Matrix factorization is at the heart of many machine learning algorithms, for example, dimensionality reduction (e.g. kernel PCA) or recommender systems relying on collaborative filtering. Understanding a singular value decomposition (SVD) of a matrix as a neural network optimization problem enables us to decompose large matrices efficiently while dealing naturally with missing values in the given matrix. But most importantly, it allows us to learn the connection between data points' feature vectors and the matrix containing information about their pairwise relations. In this paper we introduce a novel neural network architecture termed Similarity Encoder (SimEc), which is designed to simultaneously factorize a given target matrix while also learning the mapping to project the data points' feature vectors into a similarity preserving embedding space. This makes it possible to, for example, easily compute out-of-sample solutions for new data points. Additionally, we demonstrate that SimEc can preserve non-metric similarities and even predict multiple pairwise relations between data points at once.

Nowadays, recommender systems are present in many daily activities such as online shopping, browsing social networks, etc. Given the rising demand for reinvigoration of the tourist industry through information technology, recommenders have been included into tourism websites such as Expedia, Booking or Tripadvisor, among others. Furthermore, the amount of scientific papers related to recommender systems for tourism is on solid and continuous growth since 2004. Much of this growth is due to social networks that, besides to offer researchers the possibility of using a great mass of available and constantly updated data, they also enable the recommendation systems to become more personalised, effective and natural. This paper reviews and analyses many research publications focusing on tourism recommender systems that use social networks in their projects. We detail their main characteristics, like which social networks are exploited, which data is extracted, the applied recommendation techniques, the methods of evaluation, etc. Through a comprehensive literature review, we aim to collaborate with the future recommender systems, by giving some clear classifications and descriptions of the current tourism recommender systems.

The last two decades witnessed tremendous advances in the Information and Communications Technologies. Beside improvements in computational power and storage capacity, communication networks carry nowadays an amount of data which was not envisaged only few years ago. Together with their pervasiveness, network complexity increased at the same pace, leaving operators and researchers with few instruments to understand what happens in the networks, and, on the global scale, on the Internet. Fortunately, recent advances in data science and machine learning come to the rescue of network analysts, and allow analyses with a level of complexity and spatial/temporal scope not possible only 10 years ago. In my thesis, I take the perspective of an Internet Service Provider (ISP), and illustrate challenges and possibilities of analyzing the traffic coming from modern operational networks. I make use of big data and machine learning algorithms, and apply them to datasets coming from passive measurements of ISP and University Campus networks. The marriage between data science and network measurements is complicated by the complexity of machine learning algorithms, and by the intrinsic multi-dimensionality and variability of this kind of data. As such, my work proposes and evaluates novel techniques, inspired from popular machine learning approaches, but carefully tailored to operate with network traffic.

Over the past 15 years, the volume, richness and quality of data collected from the combined social networking platforms has increased beyond all expectation, providing researchers from a variety of disciplines to use it in their research. Perhaps more impactfully, it has provided the foundation for a range of new products and services, transforming industries such as advertising and marketing, as well as bringing the challenges of sharing personal data into the public consciousness. But how to make sense of the ever-increasing volume of big social data so that we can better understand and improve the user experience in increasingly complex, data-driven digital systems. This link with usability and the user experience of data-driven system bridges into the wider field of HCI, attracting interdisciplinary researchers as we see the demand for consumer technologies, software and systems, as well as the integration of social networks into our everyday lives. The fact that the data largely posted on social networks tends to be textual, provides a further link to linguistics, psychology and psycholinguistics to better understand the relationship between human behaviours offline and online. In this thesis, we present a novel conceptual framework based on a complex digital system using collected longitudinal datasets to predict system status based on the personality traits and emotions extracted from text posted by users. The system framework was built using a dataset collected from an online scholarship system in which 2000 students had their digital behaviour and social network behaviour collected for this study. We contextualise this research project with a wider review and critical analysis of the current psycholinguistics, artificial intelligence and human-computer interaction literature, which reveals a gap of mapping and understanding digital profiling against system status.

