Sharif Digital Repository

The existing continual learning methods are mainly focused on fully-supervised scenarios and are still not able to take advantage of unlabeled data available in the environment. Some recent works tried to investigate semi-supervised continual learning (SSCL) settings in which the unlabeled data are available, but it is only from the same distribution as the labeled data. This assumption is still not general enough for real-world applications and restricts the utilization of unsupervised data. In this work, we introduce Open-Set Semi-Supervised Continual Learning (OSSCL), a more realistic semi-supervised continual learning setting in which out-of-distribution (OoD) unlabeled samples in the... 

Feeder reconfiguration is one of the most prominent tasks for loss reduction, reliability improvement and also congestion alleviation in distribution networks. Besides improving abovementioned operational issues, during a long period of time, proper switching in appropriate times can causes a lot of economic savings. Furthermore, it can be intensified when capacitors are implemented and controlled in the network. To reach this purpose, most of the previous literature focuses merely on the power grid reconfiguration. Few of these researches have considered this problem as a dynamic one and studied that in a period of time, e.g. one day, considering fixed level of loads despite the others that... 

Due to the importance of hepatitis disease, especially hepatitis C, from public health view point, this research aims to investigate and suggest an efficient medical therapy for this disease. First, available models are analyzed and the suitable one is chosen. Three controllers namely, fixed PI, adaptive PI and exact linearization control (GLC) are designed to control the hepatitis C disease using the drug efficacy as the manipulated variabel. Controller parameters are determined by the Genetic Algorithm (GA) and their performances are evaluated using the selected model. Uncertainties have been considerd in both model structure and parameters and controllers performances are investigated in... 

Quality and intelligibility are two aspects of speech that are affected by various factors, such as background noise and echo. The performance of many commercial and military speech-based systems depends on at least one of these aspects of speech. Therefore, this research aims to design an improvement model to remove background noise and reverberation from the speech signal. The model training framework is based on deep learning methods and has a supervised approach in the time domain. The input of this system is the raw waveform of the speech signal mixed with noise and reverberation, and the output is the enhanced waveform of this signal. An architecture is proposed in this thesis based on... 

Traffic classification is crucial for the execution of many security and managerial tasks in a network, such as identifying malicious data, detecting disruptive users, and preventing the passage of traffic from certain applications. Nowadays, a significant portion of traffic in computer networks is encrypted. Therefore, the classification of encrypted traffic requires the use of solutions for analyzing encrypted traffic. In recent years, many solutions based on machine learning and deep learning have been proposed for analyzing encrypted traffic. While most of these solutions focus on improving the accuracy of network traffic classification, less attention has been paid to their... 

In this study pain level estimation using facial expression is investigated. To do this, there are two approaches, one approach is sequence level pain estimation and the other one is frame level pain estimation. In sequence level, after feature extraction from all frames of sequence, each sequence is represented by a fixed length feature vector, this feature vector is constructed by concatenating min, max and mean of frame features of that specific sequence, then KLPP is applied in order to reduce feature vector dimension and in the end a linear regression is implemented to predict the pain labels of the sequence. In the frame level, two approaches are introduced, the first one is based on... 

In this thesis, a heterogeneous nanocatalyst based on graphene oxide was synthesized. To improve the graphene oxide, it was reacted with paraaminobenzoic acid and then metal salts (cobalt and manganese) were coated on it. The structure of the catalyst was confirmed by field emission scanning electron microscopy (FESEM), X-ray diffraction (XRD), quantitative X-ray diffraction (EDS) and infrared (FT-IR) spectroscopy. Finally, the catalytic properties of the catalysts were evaluated in the oxidation reaction of different alcohols in the presence of oxygen. In the presence of benzyl alcohols with lethal electron groups, the oxidation conditions are stricter and the reaction efficiency is lower.... 

Partial oxidation of methane (POX) is one of the most common processes to produce synthesis gas. In this study the kinetics of this process is examined and some reduced mechanisms are introduced to model it in this study. The experimental data from Research Institute of Petroleum Industry are used as reference experimental data. Inlets of this reactor are oxygen and methane with equivalence ratio between 2.53 and 2.9. Moreover, the experimental results of previous studies with the air as the oxidant in equivalence ration between 2.4 and 2.6 are used as basis for comparison of simulation results. During the first step, partial oxidation of methane in non-catalytic porous media is simulated by...