Automatic Learning System Research Articles

Text/Image Region Separation for Document Layout Detection of Old Document Images Using Non-linear Diffusion and Level Set

Text/Image region separation is the process of identifying location of various text and image regions in a scanned document image. This is particularly helpful in detecting the layout of a scanned document image. The text region thus obtained can be used for optical character recognition (OCR) operation. The text region can be used to label and train automatic layout learning system to detect locations of title, keywords, subheadings, paragraphs, image locations etc. In case of regular image and text boundaries, Profiling or morphological operations can be used for separating the text and image regions and to achieve correct document layout out detection. However, the real-world documents will have irregular boundaries and noise, the usual profile based methods and its heuristic often fails. This will lead to incorrect document layouts. This paper proposes to use edge enhancement diffusion and level set method for text/image region separation from scanned document images. The result obtained shows that the proposed method works when the document contain multiple images. The proposed method detects the layout of the scanned document even when the image and the text regions have irregular shape.

A Semantic Rule-based Approach Supported by Process Mining for Personalised Adaptive Learning

Currently, automated learning systems are widely used for educational and training purposes within various organisations including, schools, universities and further education centres. There has been a big gap between the extraction of useful patterns from data sources to knowledge, as it is crucial that data is made valid, novel, potentially useful and understandable. To meet the needs of intended users, there is requirement for learning systems to embody technologies that support learners in achieving their learning goals and this process don’t happen automatically. This paper propose a novel approach for automated learning that is capable of detecting changing trends in learning behaviours and abilities through the use of process mining techniques. The goal is to discover user interaction patterns within learning processes, and respond by making decisions based on adaptive rules centred on captured user profiles. The approach applies semantic annotation of activity logs within the learning process in order to discover patterns automatically by means of semantic reasoning. Therefore, our proposed approach is grounded on Semantic Modelling and Process Mining techniques. To this end, it is possible to apply effective reasoning methods to make inferences over a Learning Process Knowledge-Base that leads to automated discovery of learning patterns or behaviour.

Supervised Neural Networks for the Automatic Classification of Leukocytes in Blood Microscope Images

In this paper, the application of the supervised learning system to automatic classification of leukocytes processing for the microscopic images analysis is presented. The traditional pattern classification in cellular images is typically made by experienced operators. Such procedures may present a non-standard and unstable accuracy when it depends on the operator’s capabilities and tiredness. In this study, we propose the supervised learning system to achieve an automated segmentation and classification of leukocytes based on supervised neural networks and image processing methods. The experimental results show that the proposed automatic classification learning system can effectively classify the five types of the leukocytes in microscopic cell images, as well as to compare the classification results to those obtained by the medical experts.

Comparison of Question Answering Systems Based on Ontology and Semantic Web in Different Environment

Problem statement: Question Answering (QA) system is taking an important role in current search engine optimization concept. Natural language processing technique is mostly implemented in QA system for asking user's question and several steps are also followed for conversion of questions to query form for getting a n exact answer. Approach: This paper surveys different types of question answering system based on ontology and semantic web model with different query format. For comparison, the types of input, q uery processing method, input and output format of each system and the performance metrics with its li mitations are analyzed and discussed. Our question answering for automatic learning system architectur e is used to overcome the difficulties raised from the different QA models. Results: The semantic search methodology is implemented by using RDF graph in the application of data structure domain a nd the performance is also analyzed. Answers are retrieved from ontology using Semantic Search approach and question-to-query algorithm is evaluated in our system for analyzing performance evaluation. Conclusion: Performance of question answering system of getting exact result can be improved by u sing semantic search methodology for retrieving answers from ontology model. Our system successfully implements this technique and the system is also used in intelligent manner for automatic learn ing method.

Development of Diagnostic Classifier for Ultrasound Liver Lesion Images

Liver cancer is the fifth most common cancer worldwide in men and eighth in women, and is one of the few cancers still on the rise. In this study, we focus on Development of Diagnostic Classifier for Ultrasound liver lesion. Naturaly, Ultrasound liver lesion images are having more spackle noise. Developing classifier for ultrasound liver lesion image is a challenging task. We approach fully automatic machine learning system for developing this classifier. First, we segment the liver image by calculating the textural features from co-occurrence matrix and run length method. For classification, Support Vector Machine is used based on the risk bounds of statistical learning theory. The textural features for different features methods are given as input to the SVM individually. Performance analysis train and test datasets carried out separately using SVM Model. Whenever an ultrasonic liver lesion image is given to the SVM classifier system, the features are calculated, classified, as normal, benign and malignant liver lesion. We hope the result will be helpful to the physician to identify the liver cancer in non invasive method.

