Stamatis Karlos

Speaker Identification Using Semi-supervised Learning

Semi-supervised classification methods use available unlabeled data, along with a small set of labeled examples, to increase the classification accuracy in comparison with training a supervised method using only the labeled data. In this work, a new semi-supervised method for speaker identification is presented. We present a comparison with other well-known semi-supervised and supervised classification methods on benchmark datasets and verify that the presented technique exhibits better accuracy in most cases.

Self-trained LMT for semisupervised learning

The most important asset of semisupervised classification methods is the use of available unlabeled data combined with a clearly smaller set of labeled examples, so as to increase the classification accuracy compared with the default procedure of supervised methods, which on the other hand use only the labeled data during the training phase. Both the absence of automated mechanisms that produce labeled data and the high cost of needed human effort for completing the procedure of labelization in several scientific domains rise the need for semisupervised methods which counterbalance this phenomenon. In this work, a self-trained Logistic Model Trees (LMT) algorithm is presented, which combines the characteristics of Logistic Trees under the scenario of poor available labeled data. We performed an in depth comparison with other well-known semisupervised classification methods on standard benchmark datasets and we finally reached to the point that the presented technique had better accuracy in most cases.

Self-trained Rotation Forest for semi-supervised learning

The most important asset of semi-supervised learning (SSL) methods is the use of available unlabeled data combined with an enough smaller set of labeled examples, so as to increase the classification accuracy compared with the default procedure of supervised methods, in which during the training only the labeled data are used. The encapsulation of classifier ensembles that produce different models through training process into semi-supervised schemes seems to be a promising strategy for enhanced learning ability. In this work, a Self-trained Rotation Forest (Self-RotF) algorithm and a variant of this (Weighted-Self-RotF) are presented. We performed an in depth comparison with other well-known semi- supervised classification methods on standard benchmark datasets and after having tested their performance with statistical tests, we finally reached to the point that the presented technique had better accuracy in most cases.

Self-Trained Stacking Model for Semi-Supervised Learning

The most important characteristic of semi-supervised learning methods is the combination of available unlabeled data along with an enough smaller set of labeled examples, so as to increase the learning accuracy compared with the default procedure of supervised methods, which on the other hand use only the labeled data during the training phase. In this work, we have implemented a hybrid Self-trained system that combines a Support Vector Machine, a Decision Tree, a Lazy Learner and a Bayesian algorithm using a Stacking variant methodology. We performed an in depth comparison with other well-known Semi-Supervised classification methods on standard benchmark datasets and we finally reached to the point that the presented technique had better accuracy in most cases.

Locally application of naive Bayes for self-training

Semi-supervised algorithms are well-known for their ability to combine both supervised and unsupervised strategies for optimizing their learning ability under the assumption that only a few examples together with their full feature set are given. In such cases, the use of weak learners as base classifiers is usually preferred, since the iterative behavior of semi-supervised schemes require the building of new temporal models during each new iteration. Locally weighted naïve Bayes classifier is such a classifier that encompasses the power of NB and k-NN algorithms. In this work, we have implemented a self-labeled weighted variant of local learner which uses NB as the base classifier of self-training scheme. We performed an in depth comparison with other well-known semi-supervised classification methods on standard benchmark datasets and we reached to the conclusion that the presented technique had better accuracy in most cases.

Semi-supervised forecasting of fraudulent financial statements

Prediction of potential fraudulent activities may prevent both the stakeholders and the appropriate regulatory authorities of national or international level from being deceived. The objective difficulties on collecting adequate data that are obsessed by completeness affects the reliability of the most supervised Machine Learning methods. This work examines the effectiveness of forecasting fraudulent financial statements using semi-supervised classification techniques (SSC) that require just a few labeled examples for achieving robust learning behaviors mining useful data patterns from a larger pool of unlabeled examples. Based on data extracted from Greek firms, a number of comparisons between supervised and semi-supervised algorithms has been conducted. According to the produced results, the later algorithms are favored being examined over several scenarios of different Labeled Ratio (R) values.

Automated hand gesture recognition for educational applications

Introduction of new schemes for recognizing hand gestures is an active research field related with these of Machine Learning (ML) and Human Computer Interaction (HCI). Numerous applications can be integrated with such automated algorithms for facilitating either more advanced or less experienced users inside their working or educational environment. A simple method for hand gesture recognition tasks is presented here, using effective enough methods for extracting the final training data and Random Forest learner for prediction. The contribution of this work is the low-dimensionality of the final dataset, based on well-known feature metrics and the high level of accuracy that is achieved using cheap enough equipment, being affordable both for personal use and for more academic activities or online courses.

Speech Recognition Combining MFCCs and Image Features

Automatic speech recognition (ASR) task constitutes a well-known issue among fields like Natural Language Processing (NLP), Digital Signal Processing (DSP) and Machine Learning (ML). In this work, a robust supervised classification model is presented (MFCCs + autocor + SVM) for feature extraction of solo speech signals. Mel Frequency Cepstral Coefficients (MFCCs) are exploited combined with Content Based Image Retrieval (CBIR) features extracted from spectrogram produced by each frame of the speech signal. Improvement of classification accuracy using such extended feature vectors is examined against using only MFCCs with several classifiers for three scenarios of different number of speakers.

Effectiveness of semi-supervised learning in bankruptcy prediction

Adoption of techniques from fields related with Data Science, such as Machine Learning, Data Mining and Predictive Analysis, in the task of bankruptcy prediction can produce useful knowledge for both the policy makers and the organizations that are already funding or are interested in acting towards this direction in the near future. The nature of this task prevents analysts from collecting large amount of data for building accurate predictive models. Semi-supervised algorithms overcome this phenomenon and can perform robust behavior based on a few data. Experiments using data from Greek firms have been made in this work, comparing many semi-supervised schemes against well-known supervised algorithms and the results are promising.

Optimized Active Learning Strategy for Audiovisual Speaker Recognition

The purpose of this work is to investigate the improved recognition accuracy caused from exploiting optimization stages for tuning parameters of an Active Learning (AL) classifier. Since plenty of data could be available during Speaker Recognition (SR) tasks, the AL concept, which incorporates human entities inside its learning kernel for exploring hidden insights into unlabeled data, seems extremely suitable, without demanding much expertise on behalf of the human factor. Six datasets containing 8 and 16 speakers’ utterances under different recording setups, are described by audiovisual features and evaluated through the time-efficient Uncertainty Sampling query strategy (UncS). Both Support Vector Machines (SVMs) and Random Forest (RF) algorithms were selected to be tuned over a small subset of the initial training data and then applied iteratively for mining the most suitable instances from a corresponding pool of unlabeled instances. Useful conclusions are drawn concerning the values of the selected parameters, allowing future optimization attempts to get employed into more restricted regions, while remarkable improvements rates were obtained using an ideal annotator.