Image classification is a fundamental task in computer vision, with wide applications including autonomous navigation, content recommendation, and environmental monitoring. This paper offers a comprehensive comparative analysis of deep learning techniques using the Intel Natural Scenes Image dataset, which features diverse scenes such as forests, mountains, seas, streets, buildings, and glaciers. Our study evaluates the performance of various Convolutional Neural Network (CNN) architectures, focusing on aspects such as network design, hyperparameters, data augmentation, and transfer learning strategies. We describe our experimental setup in detail, including the CNN models used, preprocessing techniques applied, and evaluation metrics employed. The results and discussions present key findings, highlighting the strengths and limitations of the approaches studied and providing guidance for future research and practical applications. Our systematic analysis yields valuable insights into effective strategies for recognizing natural scenes.

White Blood Cell (WBC) image classification is pivotal for early disease detection and diagnosis. Convolutional Neural Networks (CNNs) have emerged as potent tools for such tasks due to their ability to learn intricate features from raw pixel data. In this study, we present a CNN-based approach for automated WBC classification. Our methodology encompasses image preprocessing to enhance contrast and normalize color, succeeded by CNN training with multiple convolutional and pooling layers, thereby enabling feature acquisition from diverse WBC classes. We evaluate our approach using a publicly accessible WBC image dataset, comparing our results against other contemporary methods. Our proposed method achieves an impressive 96.2% accuracy for six distinct WBC classes, surpassing prior techniques by a considerable margin. This showcases CNNs’ potential in automated WBC classification, underscoring its significance in medical diagnosis and research. In summary, we introduce a CNN-based approach for automated WBC classification that attains state-of-the-art performance on a publicly available dataset. Our methodology encompasses image preprocessing, contrast enhancement, color normalization, and CNN training to capture distinctive features of diverse WBC classes. Our findings underscore CNNs’ promise in this domain and propose its deployment as a valuable tool in medical research and diagnosis. Subsequent efforts will explore advanced techniques like transfer learning to further elevate our method’s performance.

Privacy preserving analysis of a social network aims at a better understanding of the network and its behavior, while at the same time protecting the privacy of its individuals. We propose an anonymization method for weighted graphs, i.e., for social networks where the strengths of links are important. This is in contrast with many previous studies which only consider unweighted graphs. Weights can be essential for social network analysis, but they pose new challenges to privacy preserving network analysis. In this paper, we mainly consider prevention of identity disclosure, but we also touch on edge and edge weight disclosure in weighted graphs. We propose a method that provides k-anonymity of nodes against attacks where the adversary has information about the structure of the network, including its edge weights. The method is efficient, and it has been evaluated in terms of privacy and utility on real word datasets.

The development of 'intelligent' medical equipment, which can not only acquire various signals from the human body, but also process them and provide recommendations as to probable pathological conditions, will be highly beneficial for both the medical personnel and the patients. However, this necessitates the development and exploitation of advanced highly efficient classification techniques. In this direction this paper presents a novel ensemble classification technique, combining Random Forests with the `Markov Blanket' notion, which is used for the automated diagnosis of aortic and mitral heart valves diseases from low-cost and easily acquired heart sound signals. It has been tested in a highly 'difficult' global and heterogeneous dataset of 198 heart sound signals, which been acquired from both healthy and pathological medical cases. The proposed ensemble classification technique exhibited a higher classification performance in comparison with the classical Random Forest algorithms, and also other widely used classification algorithms.

