Fast Pose Estimation with Parameter Sensitive Hashing

Example-based methods are effective for parameter estimation problems when the underlying system is simple or the dimensionality of the input is low. For complex and high-dimensional problems such as pose estimation, the number of required examples and the computational complexity rapidly becme prohibitively high. We introduce a new algorithm that learns a set of hashing functions that efficiently index examples relevant to a particular estimation task. Our algorithm extends a recently developed method for locality-sensitive hashing, which finds approximate neighbors in time sublinear in the number of examples. This method depends critically on the choice of hash functions; we show how to find the set of hash functions that are optimally relevant to a particular estimation problem. Experiments demonstrate that the resulting algorithm, which we call Parameter-Sensitive Hashing, can rapidly and accurately estimate the articulated pose of human figures from a large database of example images.

Parameter sensitivity in simulations of snowmelt

Essery, RLH & P, Etchevers, (2004). Parameter sensitivity in simulations of snowmelt. Journal of Geophysical Research, 109, doi:10. 1029/2004JD005036.

Parameter Sensitive Detectors

Object detection can be challenging when the object class exhibits large variations. One commonly-used strategy is to first partition the space of possible object variations and then train separate classifiers for each portion. However, with continuous spaces the partitions tend to be arbitrary since there are no natural boundaries (for example, consider the continuous range of human body poses). In this paper, a new formulation is proposed, where the detectors themselves are associated with continuous parameters, and reside in a parameterized function space. There are two advantages of this strategy. First, a-priori partitioning of the parameter space is not needed; the detectors themselves are in a parameterized space. Second, the underlying parameters for object variations can be learned from training data in an unsupervised manner. In profile face detection experiments, at a fixed false alarm number of 90, our method attains a detection rate of 75% vs. 70% for the method of Viola-Jones. In hand shape detection, at a false positive rate of 0.1%, our method achieves a detection rate of 99.5% vs. 98% for partition based methods. In pedestrian detection, our method reduces the miss detection rate by a factor of three at a false positive rate of 1%, compared with the method of Dalal-Triggs.

Computing a Uniform Scaling Parameter for 3D Registration of Lung Surfaces

A difficulty in lung image registration is accounting for changes in the size of the lungs due to inspiration. We propose two methods for computing a uniform scale parameter for use in lung image registration that account for size change. A scaled rigid-body transformation allows analysis of corresponding lung CT scans taken at different times and can serve as a good low-order transformation to initialize non-rigid registration approaches. Two different features are used to compute the scale parameter. The first method uses lung surfaces. The second uses lung volumes. Both approaches are computationally inexpensive and improve the alignment of lung images over rigid registration. The two methods produce different scale parameters and may highlight different functional information about the lungs.

Nonlinear parameter retrieval from three- and four-wave mixing in metamaterials

We present a generalized nonlinear susceptibility retrieval method for metamaterials based on transfer matrices and valid in the nondepleted pump approximation. We construct a general formalism to describe the transfer matrix method for nonlinear media and apply it to the processes of three- and four-wave mixing. The accuracy of this approach is verified via finite element simulations. The method is then reversed to give a set of equations for retrieving the nonlinear susceptibility. Finally, we apply the proposed retrieval operation to a three-wave mixing transmission experiment performed on a varactor loaded split ring resonator metamaterial sample and find quantitative agreement with an analytical effective medium theory model. © 2010 The American Physical Society.

Role of Landau quantization on the neutron-drip transition in magnetar crusts

The role of a strong magnetic field on the neutron-drip transition in the crust of a magnetar is studied. The composition of the crust and the neutron-drip threshold are determined numerically for different magnetic field strengths using the experimental atomic mass measurements from the 2012 Atomic Mass Evaluation complemented with theoretical masses calculated from the Brussels-Montreal Hartree-Fock-Bogoliubov nuclear mass model HFB-24. The equilibrium nucleus at the neutron-drip point is found to be independent of the magnetic field strength. As demonstrated analytically, the neutron-drip density and pressure increase almost linearly with the magnetic field strength in the strongly quantizing regime for which electrons lie in the lowest Landau level. For weaker magnetic fields, the neutron-drip density exhibits typical quantum oscillations. In this case, the neutron-drip density can be either increased by about 14% or decreased by 25% depending on the magnetic field strength. These variations are shown to be almost universal, independently of the nuclear mass model employed. These results may have important implications for the physical interpretation of timing irregularities and quasiperiodic oscillations detected in soft gamma-ray repeaters and anomalous x-ray pulsars, as well as for the cooling of strongly magnetized neutron stars.

The FUELCON meta-architecture, II: alternatives for parameter prediction

We continue the discussion of the decision points in the FUELCON metaarchitecture. Having discussed the relation of the original expert system to its sequel projects in terms of an AND/OR tree, we consider one further domain for a neural component: parameter prediction downstream of the core reload candidate pattern generator, thus, a replacement for the NOXER simulator currently in use in the project.

A comprehensive copper stockpile leach model: background and parameter sensitivity

The design and development of a comprehensive computational model of a copper stockpile leach process is summarized. The computational fluid dynamic software framework PHYSICA+ and various phenomena were used to model transport phenomena, mineral reaction kinetics, bacterial effects, and heat, energy and acid balances for the overall leach process. In this paper, the performance of the model is investigated, in particular its sensitvity to particle size and ore permeability. A combination of literature and laboratory sources was used to parameterize the model. The simulation results from the leach model are compared with closely controlled column pilot scale tests. The main performance characteristics (e.g. copper recovery rate) predicted by the model compare reasonably well with the experimental data and clearly reflect the qualitiative behavior of the process in many respects. The model is used to provide a measure of the sensitivity of ore permeability on leach behavior, and simulation results are examined for several different particle size distributions.

Parameter sensitivity for optimal design of 65 nm node SOI transistors

Mixed-mode simulation, where device simulation is embedded directly within a circuit simulator, is used for the first time to provide scaling guidelines to achieve optimal digital circuit performance for double gate SOI MOSFETs. This significant advance overcomes the lack of availability of SPICE model parameters. The sensitivity of the gate delay and on-off current ratio to each of the key geometric and technological parameters of the transistor is quantified. The impact of the source-drain doping profile on circuit performance is comprehensively investigated.