Sharif Digital Repository

Deep learning algorithms combined with supervision rely heavily on labeled data, posing challenges in the data labeling process. Addressing this issue, researchers in the field of machine learning have focused on developing approaches to reduce the dependency on labeled data and improve the efficiency of data collection for labeling purposes. This thesis investigates the training of a classification model using data collected through a human-in-the-loop system. Notably, this research pioneers the application of active learning techniques to differentiate between political and non-political Persian tweets. The dataset introduced in this study is the sole available collection for this specific... 

Nowadays, methods based on deep learning networks are the most effective artificial in­ telligence methods. Although they have achieved success in various fields (such as machine vision and object recognition), practical and experimental cases show the fragility of deep learning networks against perturbations and unwanted changes of the input pattern. All these perturbations must be in a way that the main class of the perturbed input pattern can be rec­ ognized by human, but the network makes a mistake in recognizing its correct class. This thesis seeks a more accurate evaluation by designing adversarial attacks such that the main class of the adversarial pattern is detectable by human...