Humans excel at identifying and locating multiple instances of objects or persons in a scene, despite large variations in lighting conditions, pose or scale. In order to achieve this level of robustness, humans naturally use high-level semantic information. The notion of semantics is hierarchical in nature, for example, a house is constituted of walls, floor and ceiling. When certain entities in this hierarchy can be modeled using a functional form known a priori, we refer to the data as structured, while when the functional form is unknown, the data is unstructured. In this thesis, we address the problem of discovering multiple instances of models in structured and unstructured data. The first part of this thesis deals with structured data. We identify planar regions in an indoor scene by using a single depth image from a Microsoft Kinect sensor. The clutter in the indoor scenes, the depth dependent measurement noise and the unknown number of planar regions pose serious challenges in model discovery. We propose a scalable bottom-up approach that leverages from a heteroscedastic, i.e., point dependent model of the measurement noise. The second part of the thesis addresses multiple model discovery in unstructured data in a semi-supervised setup. We develop a framework for using mean shift clustering in kernel spaces with a few user-specified pairwise constraints. A linear transformation of the initial kernel space is learned by the constrained minimization of a Bregman divergence based objective function. We automatically determine the adaptive bandwidth parameter to be used with mean shift clustering. Finally, we compare the performance with state-of-the-art semi-supervised clustering methods and show that kernel mean shift clustering performs particularly well when the number of clusters is large. We also propose a few directions for future research. Using the planar regions detected from the first frame of Kinect, a sequence of RGB-D images can be rapidly processed to dynamically generate a consistent 3D model of the scene. We also show that for the kernel learning problem, we can use ideas from group theory and semi-definite programming to devise a more efficient algorithm that only uses linearly independent constraints.
Humans excel at identifying and locating multiple instances of objects or persons in a scene, despite large variations in lighting conditions, pose or scale.
The volume is based on the papers that were presented at the Interna tional Conference Model-Based Reasoning in Scientific Discovery (MBR'98), held at the Collegio Ghislieri, University of Pavia, Pavia, Italy, in December 1998. The papers explore how scientific thinking uses models and explanatory reasoning to produce creative changes in theories and concepts. The study of diagnostic, visual, spatial, analogical, and temporal rea soning has demonstrated that there are many ways of performing intelligent and creative reasoning that cannot be described with the help only of tradi tional notions of reasoning such as classical logic. Traditional accounts of scientific reasoning have restricted the notion of reasoning primarily to de ductive and inductive arguments. Understanding the contribution of model ing practices to discovery and conceptual change in science requires ex panding scientific reasoning to include complex forms of creative reasoning that are not always successful and can lead to incorrect solutions. The study of these heuristic ways of reasoning is situated at the crossroads of philoso phy, artificial intelligence, cognitive psychology, and logic; that is, at the heart of cognitive science. There are several key ingredients common to the various forms of model based reasoning to be considered in this book. The models are intended as in terpretations of target physical systems, processes, phenomena, or situations. The models are retrieved or constructed on the basis of potentially satisfying salient constraints of the target domain.
The volume is based on the papers that were presented at the Interna tional Conference Model-Based Reasoning in Scientific Discovery (MBR'98), held at the Collegio Ghislieri, University of Pavia, Pavia, Italy, in December 1998.
Fruit flies are "little people with wings" goes the saying in the scientific community, ever since the completion of the Human Genome Project and its revelations about the similarity amongst the genomes of different organisms. It is humbling that most signalling pathways which "define" humans are conserved in Drosophila, the common fruit fly. Feed a fruit fly caffeine and it has trouble falling asleep; feed it antihistamines and it cannot stay awake. A C. elegans worm placed on the antidepressant flouxetine has increased serotonin levels in its tiny brain. Yeast treated with chemotherapeutics stop their cell division. Removal of a single gene from a mouse or zebrafish can cause the animals to develop Alzheimer’s disease or heart disease. These organisms are utilized as surrogates to investigate the function and design of complex human biological systems. Advances in bioinformatics, proteomics, automation technologies and their application to model organism systems now occur on an industrial scale. The integration of model systems into the drug discovery process, the speed of the tools, and the in vivo validation data that these models can provide, will clearly help definition of disease biology and high-quality target validation. Enhanced target selection will lead to the more efficacious and less toxic therapeutic compounds of the future. Leading experts in the field provide detailed accounts of model organism research that have impacted on specific therapeutic areas and they examine state-of-the-art applications of model systems, describing real life applications and their possible impact in the future. This book will be of interest to geneticists, bioinformaticians, pharmacologists, molecular biologists and people working in the pharmaceutical industry, particularly genomics.
Until the late 20th century, drug discovery was mainly a linear process based on
the screening and testing of thousands of chemical substances for therapeutic
activity. The drug discovery process could be broken down into the following
steps: target selection, assay development, primary screening for chemical hits,
hit to lead compound optimization, preclinical and clinical development and,
finally, market launch. Early bottlenecks such as the typically limited availability of
discovery ...
SMITHA V.P.. It can be seen that in spite of many findings and arguments against
minimally guided approach like discovery learning, variations in discovery
learning such as guided discovery learning has gained certain popularity in
recent times due to their relative advantages over pure discovery. Traditional
methods give emphasis more on content learning as compared to teaching
models like guided discovery learning specifically designed to increase thinking
skills in students.