Learning from preferences, which provide means for expressing a subject's desires, constitutes an important topic in machine learning research. This paper presents a comparative study of four alternative instance preference learning algorithms (both linear and nonlinear). The case study investigated is to learn to predict the expressed entertainment preferences of children when playing physical games built on their personalized playing features (entertainment modeling). Two of the approaches are derived from the literature-the large-margin algorithm (LMA) and preference learning with Gaussian processes-while the remaining two are custom-designed approaches for the problem under investigation: meta-LMA and neuroevolution. Preference learning techniques are combined with feature set selection methods permitting the construction of effective preference models, given suitable individual playing features. The underlying preference model that best reflects children preferences is obtained through neuroevolution: 82.22% of cross-validation accuracy in predicting reported entertainment in the main set of game survey experimentation. The model is able to correctly match expressed preferences in 66.66% of cases on previously unseen data (p-value = 0.0136) of a second physical activity control experiment. Results indicate the benefit of the use of neuroevolution and sequential forward selection for the investigated complex case study of cognitive modeling in physical games.

Researchers have portrayed an increasing effort towards providing formal computational frameworks to consolidate the plethora of notions and relations used in the medical domain. Despite the fact that there are many reasons for this, the need for standardization of protocols and terminology is critical, not only for the provision of uniform levels of health care, but also to facilitate medical science research. In the domain of skin lesions, the variability of semantic features contained within images is major among the barriers to the medical understanding of the symptoms and development of early skin cancers. Such variability is acknowledged across specialist fields of medicine, motivating standardization of terminologies for reporting medical practice. The desideratum of making these standards machine-readable has led to their formalization in the form of ontologies. Ontologies are computational artifacts designed to provide semantic representations of a particular domain of interest. Such a representation can be encoded and reused, allowing navigation of the key concepts recorded and retrieval of information indexed against it. This fact bridges the required standardization gap by offering a set of labeling options to record observations and events encountered by medical experts. Given the twin goals of ontologies - representation and standardization - the present paper deals with the creation of an ontology for skin lesion images, encoded using Description Logic in OWL format, which can be used for decision support systems that operate on a classification basis. The declarative framework within which ontologies are encoded can transcend any specific application context and help dermatologists perform semantic inference on skin lesion images. Furthermore, extensions are easier to be accomplished and new classes can be straightforwardly incorporated.

In the present paper, Random Forests are used in a critical and at the same time non trivial problem concerning the diagnosis of Gas Turbine blading faults, portraying promising results. Random forests-based fault diagnosis is treated as a Pattern Recognition problem, based on measurements and feature selection. Two different types of inserting randomness to the trees are studied, based on different theoretical assumptions. The classifier is compared against other Machine Learning algorithms such as Neural Networks, Classification and Regression Trees, Naive Bayes and K-Nearest Neighbor. The performance of the prediction model reaches a level of 97% in terms of precision and recall, improving the existing state-of-the-art levels achieved by Neural Networks by a factor of 1.5%–2%.

Abstract Attempting to automatically learn to identify verb complements from natural language corpora without the help of sophisticated linguistic resources like grammars, parsers or treebanks leads to a significant amount of noise in the data. In machine learning terms, where learning from examples is performed using class-labelled feature-value vectors, noise leads to an imbalanced set of vectors: assuming that the class label takes two values (in this work complement/non-complement), one class (complements) is heavily underrepresented in the data in comparison to the other. To overcome the drop in accuracy when predicting instances of the rare class due to this disproportion, we balance the learning data by applying one-sided sampling to the training corpus and thus by reducing the number of non-complement instances. This approach has been used in the past in several domains (image processing, medicine, etc) but not in natural language processing. For identifying the examples that are safe to remove, we use the value difference metric, which proves to be more suitable for nominal attributes like the ones this work deals with, unlike the Euclidean distance, which has been used traditionally in one-sided sampling. We experiment with different learning algorithms which have been widely used and their performance is well known to the machine learning community: Bayesian learners, instance-based learners and decision trees. Additionally we present and test a variation of Bayesian belief networks, the COr-BBN (Class-oriented Bayesian belief network). The performance improves up to 22% after balancing the dataset, reaching 73.7% f-measure for the complement class, having made use only a phrase chunker and basic morphological information for preprocessing.

