Sharif Digital Repository

Accurate prediction of pavement performance is essential to a pavement infrastructure management system. Selection of the prediction model is based on the extent of available data, assumptions used in performance modeling, ease of use and management purposes. Therefore, two methods were proposed in this thesis based on the assumptions of availability and unavailibility of accurate data. The first method presents a two-stage model to classify and accurately predict the performance of a pavement infrastructure system. Sections with similar characteristics are classified into groups using a support vector classifier (SVC). Then, a recurrent neural network (RNN) is utilized to predict... 

Path planning of a robot inside an environment with obstacles is to determine an appropriate path for moving from an initial point to a destination without colliding the obstacles. The main considerations in selecting such a path are its length and simplicity in terms of links or turn angles. In this paper, we study this problem for a point robot in the plane and our goal is to minimize the path length. We solve this problem by converting it to an optimization problem and solving the resulted optimization problem by a recurrent neural network. According to the implementation results, the obtained path is a proper approximation of the minimum length path, especially when obstacles are not too... 

Iran is regarded as a dry land and has suffered from extreme to severe drought conditions in recent years. Drought – which is mainly caused by shortage in rainfall – affects the normal life in Iran. Development of tools for effectively monitoring and predicting drought intensity might help the policy makers to reduce the vulnerability of the areas affected by drought. In this thesis, we showed that the intensity of drought can be predicted using satellite imagery data and recurrent neural networks. To this end, the standardized precipitation index (SPI) was chosen as an index for drought and normalized deviation of vegetation index (NDVI) as a remote sensing measure extracted from NOAA-AVHRR... 

This thesis examines the differences and the similarities between the two famous RNN blocks the Long Short Term Memory and the Gated Recurrent Unit. It measure different aspects such as computational complexity, Word Error Rate, and subjective human evaluation in the task of text generation.In the computational complexity experiment results show that the LSTM takes more time to compute, in comparison to the GRU. Moving on into the next experiment the GRU slightly outperforms the LSTM in terms of WER but the perplexity for the language models tested was the same. This shows that slight differences in the perplexity does not drastically change the WER. Having said, the results suggest that the... 

Learning is an essential mechanism for the survival of living things. There are different types of learning, and value learning is among the most important types. A child learns that water resolves the thirst need by repeatedly experiencing this situation. Eventually, the value of water, which has been valueless before that, increases gradually in his mind. How this concept is encoded in the brain? previous works reveal the role of different neurons and regions that are relevant to value learning. However, population analysis and dynamic modeling are less considered. Moreover, the links between different brain regions are unknown.Finding the relationship between two relevant regions of the... 

The importance of using a fast and automatic method to classify variable stars for large amounts of data is undeniable. There have been many attempts for classifying variable stars by traditional algorithms, which require long pre-processing time. In recent years, neural networks as classifiers have come to notice. This thesis proposes the Hierarchical Classification technique, which contains several models with the same network structure. Our pre-processing method produces input data by using light curves and the period. We use OGLE-IV variable stars database to train and test the performance of Convolutional Neural Networks based on the Hierarchical Classification technique. We see that... 

Action recognition is one of the important tasks in computer vision that detects the action label in videos that contain only one action. In recent years, action recognition has attracted much attention and researchers have tried to solve it by different approaches.Action recognition by itself does not have many applications in the real world because videos are untrimmed and do not contain only one action. So Temporal Action Localization(TAL) task in which we want to predict the start and end time of each action in addition to the action label has a lot of applications in the real world and for this reason, TAL is a hot research topic. But due to its complexity, researchers have not reached... 

Indoor positioning plays a pivotal role in a wide range of applications, from smart homes to industrial automation. In this thesis, we propose a comprehensive approach for accurate positioning in indoor environments through the integration of existing Wi-Fi and Bluetooth Low Energy (BLE) devices. The proposed algorithm involves acquiring the received signal strength indicator (RSSI) data from these devices and capturing the complex interactions between RSSI and positions. To enhance the accuracy of the collected data, we first use a Kalman filter for denoising RSSI values, then categorize them into distinct classes using the K-nearest neighbor (KNN) algorithm. Incorporating the filtered...