We define instance compressibility ([13, 17]) for parametric problems in the classes PH and PSPACE.We observe that the problem ƩiCIRCUITSAT of deciding satisfiability of a quantified Boolean circuit with i-1 alternations of quantifiers starting with an existential quantifier is complete for parametric problems in the class Ʃp/i with respect to w-reductions, and that analogously the problem QBCSAT (Quantified Boolean Circuit Satisfiability) is complete for parametric problems in PSPACE with respect to w-reductions. We show the following results about these problems: 1. If CIRCUITSAT is non-uniformly compressible within NP, then ƩiCIRCUITSAT is non-uniformly compressible within NP, for any i≥1. 2. If QBCSAT is non-uniformly compressible (or even if satisfiability of quantified Boolean CNF formulae is non-uniformly compressible), then PSPACE ⊆ NP/poly and PH collapses to the third level. Next, we define Succinct Interactive Proof (Succinct IP) and by adapting the proof of IP = PSPACE ([11, 6]) , we show that QBCNFSAT (Quantified Boolean Formula (in CNF) Satisfiability) is in Succinct IP. On the contrary if QBCNFSAT has Succinct PCPs ([32]) , Polynomial Hierarchy (PH) collapses. After extending the notion of instance compression to higher classes, we study the hierarchical structure of the parametric problems with respect to compressibility. For that purpose, we extend the existing definition of VC-hierarchy ([13]) to parametric problems. After that, we have considered a long list of natural NP problems and tried to classify them into some level of VC-hierarchy. We have shown some of the new w-reductions in this context and pointed out a few interesting results including the ones as follows. 1. CLIQUE is VC1-complete (using the results in [14]). 2. SET SPLITTING and NAE-SAT are VC2-complete. We have also introduced a new complexity class VCE in this context and showed some hardness and completeness results for this class. We have done a comparison of VC-hierarchy with other related hierarchies in parameterized complexity domain as well. Next, we define the compression of counting problems and the analogous classification of them with respect to the notion of instance compression. We define #VC-hierarchy for this purpose and similarly classify a large number of natural counting problems with respect to this hierarchy, by showing some interesting hardness and completeness results. We have considered some of the interesting practical problems as well other than popular NP problems (e.g., #MULTICOLOURED CLIQUE, #SELECTED DOMINATING SET etc.) and studied their complexity for both decision and counting version. We have also considered a large variety of circuit satisfiability problems (e.g., #MONOTONE WEIGHTED-CNFSAT, #EXACT DNF-SAT etc.) and proved some interesting results about them with respect to the theory of instance compressibility.

No description available.

The implementation of data mining techniques in the medical area has generated great interest because of its potential for more efficient, economic and robust performance when compared to physicians. In this paper, we focus on the implementation of Multiple-Instance Learning (MIL) in the area of medical image mining, particularly to hard exudates detection in retinal images from diabetic patients. Our proposed approach deals with the highly noisy images that are common in the medical area, improving the detection specificity while keeping the sensitivity as high as possible. We have also investigated the effect of feature selection on system performance. We describe how we implement the idea of MIL on the problem of retinal image mining, discuss the issues that are characteristic of retinal images as well as issues common to other medical image mining problems, and report the results of initial experiments. / Singapore-MIT Alliance (SMA)

The objective of the present work is to propose a novel method to extract a stable feature set representative of image content. Each image is represented by a linear combination of fractal orthonormal basis vectors. The mapping coefficients of an image projected onto each orthonormal basis constitute the feature vector. The set of orthonormal basis vectors are generated by utilizing fractal iterative function through target and domain blocks mapping. The distance measure remains consistent, i.e., isometric embedded, between any image pairs before and after the projection onto orthonormal axes. Not only similar images generate points close to each other in the feature space, but also dissimilar ones produce feature points far apart. The above statements are logically equivalent to that distant feature points are guaranteed to map to images with dissimilar contents, while close feature points correspond to similar images. In this paper, we adapt the Multiple Instance Learning paradigm using the Diverse Density algorithm as a way of modeling the ambiguity in images in order to learning concepts used to classify images. A user labels an image as positive if the image contains the concepts, as negative if the image far from the concepts. Each example image is a bag of blocks where only the bag is labeled. The User selects positive and negative image examples to train the concepts in feature space. From a small collection of positive and negative examples, the system learns the concepts using them to retrieve images that contain the concepts from database. Each concept having similar blocks becomes the group in each image. According groups¡¦ location distribution, variation and spatial relations computes positive examples and database images similarity.

During the Digital Period, the more people using these digital video. When there are more and more users and amount of video data, the management of video data becomes a significant dimension during development. Therefore, there are more and more studying of accomplishing video database system, which provide users to search and get them. In this paper, a novel method for Video Scene Change Detection and video database retrieval is proposed. Uses Fractal orthonormal bases to guarantee the similar index has the similar image the characteristic union support vector clustering, splits a video into a sequence of shots, extracts a few representative frames(key-frames) to take the video database index from each shot. When image search compared to, according to MIL to pick up the characteristic, which images pursues the video database to have the similar characteristic, computation similar, makes the place output according to this.

Diss.--Bielefeld, 2000. / Bibliogr. p. 133-147.

