[pt] Este trabalho é um estudo sobre o desempenho das principais versões do protocolo TCP e de sua variação denominada STP na transmissão via satélite. Devido à grande distância da terra aos satélites geoestacionários, a transmissão via satélite se caracteriza por um grande atraso de propagação e maior ocorrência de erros em comparação com a transmissão ótica. Estes problemas causam sério impacto no desempenho do TCP pois eventuais perdas de pacote implicam em retransmissão e demorada reinicialização do protocolo. Neste estudo, abordamos especificamente o efeito dos erros no canal como fonte de perdas de pacote, considerando modelos de erros estatisticamente independentes e erros em surto. Este tipo de erro ocorre geralmente pela utilização de códigos convolucionais associados ao decodificador de Viterbi. Através de simulação em computador usando o software Matlab, foram inicialmente caracterizadas as seqüências de erros geradas por alguns codificadores convolucionais típicos. Em seguida, o efeito destas seqüências foi introduzida em simulações de uma aplicação FTP/TCP em canal satélite, comparado-se a vazão obtida com as diferentes versões do TCP. / [en] This work is a study on the performance of the main versions of protocol TCP and of its variation STP satellite transmission. Due to great distance of the land to the geostationary satellites, the satellite transmission is characterizes for a great delay of propagation and greater occurrence of errors in comparison with the optics transmission. These problems cause serious impact in the performance of the TCP therefore eventual losses of package imply in retransmission and delayed reset of the protocol. In this study, we specifically approach the effect of the errors in the canal as source of losses of package, considering models of statistically independent errors and burst errors. This type of error generally occurs for the use of convolutional code associates with Viterbi decoder. Through simulation in computer using Matlab software, we initially had been characterized the sequences of errors generated for some typical convolutional codes. After that, the effect of these sequences was introduced in simulation of an application FTP/TCP in canal satellite, compared it outflow gotten with the different versions of the TCP, using ns-2.

Major advances in machine learning have made image recognition applications, with Artificial Neural Network, blossom over the recent years. The aim of this thesis was to find a solution to recognize digits from a number sequence on an ID tag, used to identify farm animals, with the help of image recognition. A Recurrent Convolutional Neural Network solution called PPNet was proposed and tested on a data set called Animal Identification Tags. A transfer learning method was also used to test if it could help PPNet generalize and better recognize digits. PPNet was then compared against Microsoft Azures own image recognition API, to determine how PPNet compares to a general solution. PPNet, while not performing as good, still managed to achieve competitive results to the Azure API.

Understanding road surface conditions is an important component in active vehicle safety. Estimations can be achieved through image classification using increasingly popular convolutional neural networks (CNNs). In this paper, we explore the effects of multi-task learning by creating CNNs capable of simultaneously performing the two tasks road surface condition classification (RSCC) and road scene semantic segmentation (RSSS). A multi-task network, containing a shared feature extractor (VGG16, ResNet-18, ResNet-101) and two taskspecific network branches, is built and trained using the Road-Conditions and Cityscapes datasets. We reveal that utilizing task-dependent homoscedastic uncertainty in the learning process improvesmulti-task model performance on both tasks. When performing task adaptation, using a small set of additional data labeled with semantic information, we gain considerable RSCC improvements on complex models. Furthermore, we demonstrate increased model generalizability in multi-task models, with up to 12% higher F1-score compared to single-task models.

Most research behind the use of Machine Learning models in the field of Human Activity Recognition focuses mainly on the classification of daily human activities and aerobic exercises. In this study, we focus on the use of 1 accelerometer and 2 gyroscope sensors to build a Deep Learning classifier to recognise 5 different strength exercises, as well as a null class. The strength exercises tested in this research are as followed: Bench press, bent row, deadlift, lateral rises and overhead press. The null class contains recordings of daily activities, such as sitting or walking around the house. The model used in this paper consists on the creation of consecutive overlapping fixed length sliding windows for each exercise, which are processed separately and act as the input for a Deep Convolutional Neural Network. In this study we compare different sliding windows lengths and overlap percentages (step sizes) to obtain the optimal window length and overlap percentage combination. Furthermore, we explore the accuracy results between 1D and 2D Convolutional Neural Networks. Cross validation is also used to check the overall accuracy of the classifiers, where the database used in this paper contains 5 exercises performed by 3 different users and a null class. Overall the models were found to perform accurately for window’s with length of 0.5 seconds or greater and provided a solid foundation to move forward in the creation of a more robust fully integrated model that can recognize a wider variety of exercises.

