Yusuf Yaslan

Co-training with relevant random subspaces

We introduce the relevant random subspace Co-training (Rel-RASCO) algorithm which produces relevant random subspaces and then does semi-supervised ensemble learning using those subspaces and unlabeled data. Ensemble learning algorithms may benefit from diversity of classifiers used. However, for high dimensional data choosing subspaces randomly, as in RASCO (Random Subspace Method for Co-training, Wang et al. 2008 [5]) algorithm, may produce diverse but inaccurate classifiers. We produce relevant random subspaces by means of drawing features with probabilities proportional to their relevances measured by the mutual information between features and class labels. We show that Rel-RASCO achieves better accuracy by this relevant and random subspace selection scheme. Experiments on five real and one synthetic datasets show that Rel-RASCO algorithm outperforms both RASCO and Co-training in terms of the accuracy achieved at the end of Co-training.

Music genre classification using MIDI and audio features

We report our findings on using MIDI files and audio features from MIDI, separately and combined together, for MIDI music genre classification. We use McKay and Fujinagas 3-root and 9-leaf genre data set. In order to compute distances between MIDI pieces, we use normalized compression distance (NCD). NCD uses the compressed length of a string as an approximation to its Kolmogorov complexity and has previously been used for music genre and composer clustering. We convert the MIDI pieces to audio and then use the audio features to train different classifiers. MIDI and audio from MIDI classifiers alone achieve much smaller accuracies than those reported by McKay and Fujinaga who used not NCD but a number of domain-based MIDI features for their classification. Combining MIDI and audio from MIDI classifiers improves accuracy and gets closer to, but still worse, accuracies than McKay and Fujinagas. The best root genre accuracies achieved using MIDI, audio, and combination of them are 0.75, 0.86, and 0.93, respectively, compared to 0.98 of McKay and Fujinaga. Successful classifier combination requires diversity of the base classifiers. We achieve diversity through using certain number of seconds of the MIDI file, different sample rates and sizes for the audio file, and different classification algorithms.

Audio Music Genre Classification Using Different Classifiers and Feature Selection Methods

We examine performance of different classifiers on different audio feature sets to determine the genre of a given music piece. For each classifier, we also evaluate performances of feature sets obtained by dimensionality reduction methods. Finally, we experiment on increasing classification accuracy by combining different classifiers. Using a set of different classifiers, we first obtain a test genre classification accuracy of around 79.6 plusmn 4.2% on 10 genre set of 1000 music pieces. This performance is better than 71.1 plusmn 7.3% which is the best that has been reported on this data set. We also obtain 80% classification accuracy by using dimensionality reduction or combining different classifiers. We observe that the best feature set depends on the classifier used

An integrated decoding framework for audio watermark extraction

This paper proposes a blind audio watermark extraction technique that allows performing watermark decoding while installing data synchronization. The proposed decoding algorithm employs correlation techniques supported by a wavelet denoising process, thus improves the decoding performance significantly. A data adaptive nonlinear MPEG Layer 1 Model 1 compatible watermark encoder is designed for watermark embedding. A channel encoder is also included into the system to take the advantage of error correction. The method does not require the original audio for decoding and it is robust to channel noise, filtering as well as stereo-to-mono conversions. It allows working at very low watermark-to-signal ratios thus preserves inaudibility.

A hybrid method for time series prediction using EMD and SVR

Forecasting in several areas such as stock price, electricity power consumption, tourist arrival rates or capacity planning allows us to give decisions for future events. The rising up or falling down of the values can support researchers, economists or investors while giving their important decisions. This study aims to forecast the directional movements of electricity load demands and evaluates the performance on 3 load datasets. In experimental results, the proposed Empirical Mode Decomposition (EMD) and Support Vector Regression (SVR) based hybrid method is compared with single SVR. It is observed that the proposed EMD-SVR method outperforms the single SVR performance on direction measurements including Direction Accuracy, Correct Up and Correct Down trends.

A Genetic Programming Classifier Design Approach for Cell Images

This paper describes an approach for the use of genetic programming (GP) in classification problems and it is evaluated on the automatic classification problem of pollen cell images. In this work, a new reproduction scheme and a new fitness evaluation scheme are proposed as advanced techniques for GP classification applications. Also an effective set of pollen cell image features is defined for cell images. Experiments were performed on Bangor/Aberystwyth Pollen Image Database and the algorithm is evaluated on challenging test configurations. We reached at 96 % success rate on the average together with significant improvement in the speed of convergence.

Emotion recognition via random forest and galvanic skin response: Comparison of time based feature sets, window sizes and wavelet approaches

Emotions play a significant and powerful role in everyday life of human beings. Developing algorithms for computers to recognize emotional expression is a widely studied area. In this study, emotion recognition from Galvanic signals was performed using time domain and wavelet based features. Feature extraction has been done with various feature set attributes. Various length windows have been used for feature extraction. Various feature attribute sets have been implemented. Valence and arousal have been categorized and relationship between physiological signals and arousal and valence has been studied using Random Forest machine learning algorithm. We have achieved 71.53% and 71.04% accuracy rate for arousal and valence respectively by using only galvanic skin response signal. We have also showed that using convolution has positive affect on accuracy rate compared to non-overlapping window based feature extraction.

A comparison study on active learning integrated ensemble approaches in sentiment analysis

One of the most challenging problems of sentiment analysis on social media is that labelling huge amounts of instances can be very expensive. Active learning has been proposed to overcome this problem and to provide means for choosing the most useful training instances. In this study, we introduce active learning to a framework which is comprised of most popular base and ensemble approaches for sentiment analysis. In addition, the implemented framework contains two ensemble approaches, i.e. a probabilistic algorithm and a derived version of Behavior Knowledge Space (BKS) algorithm. The Shannon Entropy approach was utilized for choosing among training data during active learning process and it was compared with maximum disagreement method and random selection of instances. It was observed that the former method causes better accuracies in less number of iterations. The above methods were tested on Cornell movie review dataset and a popular multi-domain product review dataset.

Co-training with adaptive Bayesian classifier combination

In a classification problem, when there are multiple feature views and unlabeled examples, co-training can be used to train two separate classifiers, label the unlabeled data points iteratively and then combine the resulting classifiers. Especially when the number of labeled examples is small due to expense or difficulty of obtaining labels, co-training can improve classifier performance. For binary classification problems, mostly, the product rule has been used to combine classifier outputs. In this paper, we propose an adaptive Bayesian classifier combination method which selects either the Bayesian or the product combination method based on the belief values. We compare our adaptive Bayesian method with Bayesian, product and maximum classifier combination methods for the multi-class pollen image classification problem. Two different feature sets, Haralickpsilas texture features and features obtained using local linear transforms are used for co-training. Experimental results show that adaptive Bayesian combination with co-training performs better than the other three methods.

Robust audio watermarking by adaptive psychoacoustic masking

This paper proposes an adaptive audio watermarking algorithm that allows watermark embedding at low watermark-to-signal ratio levels. It is realized by modifying the MPEG Layer 1 Model 1 psychoacoustic masking technique in an iterative way, while preserving inaudibility. The developed watermarking algorithm does not require the original audio file at the decoding stage and it is robust to stereo-mono conversions, additive channel noise and changes in the sampling rate. Decoding performance of the proposed watermarking algorithm is compared to the traditional spread spectrum based methods as well as a DC level shifting method which performs time domain watermarking.