The Design of Advertising Exchanges

Internet advertising exchanges possess three characteristics—fast delivery, low values, and automated systems—that influence market design. Automated learning systems induce the winner’s curse when several pricing types compete. Bidders frequently compete with different data, which induces randomization in equilibrium. Machine learning causes the value of information to leak across participants. Discrimination may be used to induce efficient exploration, although publishers (websites) may balk at participating. The creation of “learning accounts,” which divorce payments from receipts, may be used to internalize learning externalities. Under some learning mechanisms the learning account eventually shows a surplus. The solution is illustrated computationally.

901 CADの多様な操作の自動学習システム(生産加工・工作機械I)

901 CADの多様な操作の自動学習システム(生産加工・工作機械I)

Comparing Learning Methods

In this article we perform some experiments to study how an automatic system learns a set of rules from its interaction with an artificial environment. In particular, we are interested in comparing these capabilities to the skills shown by humans to learn the same rules in similar conditions. We perform this analysis by conducting two experiments. On the one hand, we observe the evolution of the automatic learning system in terms of its performance along time. At the beginning, the system does not know the rules, but it can observe the positive/negative results of its decisions. As its knowledge about the environment becomes more precise, its performance improves. On the other hand, seventy students faced the same artificial environment in the same conditions, though this time the experiment was presented as a game. The objective of the game consists in gaining points, but the rules of the game are not known a priori. So, there is a clear incentive for finding them out. We use these experiments to compare the learning curves of both humans and automatic systems, and we use this information to analyze the similarities/differences between both learning processes. In particular, we are interested in assessing how close the automatic system is from passing the Turing test.

Differential involvement of the central amygdala in appetitive versus aversive learning

Understanding the function of the distinct amygdaloid nuclei in learning comprises a major challenge. In the two studies described herein, we used c-Fos immunolabeling to compare the engagement of various nuclei of the amygdala in appetitive and aversive instrumental training procedures. In the first experiment, rats that had already acquired a bar-pressing response to a partial food reinforcement were further trained to learn that an acoustic stimulus signaled either continuous food reinforcement (appetitive training) or a footshock (aversive training). The first training session of the presentation of the acoustic stimulus resulted in significant increases of c-Fos immunolabeling throughout the amygdala; however, the pattern of activation of the nuclei of the amygdala differed according to the valence of motivation. The medial part of the central amygdala (CE) responded, surprisingly, to the appetitive conditioning selectively. The second experiment was designed to extend the aversive versus appetitive conditioning to mice, trained either for place preference or place avoidance in an automated learning system (INTELLICAGE). Again, much more intense c-Fos expression was observed in the medial part of the CE after the appetitive training as compared to the aversive training. These data, obtained in two species and by means of novel experimental approaches balancing appetitive versus aversive conditioning, support the hypothesis that the central nucleus of the amygdala is particularly involved in appetitively motivated learning processes.

An Automated Learning System for Java Programming

The increase in the numbers enrolling in university computer courses makes huge demands on resources, and maintaining standards of teaching and tutorial support is almost impossible without a massive increase in staff. To overcome this situation, the authors have developed RoboProf, an automated learning environment which, as well as generating and assessing programming exercises, provides ongoing assistance and feedback to students without extra demands on lecturer and tutors' time. This system also contains a technique for detecting plagiarism, an increasing problem in computing courses worldwide. For this research, RoboProf was used to teach Java programming to a class containing nearly 300 students in the first year of a computing degree. Use of the system by students was monitored and recorded on log files in order to investigate the extent to which usage patterns influence achieved programming skill. An analysis shows that students who complete the set of RoboProf exercises perform significantly better than those who do not. The timeliness in which these exercises are completed relative to other students is significant: early solvers get higher marks, and students solving the problems with fewer attempts get higher marks. Not surprisingly, plagiarists achieve a lower score than those who do their own work. Other factors that were found to influence programming performance included entry standards and gender. Entry qualifications impacted positively on performance, and males performed significantly better than females. There was a significant positive correlation between the score achieved in the RoboProf course and the performance in a subsequent computing course administered in the traditional manner.

Hierarchical structure and word strength prediction of Mandarin prosody

We use Stem-ML to build an automatic learning system for Mandarin prosody that allows us to make quantitative measurements of prosodic strengths. Stem-ML is a phenomenological model of the muscle dynamics and planning process that controls the tension of the vocal folds. Because Stem-ML describes the interactions between nearby tones or accents, we were able to use a highly constrained model with only one accent template for each lexical tone category, and a single prosodic strength per word. The model accurately reproduces the intonation of the speaker, capturing 87% of the variance of the speech's fundamental frequency, f0. The result reveals strong alternating metrical patterns in words, and suggests that the speaker uses word strength to mark a hierarchy of sentence, clause, phrase, and word boundaries.

