approach is to simply use larger metal particles. Indeed most reports on the DNA hybridization induced assembly of Au nanoparticles have made use of relatively large particles (>10 nm), hinting at a weaker non-specific interaction between the oligonucleotides and large Au nanoparticles. However, most current methods of nanoparticle synthesis are optimized to produce metal nanoparticles only within a narrow size range. We find that core-shell nanoparticles formed by the seeded growth method may be used to artificially enlarge the size of the metal particles to reduce the nonspecific binding of oligonucleotides. We demonstrate herein a core-shell assisted growth method to assemble Pt and Ru nanoparticles by DNA hybridization. This method involves firstly synthesizing approximately 16 nm core-shell Ag-Pt and 21 nm core-shell Au-Ru nanoparticles from 9.6 nm Ag seeds and 17.2 nm Au seeds respectively by the seed-mediated growth method. The core-shell nanoparticles were then functionalized by complementary thiolated oligonucleotides followed by aging in 0.2 M PBS buffer for 6 hours. The DNA hybridization induced bimetallic assembly of Pt and Ru nanoparticles could then be carried out in 0.3 M PBS buffer for 10 hours.

Molecular computations for reactions and phase transitions: applications to protein stabilization, hydrates and catalysis

In this work we have made significant contributions in three different areas of interest: therapeutic protein stabilization, thermodynamics of natural gas clathrate-hydrates, and zeolite catalysis. In all three fields, using our various computational techniques, we have been able to elucidate phenomena that are difficult or impossible to explain experimentally. More specifically, in mixed solvent systems for proteins we developed a statistical-mechanical method to model the thermodynamic effects of additives in molecular-level detail. It was the first method demonstrated to have truly predictive (no adjustable parameters) capability for real protein systems. We also describe a novel mechanism that slows protein association reactions, called the “gap effect.” We developed a comprehensive picture of methioine oxidation by hydrogen peroxide that allows for accurate prediction of protein oxidation and provides a rationale for developing strategies to control oxidation. The method of solvent accessible area (SAA) was shown not to correlate well with oxidation rates. A new property, averaged two-shell water coordination number (2SWCN) was identified and shown to correlate well with oxidation rates. Reference parameters for the van der Waals Platteeuw model of clathrate-hydrates were found for structure I and structure II. These reference parameters are independent of the potential form (unlike the commonly used parameters) and have been validated by calculating phase behavior and structural transitions for mixed hydrate systems. These calculations are validated with experimental data for both structures and for systems that undergo transitions from one structure to another. This is the first method of calculating hydrate thermodynamics to demonstrate predictive capability for phase equilibria, structural changes, and occupancy in pure and mixed hydrate systems. We have computed a new mechanism for the methanol coupling reaction to form ethanol and water in the zeolite chabazite. The mechanism at 400°C proceeds via stable intermediates of water, methane, and protonated formaldehyde.