Spurious Valleys in the Error Surface of Recurrent Networks—Analysis and Avoidance

This paper gives a detailed analysis of the error surfaces of certain recurrent networks and explains some difficulties encountered in training recurrent networks. We show that these error surfaces contain many spurious valleys, and we analyze the mechanisms that cause the valleys to appear. We demonstrate that the principle mechanism can be understood through the analysis of the roots of random polynomials. This paper also provides suggestions for improvements in batch training procedures that can help avoid the difficulties caused by spurious valleys, thereby improving training speed and reliability.     

Microbiological survey of a South African poultry processing plant

Bacterial populations associated with poultry processing were determined on neck skin samples, equipment surfaces and environmental samples by replicate surveys. Aerobic plate counts, Enterobacteriaceae counts, Enterobacteriaceae counts and Pseudomonas counts were performed by standard procedures and the prevalence of Listeria, presumptive Salmonella and Staphylococcus aureus determined. Statistically significant (P < 0.05) increases in counts of all types of bacteria were obtained on product samples as a result of processing. Although bacterial counts on neck skin samples decreased by 0.3 to 0.4 log CFU g-1 after spray washing of carcasses, subsequent spinchilling and packaging of whole carcasses resulted in 0.7 to 1.2 log CFU g-1 increases. Bacterial numbers on equipment surfaces, however, decreased significantly from the "dirty" to the "clean" areas of the abattoir. Transport cages, "rubber fingers", defeathering curtains, shackles and conveyor belts repeatedly showed aerobic plate counts in excess of 5.0 log CFU 25 cm-2. Aerobic plate counts of scald tank and spinchiller water were 2 log CFU ml-1 higher than those of potable water samples. Bacterial numbers of the air in the "dirty" area were higher than those of the "clean" area. Listeria, presumptive Salmonella and Staphylococcus aureus were isolated from 27.6, 51.7 and 24.1% of all product samples, respectively, and Listeria and Staphylococcus aureus were also isolated from selected equipment surfaces.     

A Minimal Closed-form Solution for the Perspective Three Orthogonal Angles (P3oA) Problem: Application To Visual Odometry

We provide a simple closed-form solution to the Perspective three orthogonal angles (P3oA) problem: given the projection of three orthogonal lines in a calibrated camera, find their 3D directions. Upon this solution, an algorithm for the estimation of the camera relative rotation between two frames is proposed. The key idea is to detect triplets of orthogonal lines in a hypothesize-and-test framework and use all of them to compute the camera rotation in a robust way. This approach is suitable for human-made environments where numerous groups of orthogonal lines exist. We evaluate the numerical stability of the P3oA solution and the estimation of the relative rotation with synthetic and real data, comparing our results to other state-of-the-art approaches.     

Face tracking with automatic model construction

This paper describes an active model with a robust texture model built on-line. The model uses one camera and it is able to operate without active illumination. The texture model is defined by a series of clusters, which are built in a video sequence using previously encountered samples. This model is used to search for the corresponding element in the following frames. An on-line clustering method, named leaderP is described and evaluated on an application of face tracking. A 20-point shape model is used. This model is built offline, and a robust fitting function is used to restrict the position of the points. Our proposal is to serve as one of the stages in a driver monitoring system. To test it, a new set of sequences of drivers recorded outdoors and in a realistic simulator has been compiled. Experimental results for typical outdoor driving scenarios, with frequent head movement, turns and occlusions are presented. Our approach is tested and compared with the Simultaneous Modeling and Tracking (SMAT) [1], and the recently presented Stacked Trimmed Active Shape Model (STASM) [2], and shows better results than SMAT and similar fitting error levels to STASM, with much faster execution times and improved robustness.     

Using OGRO and CertiVeR to improve OCSP validation for Grids

Authentication and authorization in many distributed systems rely on the use of cryptographic credentials that in most of the cases have a defined lifetime. This feature mandates the use of mechanisms able to determine whether a particular credential can be trusted at a given moment. This process is commonly named validation. Among available validation mechanisms, the Online Certificate Status Protocol (OCSP) stands out due to its ability to carry near real time certificate status information. Despite its importance for security, OCSP faces considerable challenges in the computational Grid (i.e. Proxy Certificate’s validation) that are being studied at the Global Grid Forum’s CA Operations Work Group (CAOPS-WG). As members of this group, we have implemented an OCSP validation infrastructure for the Globus Toolkit 4, composed of the CertiVeR Validation Service and our Open GRid Ocsp (OGRO) client library, which introduced the Grid Validation Policy. This paper summarizes our experiences on that work and the results obtained up to now. Furthermore we introduce the prevalidation concept, a mechanism analogous to the Authorization Push-Model, capable of improving OCSP validation performance in Grids. This paper also reports the results obtained with OGRO’s prevalidation rules for Grid Services as a proof of concept.