The present paper describes a Bayesian network approach to information retrieval (IR) from natural language texts in Greek. The network structure provides an intuitive representation of uncertainty relationships and the embedded conditional probability table is used by inference algorithms in an attempt to identify documents that are relevant to the user's needs, expressed in the form of Boolean queries. Our research has been directed in constructing a probabilistic IR framework that focus on assisting users perform ad-hoc retrieval of Greek documents from the domain of economics. Furthermore, users can integrate feedback regarding the relevance of the retrieved documents in an attempt to improve performance on upcoming requests. Towards these goals, we have developed the Bayesian network IR system and tested it on several web corpora with different application domains. We have developed two different approaches with regard to the structure: a simple one, where the structure is manually provided, and an automated one, where data mining is used in order to extract the network's structure. Results have depicted satisfactory performance in terms of precision-recall curves.

For the present paper, a Bayesian probabilistic framework for the task of automatic acquisition of intonational phrase breaks was established. By considering two different conditional independence assumptions, the naïve Bayes and Bayesian networks approaches were regarded and evaluated against the CART algorithm, which has been previously used with success. A finite length window of minimal morphological and syntactic resources was incorporated, i.e. the POS label and the kind of phrase boundary, a novel syntactic feature that has not been applied to intonational phrase break detection before. This feature can be used in languages where syntactic parsers are not available and proves to be important, not only for the proposed Bayesian methodologies but for other algorithms, like CART. Trained on a 5500 word database, Bayesian networks proved to be the most effective in terms of precision (82,3%) and recall (77,2%) for predicting phrase breaks. 1.

Learning Bayesian Belief Networks (BBN) from corpora and incorporating the extracted inferring knowledge with a Support Vector Machines (SVM) classifier has been applied to character segmentation for unconstrained handwritten text. By taking advantage of the plethora in unlabeled data found in image databases in addition to some available labeled examples, we overcome the expensive task of annotating the whole set of training data and the performance of the character segmentation learner is increased. Apart from this approach, which has not previously used for this task, we have experimented with two well-known machine learning methods (Learning Vector Quantization and a simplified version of the Transformation-Based Learning theory). We argue that a classifier generated from BBN and SVM is well suited for learning to identify the correct segment boundaries. Empirical results will support this claim. Performance has been methodically evaluated using both English and Modem Greek corpora in order to determine the unbiased behaviour of the trained models. Limited training data are proved to endow with satisfactory results. We have been able to achieve precision exceeding 86%.

This paper deals with the use of Bayesian Belief Networks in order to improve the accuracy and training time of character segmentation for unconstrained handwritten text. Comparative experimental results have been evaluated against Naive Bayes classification, which is based on the assumption of the independence of the parameters and two additional previous commonly used methods. Results have depicted that obtaining the inferential dependencies of the training data, could lead to the reduction of the required training time and size by a factor of 55%. Moreover, the achieved accuracy in detecting segment boundaries exceeds 86% whereas limited training data are proved to endow with very satisfactory results.

Learning Bayesian belief networks (BBN) from corpora and support vector machines (SVM) have been applied to the automatic acquisition of verb subcategorization frames for Modern Greek. We are incorporating minimal linguistic resources, i.e. basic morphological tagging and phrase chunking, to demonstrate that verb subcategorization, which is of great significance for developing robust natural language human computer interaction systems, could be achieved using large corpora, without having any general-purpose, syntactic parser at all. In addition, apart from BBN and SVM, which have not previously used for this task, we have experimented with three well-known machine learning methods (feedforward backpropagation neural networks, learning vector quantization and decision tables), which are also being applied to the task of verb subcategorization frame defection for the first time. We argue that both BBN and SVM are well suited for learning to identify verb subcategorization frames. Empirical results will support this claim. Performance has been methodically evaluated using two different corpora types, one balanced and one domain-specific in order to determine the unbiased behaviour of the trained models. Limited training data are proved to endow with satisfactory results. We have been able to achieve precision exceeding 80% on the identification of subcategorization frames which were not known beforehand.