<div>Convolutional neural networks lie at the heart of nearly every object recognition system today. While their performance continues to improve through new architectures and techniques, some of their deciencies have not been fully addressed to date. Two of these deciencies are their inability to distinguish the spatial relationships between features taken from the data, as well as their need for a vast amount of training data. Capsule networks, a new type of convolutional neural network, were designed specically to address these two issues. In this work, several capsule network architectures are utilized to classify objects taken from overhead satellite imagery. These architectures are trained and tested on small datasets that were constructed from the xView dataset, a comprehensive collection of satellite images originally compiled for the task of object detection. Since the objects in overhead satellite imagery are taken from the same viewpoint, the transformations exhibited within each individual object class consist primarily of rotations and translations. These spatial relationships are exploited by capsule networks. As a result it is shown that capsule networks achieve considerably higher accuracy when classifying images from these constructed datasets than a traditional convolutional neural network of approximately the same complexity.</div>

Two dimensional (2D) materials with interesting fundamental physics and potential applications attract tremendous efforts to study. The versatile properties of 2D materials can be further tailored by tuning the electronic structure with the layer-stacking arrangement, of which the main adjustable parameters include the thickness and the in-plane twist angle between layers. The Angle-Resolved Photoemission Spectroscopy (ARPES) has become a canonical tool to study the electronic structure of crystalline materials. The recent development of ARPES with sub-micrometre spatial resolution (micro-ARPES) has made it possible to study the electronic structure of materials with mesoscopic domains. In this thesis, we use micro-ARPES to investigate the spatially-resolved electronic structure of a series of few-layer materials: 1. We explore the electronic structure of the domains with different number of layers in few-layer graphene on copper substrate. We observe a layer- dependent substrate doping effect in which the Fermi surface of graphene shifts with the increase of number of layers, which is then explained by a multilayer effective capacitor model. 2. We systematically study the twist angle evolution of the energy band of twisted few-layer graphene over a wide range of twist angles (from 5° to 31°). We directly observe van Hove Singularities (vHSs) in twisted bilayer graphene with wide tunable energy range over 2 eV. In addition, the formation of multiple vHSs (at different binding energies) is observed in trilayer graphene. The large tuning range of vHS binding energy in twisted few-layer graphene provides a promising material base for optoelectrical applications with broad-band wavelength selectivity. 3. To better extract the energy band features from ARPES data, we propose a new method with a convolutional neural network (CNN) that achieves comparable or better results than traditional derivative based methods. Besides ARPES study, this thesis also includes the study of surface reconstruction for the layered material Bi2O2Se with the analysis of Scanning Tunnelling Microscopy (STM) images. To explain the origin of the pattern, we propose a tile model that produces the identical statistics with the experiment.

