Sharif Digital Repository

Semi-structured documents are a new type of textual documents which has gained a lot of attention nowadays to itself. A specific document modeling for boosting classifiers is needed more than ever which reflects major document specifications. The main goal of this thesis is presenting new adaptive model based on semi-structured documents features. We also aim to use some problem solving approaches such as Statistical approach, Machine Learning and few Algorithmic solutions. In some cases 10% precision optimization can be seen compare to the best approaches available nowadays  

Cardiovascular diseases are very common nowadays and are one of the main reasons of death. Being among the major types of these diseases, correct and in time diagnosis of Coronary Artery Disease (CAD) is very important. The best and most accurate CAD diagnosis method by now is recognized as Angiography, which has many side effects and is costly. Thus researchers are seeking for inexpensive, though still accurate, methods. Existing studies have used several features in collecting data from patients, while applying different data mining algorithms to increase accuracy. In this thesis, a data set is introduced which utilizes several new and effective features for CAD diagnosis, as well as a... 

Computer-Aided Diagnosis (CAD) systems are widely used for detection of various kinds of abnormalities in mammography images. In this work, mass classification is investigated and its steps are explained in detail, for each step a main method is presented and other methods are also discussed. For mass segmentation a relatively new method based on level set and Morphological Component Analysis (MCA) is used.After this step, various kinds of features such as shape, geometrical, and textural ones are introduced. Moreover, a set of proposed features based on wavelet transformation,for this application are presented. The proposed features can describe margin and texture characterizations of a... 

Scene classification and understanding is one of the most important fields in computer vision. Its applications are such as exploring robot navigation enviroment, content-based image retrieval (CBIR), organization in image databases, highly semantic describing images and videos and content extraction of videos.Many methods and algorithm are proposed till today to deal with diversity of this field by emphesizing on feature based methods or machine learning based methods. In this research we have focoused on proposing a new algorithm which is using principals of NBNN image classification method but major changes in how to exract distance metric from Nearest neighbour and how to use local... 

Traffic density estimation is one of the most challenging problems in Intelligent Transportation Systems. One of the important traffic information that is broadcasted to drivers is Traffic Density information. In many traffic control centers; human operators are responsible for estimating traffic density from captured video data. Increasing traffic cameras and constraint number of operators introduce an updating delay to broadcasted information. So it is important to have an automatic traffic density estimation system. In this thesis, machine vision is used to solve this problem. Supervised Image classification is our approach. In supervised Image classification, images are classified to... 

Distance Metric Learning algorithms have been widely used in Machine Learning methods recently. In these algorithms a distance function between objecs (data points) is learned based on their labels or similarity and dissimilarity constraints. Recent works have shown that a good precision is obtained in classification or clustering methods which use these functions. Since in the current systems many of data points do not exist at the beginning and are added to the training set as the algorithm is run, online methods are needed to update learned metric due to new data.
In this thesis, we proposed a new online distance metric learning method that has higher performance than existing... 

Classification is one the contrivesial problems in machine vision and pattern recongnition. Traditional feature extraction methods which are based on low level feature extraction do not have high classification accuracy, thus they do not have the ability to represent images in feature space in discriminative way. In this thesis we have proposed a grid base method and used hidden Markov model (HMM) to include topological and spatial information in feature vectors. Then the classifiers created based on HMM feature extraction are combind. Combination of classifiers is based on designing a convex goal function. The goal of this optimization is to determine the wight of each classifier for... 

This paper presents some approaches based on data mining techniques to solve the prediction task of Computational Intelligence and Learning (CoIL) Challenge 2000. The prediction task of the contest is a direct mailing problem and the goal is to improve its response rate. The main issue in this competition is the incompatibility of the dataset in which the distribution of the classes of the target attribute is highly unbalanced. This in turn causes high error rate in identifying the minority class samples. Three different level methods including data-level, algorithm-level, and hybrid method are used to overcome this issue. The specificity and sensitivity criteria are employed to compare the... 

This thesis provids a Persian End-To-End Speech Recognition system. In this system, the input is low-level features of speech signal. Deep recurrent neural networks (RNNs) and Long Short-Term Memory (LSTM) units as the RNN building blocks are used as the acoustic model. Continuous speech data is labeled by the CTC which is applied as the output layer of a recurrent neural network. By using the CTC objective function, acoustic modeling problem is simplified to just an RNN learning problem over pairs of speech and context-independent (CI) label sequences. A distinctive feature of this system is a generalized decoding approach based on weighted finite-state transducers (WFSTs), which enables... 

The purpose of this project is to improve current speaker verification techniques with short utterance using phonetic information extraction. I-vector technique is widely used in speaker verification systems. Different speakers span a subspace of universal acoustic space, which is usually modeled by “Universal Background model”. Speaker-specific subspace depends on the voice of speaker. In state-of-the-art speaker verification systems i-vectors are extracted by a factor analysis technique to represent speaker characteristics. Studies demonstrate that voiced phonemes contain more speaker-specific information than unvoiced. In this thesis we have classified voiced frames in order to exploit... 