Texture-based feature tracking for effective time-varying data visualization

Analyzing, visualizing, and illustrating changes within time-varying volumetric data is challenging due to the dynamic changes occurring between timesteps. The changes and variations in computational fluid dynamic volumes and atmospheric 3D datasets do not follow any particular transformation. Features within the data move at different speeds and directions making the tracking and visualization of these features a difficult task. We introduce a texture-based feature tracking technique to overcome some of the current limitations found in the illustration and visualization of dynamic changes within time-varying volumetric data. Our texture-based technique tracks various features individually and then uses the tracked objects to better visualize structural changes. We show the effectiveness of our texture-based tracking technique with both synthetic and real world time-varying data. Furthermore, we highlight the specific visualization, annotation, registration, and feature isolation benefits of our technique. For instance, we show how our texture-based tracking can lead to insightful visualizations of time-varying data. Such visualizations, more than traditional visualization techniques, can assist domain scientists to explore and understand dynamic changes.     

Odometry-Based Viterbi Localization with Artificial Neural Networks and Laser Range Finders for Mobile Robots

This paper proposes an approach that solves the Robot Localization problem by using a conditional state-transition Hidden Markov Model (HMM). Through the use of Self Organized Maps (SOMs) a Tolerant Observation Model (TOM) is built, while odometer-dependent transition probabilities are used for building an Odometer-Dependent Motion Model (ODMM). By using the Viterbi Algorithm and establishing a trigger value when evaluating the state-transition updates, the presented approach can easily take care of Position Tracking (PT), Global Localization (GL) and Robot Kidnapping (RK) with an ease of implementation difficult to achieve in most of the state-of-the-art localization algorithms. Also, an optimization is presented to allow the algorithm to run in standard microprocessors in real time, without the need of huge probability gridmaps.     

Characteristics of an immobilized lipase for the commercial synthesis of esters

The lipase-catalyzed synthesis of ester bonds has been well-documented lately and is of much current commercial interest. Immobilization of a fungal lipase on a unique macroporous support allows not only the ability to operate in non-aqueous media but to catalyze ester synthesis in quantitative yields, employing attractive commercial conditions. Catalyst dose and process configurations will be illustrated. The capability of the catalyst to operate efficiently in reverse under a variety of unnatural, hostile, solvent-containing environments will be discussed. The range of substrates for this immobilized lipase, Lipzome, has been investigated. The enzyme will catalyze ester synthesis with saturated, unsaturated and a variety of branched carboxylic acids. The alcohol specificity for this enzyme also is equally broad. A wide variety of straight-chain, branched and polar alcohols can be substrates. In addition, some examples of alcohol specificity for kinetic isomer resolution will be cited. Presented at the symposium “The Biology, Biotechnology and Technology of Lipases” at the 78th annual meeting of the American Oil Chemists’ Society held May 17–21, 1987, in New Orleans, Louisiana.     

Augmented reality annotations to assist persons with Alzheimers and their caregivers

Persons with Alzheimer’s disease (AD) and their caregivers implement diverse strategies to cope with memory loss. A common strategy involves placing tags on drawers or removing cabinet doors to make their contents visible. This study describes the Ambient aNnotation System (ANS), aimed at assisting people suffering from AD and their caregivers with this task. The system has two main modules: The tagging subsystem allows caregivers to create and manage ambient annotations in order to assist people with memory problems. The second subsystem allows people with AD to use a mobile phone to recognize tags in the environment and to receive relevant information in the form of audio, text, or images. The identification of these tags is performed in real time by uploading images from the mobile phone to a server, which uses the SURF algorithm for object recognition. We describe the design and implementation of the system as well as results of the evaluation of its performance and efficiency. ANS can process query images approximately every 2 s and is able to locate users in their homes with a precision of 0.93. A usability study conducted with six subjects determined that audio notifications are more effective than vibrating notifications to alert the user about tags in the environment.