La notion de partie comportant de nombreuses incertitudes, tant théoriques que pratiques, l'objet de cette thèse est de proposer une approche renouvelée de cette notion, en ne recherchant pas seulement quels sont les critères de détermination de la qualité de partie mais aussi quelles sont les personnes ayant vocation à disposer de cette qualité procédurale, car beaucoup de problèmes entourant cette notion nécessitent de pouvoir répondre à cette dernière interrogation. Il est ressorti de cette étude que le respect du principe selon lequel nul ne peut être jugé sans avoir été mis en mesure de se défendre impose que toute personne susceptible d'être jugée ait la qualité de partie, et que, sauf exceptions, le respect du devoir de non-immixtion dans les affaires d'autrui et de l'objectif de bonne administration de la justice implique de ne pas octroyer la qualité de partie à d'autres personnes que celles susceptibles d'être jugées. Dès lors, nous avons pu vérifier si les personnes étant parties ou tiers étaient légitimes à l'être et proposer, lorsque tel n'était pas le cas, des solutions pour y remédier. En outre, en analysant à la lumière de l'approche centrée sur la vocation à être partie les divers cas posant des difficultés de qualification procédurale, nous avons pu démontrer que les catégories de tiers et de parties ne nécessitent pas d'être complétées de classifications intermédiaires. En revanche, étudier isolément chaque situation à l'aune de cette nouvelle approche a permis de faire en sorte, pour tous ces cas jusqu'alors incertains, que chaque personne dispose de la qualité procédurale et des prérogatives auxquelles elle pouvait légitimement prétendre. / No English summary available.

Bien que profondément réformé ces vingt dernières années, l’office du juge administratif de la légalité reste caractérisé par certaines insuffisances. Faute de pouvoir se saisir d’office de certaines conclusions ou de tout moyen d’annulation, le juge chargé de censurer les actes administratifs illégaux est parfois contraint de rejeter des requêtes en annulation pourtant recevables et fondées. Cette défectuosité du contrôle juridictionnel opéré trouve son origine dans l’insuffisante prise en considération de la singularité de la fonction juridictionnelle exercée, qui ne consiste pas exclusivement à trancher un litige mais également à contrôler la légalité d’un acte. Le parachèvement de l’office du juge administratif de la légalité suppose son émancipation des éléments constitutifs de l’instance tels que fixés par les parties afin de lui conférer les pouvoirs adaptés à l’accomplissement de la mission qui lui incombe. A cette fin, l’étude des fondements de la prohibition de l’autosaisine permet de démontrer qu’aucun obstacle d’ordre théorique ou pratique ne s’oppose véritablement ni à la consécration des conclusions d’ordre public, ni à la généralisation des moyens d’ordre public. Seules des limites d’ordre conceptuel, nécessairement contingentes et dès lors surmontables, expliquent l’inertie à laquelle s’astreint encore le juge administratif de la légalité dans la détermination de la matière litigieuse. / Although profoundly reformed in the course of the past twenty years, the judicial recourse for abuse of power led by the administrative judge still falls short.Prevented to review, on its own initiative, the legality of the whole administrative act, the judge is sometimes bound to reject some complaints although they were admissible and well-founded. This deficiency in judicial review originated in the insufficient consideration of the unique nature of the judicial recourse for abuse of power which does not only entail settling a case but also ensuring the legality review of administrative decisions.To ensure the completion of his powers the administrative judge will need to liberate himself from the sole content of the submissions and arguments of the applicants and be allowed to expand his powers in order to accomplish his purpose.The analysis of the rationale to this prohibition of judge-initiated legal argumentsdemonstrates that there isn’t any theoretical or practical obstacle to the possibility for the administrative judge to automatically examine the legality of public policy arguments and conclusions. Only conceptual limitations, necessarily incidental, explain the inertia of the administrative judge in the determination of the matter at hand.

In this thesis, we explore the use of Convolutional Neural Networks for semantic and instance segmentation, with a focus on studying the application of existing methods with cheaper neural networks. We modify a fast object detection architecture for the instance segmentation task, and study the concepts behind these modifications both in the simpler context of semantic segmentation and the more difficult context of instance segmentation. Various instance segmentation branch architectures are implemented in parallel with a box prediction branch, using its results to crop each instance's features. We negate the imprecision of the final box predictions and eliminate the need for bounding box alignment by using an enlarged bounding box for cropping. We report and study the performance, advantages, and disadvantages of each. We achieve fast speeds with all of our methods.

Most widely accepted measures of performance for learning algorithms, such as accuracy and area under the ROC curve, provide information about behavior at the data set level. They say nothing about which instances are misclassified, whether two learning algorithms with the same classification accuracy on a data set misclassify the same instances, or whether there are instances misclassified by all learning algorithms. These questions about behavior at the instance level motivate our empirical analysis of instance hardness, a measure of expected classification accuracy for an instance. We analyze the classification of 57 data sets using 9 learning algorithms. Of the over 175000 instances investigated, 5% are misclassified by all 9 of the considered learning algorithms, and 15% are misclassified by at least half. We find that the major cause of misclassification for hard instances is class overlap, manifested as outliers and border points which can be exacerbated by class skew. We analyze these causes and show to what extent each leads to misclassifications, both in isolation and jointly. 19.8% of all misclassified instances are outliers; 71.3% are border points; 21% belong to a minority class. We also find that 91.6% of all outliers and 38.3% of all border points are misclassified whereas only 3.5% of instances without class overlap are misclassified. We propose a set of heuristics to predict when an instance will be hard to correctly classify. Additionally, we analyze how different learning algorithms perform on tasks with varying degrees of outliers, border points and class skew.