PYRAMID MATCH KERNEL AND CLASSIFIER ENSEMBLE-BASED MIL ALGORITHM FOR PORNOGRAPHIC IMAGES FILTERING

In this paper, a novel multi-instance learning (MIL) algorithm based on pyramid match kernel (PMK) and classifier ensemble is proposed for recognizing pornographic scene from image database. First, an improved JSEG image segmentation technique is deployed for dividing every image into several regions, and regards the whole image as a "bag", the low-level visual features (i.e. color and texture) of each segmented region as "instance". As a result, the pornographic images filtering problem can be transferred into a typical MIL problem. Second, similarity between the multi-instance bags is measured by PMK method, which allows MIL problem to be solved directly by the support vector machine (SVM). Finally, many base classifiers based on PMK with different levels are constructed, and the performance weighting rule is used to dynamically determine the weights of them, so the strategy of classifier ensemble is used to improve the filtering accuracy. In a real condition image set that the ratio of normal image to pornographic image is 9:1, experimental results show that the proposed algorithm, named PMKCE-MIL, is robust, and its performance is superior to other algorithms.

Decision Makers' Ability To Identify Unusual Costs And Implications For Alternative Estimation Procedures

This paper reports the results of an experiment in which individuals visually fitted a cost function to data. The inclusion or omission of unusual data points within the data set was experimentally manipulated. The results indicate that individuals omit outliers from their visual fits, but do not omit influential points. Evidence also suggests that the weighting rule used by individuals is more robust that the weighting rule used in the ordinary least squares criterion.

A Systematized FMEA Practice for Magnet Relays

The Design Failure Modes and Effects Analysis (DFMEA) are generally applied to risk management of New Product Development (NPD) through standardization of potential failure modes and effect-ranking of rating criterion with failure modes. Typical 1 to 10 of effect-ranking are widely weighed the priority of classification, that framing effects and status quo senses might cause decision trap happening thus. The FMEA follows considerable indexes which are including Severity, Occurrence and Detection, and need be associated with difference between every two failures individually. However, we suspect that a more systematic construction of the analysis by which failure modes belong is necessary in order to make intellectual progress in this area. Two ways of such differentiation and construction are improvable effect-ranking and systematized indexes; here we resolve for attributes of failures with classification, maturity and experiance of indexes according to an existing rule. In Severity model, the larger differentiation is achieved by separating indexes to the classification of the Law & Regulation, Function and Cosmetic. Occurrence model has its characteristic a reliable ranking indexwhich assists decisionmakers to manage their venture. This is the model most closely associate with product maturity by grouping indexes to the new, extend and series product. Detection model offers a special perspective on cost; here the connections concerned with phase occasion of the review, verification and validation. Such differentiations will be proposed and mapped with the Life Cycle Profile (LCP) to systematize FMEA. Meanwhile, a more reasonable Risk Priority Number (RPN) with the new weighting rule will be worked out for effect-ranking and management system will be integrated systematiclly

Test of a Fuzzy Set Based Decision Model as an AID in Solving Human Factors Design Problems

Fuzzy Set Theory has proved popular for development of decision making models. However, most such models have not been tested using problems such as commonly found in Human Factors system design. This study used a decision model that combined Fuzzy Set decision rules with an eigenvector weighting rule. Five experienced Human Factors Designers solved six design problems, half manually, and half using a computer program that served as a decision making aid, using the model. On completion of the procedure, the computer model made a recommendation for a solution. The user could accept or reject the model's choice. Comparisons were made between manual and computer aided decision making, and the Fuzzy Set decision rule was compared with other possible decision rules using the same data. Results showed that use of the model-based decision aid was accepted by the users, and were reasonable. In addition, a possible measure of decision making quality was found in the measure of weighting inconsistency which is part of the eigenvector procedure.