We introduce $\textit{semi-unsupervised learning}$, an extreme case of semi-supervised learning with ultra-sparse categorisation where some classes have no labels in the training set. That is, in the training data some classes are sparsely labelled and other classes appear only as unlabelled data. Many real-world datasets are conceivably of this type. We demonstrate that effective learning in this regime is only possible when a model is capable of capturing both semi-supervised and unsupervised learning. We develop two deep generative models for classification in this regime that extend previous deep generative models designed for semi-supervised learning. By changing their probabilistic structure to contain a mixture of Gaussians in their continuous latent space, these new models can learn in both unsupervised and semi-unsupervised paradigms. We demonstrate their performance both for semi-unsupervised and unsupervised learning on various standard datasets. We show that our models can learn in an semi-unsupervised manner on Fashion-MNIST. Here we artificially mask out all labels for half of the classes of data and keep $2\%$ of labels for the remaining classes. Our model is able to learn effectively, obtaining a trained classifier with $(77.2\pm1.3)\%$ test set accuracy. We also can train on Fashion-MNIST unsupervised, obtaining $(75.2\pm1.5)\%$ test set accuracy. Additionally, doing the same for MNIST unsupervised we get $(96.3\pm0.9)\%$ test set accuracy, which is state-of-the art for fully probabilistic deep generative models.

The social media revolution has changed the way that brands interact with consumers. Instead of spending their advertising budget on interstate billboards, more and more companies are choosing to partner with so-called Internet "influencers" --- individuals who have gained a loyal following on online platforms for the high quality of the content they post. Unfortunately, it's not always easy for small brands to find the right influencer: someone who aligns with their corporate image and has not yet grown in popularity to the point of unaffordability. In this paper we sought to develop a system for brand-influencer matchmaking, harnessing the power and flexibility of modern machine learning techniques. The result is an algorithm that can predict the most fruitful brand-influencer partnerships based on the similarity of the content they post.

Electrocardiographic signal is a subject to multiple noises, caused by various factors. It is therefore a standard practice to denoise such signal before further analysis. With advances of new branch of machine learning, called deep learning, new methods are available that promises state-of-the-art performance for this task. We present a novel approach to denoise electrocardiographic signals with deep recurrent denoising neural networks. We utilize a transfer learning technique by pretraining the network using synthetic data, generated by a dynamic ECG model, and fine-tuning it with a real data. We also investigate the impact of the synthetic training data on the network performance on real signals. The proposed method was tested on a real dataset with varying amount of noise. The results indicate that four-layer deep recurrent neural network can outperform reference methods for heavily noised signal. Moreover, networks pretrained with synthetic data seem to have better results than network trained with real data only. We show that it is possible to create state-of-the art denoising neural network that, pretrained on artificial data, can perform exceptionally well on real ECG signals after proper fine-tuning.

One of the key limitations of modern deep learning based approaches lies in the amount of data required to train them. Humans, on the other hand, can learn to recognize novel categories from just a few examples. Instrumental to this rapid learning ability is the compositional structure of concept representations in the human brain - something that deep learning models are lacking. In this work we make a step towards bridging this gap between human and machine learning by introducing a simple regularization technique that allows the learned representation to be decomposable into parts. We evaluate the proposed approach on three datasets: CUB-200-2011, SUN397, and ImageNet, and demonstrate that our compositional representations require fewer examples to learn classifiers for novel categories, outperforming state-of-the-art few-shot learning approaches by a significant margin.

Which generative model is the most suitable for Continual Learning? This paper aims at evaluating and comparing generative models on disjoint sequential image generation tasks. We investigate how several models learn and forget, considering various strategies: rehearsal, regularization, generative replay and fine-tuning. We used two quantitative metrics to estimate the generation quality and memory ability. We experiment with sequential tasks on three commonly used benchmarks for Continual Learning (MNIST, Fashion MNIST and CIFAR10). We found that among all models, the original GAN performs best and among Continual Learning strategies, generative replay outperforms all other methods. Even if we found satisfactory combinations on MNIST and Fashion MNIST, training generative models sequentially on CIFAR10 is particularly instable, and remains a challenge. Our code is available online \footnote{\url{https://github.com/TLESORT/Generative\_Continual\_Learning}}.

Latent factor models for recommender systems represent users and items as low dimensional vectors. Privacy risks of such systems have previously been studied mostly in the context of recovery of personal information in the form of usage records from the training data. However, the user representations themselves may be used together with external data to recover private user information such as gender and age. In this paper we show that user vectors calculated by a common recommender system can be exploited in this way. We propose the privacy-adversarial framework to eliminate such leakage of private information, and study the trade-off between recommender performance and leakage both theoretically and empirically using a benchmark dataset. An advantage of the proposed method is that it also helps guarantee fairness of results, since all implicit knowledge of a set of attributes is scrubbed from the representations used by the model, and thus can't enter into the decision making. We discuss further applications of this method towards the generation of deeper and more insightful recommendations.

Top