In recent years, neural networks have been widely used in most machine learning domains. However, it has been shown that these networks are vulnerable to adversarial examples. adversarial examples are small and imperceptible perturbations applied to the input which lead to producing wrong output and thus, fooling the network. This will become an important issue in security related applications of deep neural networks, such as self-driving cars and medical diagnostics. Since, in the wort-case scenario, even human lives could be threatened. Although, many works have focused on crafting adversarial examples for image data, only a few studies have been done on textual data due to the existing... 

The tree structure of neuron morphologies has excited neuroscientists since their discovery in the 19-th century. Many theories assign computational meaning to morphologies, but it is still hard to generate realistic looking morphologies. There are a few growth models for generating neuron morphologies that correctly reproduce some features (e.g. branching angles) of morphologies, but they tend to fall short on other features. Here we present an approach that builds a generative model by extracting a set of human-chosen features from a database of neurons by using the naïve Bayes approach. Then by starting from a neuron with a soma we use statistical sampling techniques to generate... 

In recent years, Machine Learning models have been developed in Signal Processing, Computer Vision and Neuroscience areas. There are two categories of Machine Learning models which are supervised and unsupervised learning models. Regression and classification problems are two popular problems examples of supervised learning models. From unsupervised learning problems, we can mention the clustering problem. Support Vector Regression (SVR), Decision Tree Regression and Bagging Ensemble Regression models are some important models of the regression problem. For classification problems, we can also mention to Support Vector Classification, Decision Tree Classification, and Bagging Ensemble... 

Epilepsy is one of the diseases that affects the quality of life of epileptic patients. Epileptic patients lose control during epileptic seizures and are more likely to face problems. Designing and creating a seizure detection system can reduce casualties from epileptic attacks. In this study, we present an automatic method that reduces the artifact from the raw signals, and then classifies the seizure and non-seizure epochs. At all stages, it is assumed that no information is available about the patient and this detection is made only based on the information of other patients. The data from this study were recorded in Temple Hospital and the recording conditions were not controlled, so... 

Nowadays, researchers working on classifiers which are designed to predict minority class. In this work, we attempt to improve fraud detection performance, with minimum possible complexity. In this regard, by incrementing model sensitivity to minority class samples, we solve the problem of model ignorance to these instances. Moreover, by using clustering, we cluster similar inputs based on their features, and split each class to smaller bins. Then with considering the fact that, prediction probability threshold influences the final performance, we define statistical hypothesis testing exclusively for each cluster to evaluate predictions with expected range. In this method, model is not... 

By expanding the use of IT and public access to financial markets, the number of players in this area has increased and the nonlinearity of the market has become more complex. Hence, investors need specific strategies that can make profitable investment by determining the time of purchase and sale of stocks. The purpose of this research is to provide a stock trading framework for strategic portfolio management. This framework uses daily values of 18 indicators of technical analysis as features and daily trading signals as data labels for training various machine learning models, such as support vector regression, k nearest neighbors, decision tree, artificial neural network and random... 

Mobile application marketplaces are not only a distribution platform but also a place for users to give feedback on their experience with application. User reviews contain useful information for software evolution tasks including bug reports, user experience, and feature requests. Considering the massive number of reviews that popular apps receive every day, manual inspection of reviews is not feasible in many cases. Researchers have developed automated tools to classify reviews into fixed and general-purpose categories related to software evolution in order to assist developers. Although this classification can reduce the time and effort for mobile developers, it does not consider the... 

The research in this study aims to analyze data in healthcare, especially the diagnosis of several diseases caused by heart failure. Analyzing and analyzing this data can lead to the discovery of relationships and patterns that can play an important role in the decision-making process of relevant officials in any field. Today, medical data around the world is stored in large volumes for future research. Various infrastructures and software have been set up in many health centers and research centers affiliated with those organizations.In this research, the general process of work is such that the data related to the heart sounds, which are in the four broad categories of S1 to S4, are... 

Human has different emotions such as happiness, sadness, anger, etc. Recognizing these emotions plays an important role in human-machine interface. Emotion recognition can be divided into approaches, physiological and non-physiological signals. Non-physiological signals include facial expressions, body gesture, and voice, and physiological signals include electroencephalograph (EEG), electrocardiograph (ECG), and functional magnetic resonance imaging (fMRI). EEG signal has been absorbed a lot of attention in emotion recognition because recording of EEG signal is easy and it is non-invasive. Analysis of connectivity and interaction between different areas of the brain can provide useful... 

Diffractive deep neural network is an optical machine-learning framework that uses diffractive surfaces, optical devices, electro-optic devices and engineered matterials to optically perform computational tasks. These diffractive networks, after their desing and train phase by computers and machine learning algorithms, are physically fabricated using 3D printing or lithography, to actualize the model of trained network. Machine learning processes and alghorithms are performed through light-matter interaction and diffraction of light. This procedure is done at the speed of light and without the need of any power, except for the light illumination for the input object. In comparison with...