自己組織化マップの自動追加学習システム(SOMとその応用)

自己組織化マップの自動追加学習システム(SOMとその応用)

Utilising the student model in distance learning

A key to effective automated learning systems is the ability to dynamically adapt learning strategies to the needs of individual students. This presentation will describe the dynamic presentation facility of the DANDIE distance-based learning system that utilises a student model and intelligent system techniques to optimise presentation content and form to meet individual needs. Given a knowledge-base of course material, the system adapts the information through the selection of predefined learning strategies, dynamically constructs an interconnected set of presentation modules, and evaluates the effectiveness of the presentation strategy with respect to the individual student.

Indeterminate fine-needle aspiration of the breast

Fine-needle aspiration (FNA) of the breast, although effective for the diagnosis of breast carcinoma, has a significant drawback. A minority of cases cannot be classified as benign or malignant. These FNAs are assigned an inconclusive diagnosis, often prompting surgical biopsy. Surgery is justified in some of these cases, but many of these lesions are benign. If these inconclusive FNAs could be accurately diagnosed as benign or malignant, many of these patients might avoid having to undergo surgical biopsy. An image analysis and an automated learning system that was developed at the University of Wisconsin (Xcyt) was used to categorize 56 (37 benign and 19 malignant) breast FNAs diagnosed as "indeterminate" and the computer diagnosis compared with the surgical biopsy. For each case, an operator chose a group of cells within a single field on the FNA slide and digitized this image using a video camera. The outline of each nucleus was manually outlined, and the exact border was delineated by the computer. Based on the analysis of three nuclear features (area, texture, and smoothness), the Xcyt system computed a benign or malignant diagnosis and a corresponding probability of malignancy for each case. Probabilities of malignancy for the respective cases ranged from 0.0-1.0. Benign cases were defined as those having probabilities of malignancy < 0.3; those with probabilities above this limit were considered malignant. Using these criteria, the computer identified 33 cases as benign and 23 cases as malignant. When compared with the surgical biopsy, 42 of the cases (75%) were correctly classified with a sensitivity and specificity of 73.7% and 75.7%, respectively. There were only 5 false-negative cases with a false-negative rate of 13.5% and a predictive value of a negative test of 84.8%. When faced with inconclusive diagnoses of FNAs of breast masses, the authors believe that image analysis may be used as an aid in the further classification of such lesions, thereby providing a more appropriate triage for surgical biopsy.

Machine Learning of Weather Forecasting Rules from Large Meteorological Data Bases

Discovery of useful forecasting rules from observational weather data is an outstanding interesting topic. The traditional methods of acquiring forecasting knowledge are manual analysis and investigation performed by human scientists. This paper presents the experimental results of an automatic machine learning system which derives forecasting rules from real observational data. We tested the system on the two large real data sets from the areas of central China and Victoria of Australia. The experimental results show that the forecasting rules discovered by the system are very competitive to human experts. The forecasting accuracy rates are 86.4% and 78% of the two data sets respectively.

A New Expert System on Homoeopathy using Concept Theory based Rule Induction

The present paper describes an automated learning system for knowledge acquisition of Homoeopathic expert systems by the application of Concept Theory (CT). The treatment in Homoeopathy is solely based on individualisation, which is a process to single out a patient In Homoeopathy, if individualising symptoms are same, then irrespective of disease symptoms, the appropriate medicine to be prescribed is also same. The induction of Concept Rule (CR), Modified Concept Rule (MCR) and Generalised Modified Concept Rule (GMCR) from the input training instances, comprising individualising symptoms and diseases, rediscovers the individualising law of Homoeopathy and afterwards selects the medicine corresponding to that particular disease, provided a particular individualising symptom set is supplied.

Explanation-based machine learning in engineering design

The development of an unguided or automatic learning system in the domain of structural design is reported. The machine learning technique used is Explanation-Based Learning (EBL). A prototype system, called Structural Design Learning System (SDLS), has been developed in a combination of Prolog and Pascal languages. SDLS includes a modified implementation of EBL in Prolog, where the two stages of explanation and generalization are combined into one stage. A more-formal definition of the operationality criterion suitable for the domain of structural design is proposed and used in SDLS.

Global shape recognition of 3-D objects using a differential library storage

Shape recognition of three-dimensional rigid objects when viewed by a two-dimensional imaging system is discussed. Fourier descriptor features are used that are invariant to scale, translation, and rotation about the viewing axis. A differential library storage method is developed which allows compact storage of the many shapes generated by viewing an object from different aspect angles. Experimental results are presented using aircraft and normalized Fourier descriptors which show the effectiveness of the differential library storage. An automatic shape learning system is also demonstrated.