Dialectal Arabic speech research suffers from the lack of labelled resources and standardised orthography. There are three main challenges in dialectal Arabic speech recognition: (i) finding labelled dialectal Arabic speech data, (ii) training robust dialectal speech recognition models from limited labelled data and (iii) evaluating speech recognition for dialects with no orthographic rules. This thesis is concerned with the following three contributions: Arabic Dialect Identification: We are mainly dealing with Arabic speech without prior knowledge of the spoken dialect. Arabic dialects could be sufficiently diverse to the extent that one can argue that they are different languages rather than dialects of the same language. We have two contributions: First, we use crowdsourcing to annotate a multi-dialectal speech corpus collected from Al Jazeera TV channel. We obtained utterance level dialect labels for 57 hours of high-quality consisting of four major varieties of dialectal Arabic (DA), comprised of Egyptian, Levantine, Gulf or Arabic peninsula, North African or Moroccan from almost 1,000 hours. Second, we build an Arabic dialect identification (ADI) system. We explored two main groups of features, namely acoustic features and linguistic features. For the linguistic features, we look at a wide range of features, addressing words, characters and phonemes. With respect to acoustic features, we look at raw features such as mel-frequency cepstral coefficients combined with shifted delta cepstra (MFCC-SDC), bottleneck features and the i-vector as a latent variable. We studied both generative and discriminative classifiers, in addition to deep learning approaches, namely deep neural network (DNN) and convolutional neural network (CNN). In our work, we propose Arabic as a five class dialect challenge comprising of the previously mentioned four dialects as well as modern standard Arabic. Arabic Speech Recognition: We introduce our effort in building Arabic automatic speech recognition (ASR) and we create an open research community to advance it. This section has two main goals: First, creating a framework for Arabic ASR that is publicly available for research. We address our effort in building two multi-genre broadcast (MGB) challenges. MGB-2 focuses on broadcast news using more than 1,200 hours of speech and 130M words of text collected from the broadcast domain. MGB-3, however, focuses on dialectal multi-genre data with limited non-orthographic speech collected from YouTube, with special attention paid to transfer learning. Second, building a robust Arabic ASR system and reporting a competitive word error rate (WER) to use it as a potential benchmark to advance the state of the art in Arabic ASR. Our overall system is a combination of five acoustic models (AM): unidirectional long short term memory (LSTM), bidirectional LSTM (BLSTM), time delay neural network (TDNN), TDNN layers along with LSTM layers (TDNN-LSTM) and finally TDNN layers followed by BLSTM layers (TDNN-BLSTM). The AM is trained using purely sequence trained neural networks lattice-free maximum mutual information (LFMMI). The generated lattices are rescored using a four-gram language model (LM) and a recurrent neural network with maximum entropy (RNNME) LM. Our official WER is 13%, which has the lowest WER reported on this task. Evaluation: The third part of the thesis addresses our effort in evaluating dialectal speech with no orthographic rules. Our methods learn from multiple transcribers and align the speech hypothesis to overcome the non-orthographic aspects. Our multi-reference WER (MR-WER) approach is similar to the BLEU score used in machine translation (MT). We have also automated this process by learning different spelling variants from Twitter data. We mine automatically from a huge collection of tweets in an unsupervised fashion to build more than 11M n-to-m lexical pairs, and we propose a new evaluation metric: dialectal WER (WERd). Finally, we tried to estimate the word error rate (e-WER) with no reference transcription using decoding and language features. We show that our word error rate estimation is robust for many scenarios with and without the decoding features.

In this work, we develop an end-to-end neural network-based computer vision system to automatically identify where each person within a 2-D image of a school classroom is looking (“gaze following�), as well as who she/he is looking at. Automatic gaze following could help facilitate data-mining of large datasets of classroom observation videos that are collected routinely in schools around the world in order to understand social interactions between teachers and students. Our network is based on the architecture by Recasens, et al. (2015) but is extended to (1) predict not only where, but who the person is looking at; and (2) predict whether each person is looking at a target inside or outside the image. Since our focus is on classroom observation videos, we collect gaze dataset (48,907 gaze annotations over 2,263 classroom images) for students and teachers in classrooms. Results of our experiments indicate that the proposed neural network can estimate the gaze target - either the spatial location or the face of a person - with substantially higher accuracy compared to several baselines.

Juggling can be both a recreational activity that provides a wide variety of challenges to participants and an art form that can be performed on stage. Non-learning-based computer vision techniques, depth sensors, and accelerometers have been used in the past to augment these activities. These solutions either require specialized hardware or only work in a very limited set of environments. In this project, a 54 000 frame large video dataset of annotated juggling was created and a convolutional neural network was successfully trained that could locate the balls and hands with high accuracy in a variety of environments. The network was sufficiently light-weight to provide real-time inference on CPUs. In addition, the locations of the balls and hands were recorded for thirty-six common juggling pattern, and small neural networks were trained that could categorize them almost perfectly. By building on the publicly available code, models and datasets that this project has produced jugglers will be able to create interactive juggling games for beginners and novel audio-visual enhancements for live performances.

Knowledge Graph is a rising topic in the field of Artificial Intelligence. As the current trend of knowledge representation, Knowledge graph research is utilizing the large knowledge base freely available on the internet. Knowledge graph also allows inspection, analysis, the reasoning of all knowledge in reality. To enable the ambitious idea of modeling the knowledge of the world, different theory and implementation emerges. Nowadays, we have the opportunity to use freely available information from Wikipedia and Wikidata. The thesis investigates and formulates a theory about learning from Knowledge Graph. The thesis researches probabilistic knowledge graph. It only focuses on a branch called latent feature models in learning probabilistic knowledge graph. These models aim to predict possible relationships of connected entities and relations. There are many models for such a task. The metrics and training process is detailed described and improved in the thesis work. The efficiency and correctness enable us to build a more complex model with confidence. The thesis also covers possible problems in finding and proposes future work.