Vector Representation of Arbitrary-sized Matrices

In the thesis, we aim to compare matrices that can vary in size or shape, and their columns/rows are permutation invariant. We propose a general solution to such problem by using the vector representation to represent a matrix. Several feature representations are invented and investigated in the thesis, based on histogram, statistics,...

Stability-Based Model Selection in Non-Stationary Environment

Model selection is associated to model assessment, which is the problem of comparing different models, or model hyperparameters, for a particular learning task. It constitutes a fundamental step in building machine learning models. The central question is: How a model will work in the future? In this thesis, a new model selection scheme for...

Active Learning for Overlay Prediction in Semi-conductor Manufacturing

In the manufacturing of semi-conductor devices there is a constant demand for increasing precision and yield. Measuring and controlling overlay errors is essential in this process, but these measurements are difficult and costly. Predictive models can be used as an addition to measurements, but they required labelled data for training. To...

Quantitative evaluation of Generative Adversarial Networks and improved training techniques

Generative adversarial networks (GANs) are a class of generative models, for which the goal is to learn from training data and then to generate data with similar characteristics. Despite the wide use of GANs, a quantitative evaluation method of their performance is lacking. In the current work, we invented a series of artificial datasets,...

Model-based rare category detection for temporal data

Rare category detection is the task of discovering rare classes in unlabelled and imbalanced datasets. Existing algorithms focus almost exclusively on static data in which instances are assumed to be independent. In this thesis we propose an algorithm that is designed for temporal data. Specifically, we are interested in data with temporal...

Outlier detection in multivariate time series: Exploiting reconstructions from random projections

Recognizing and Handling Negations in Machine Learning

In several machine learning problems, a relatively small subproblem is present in which combinations of (negating) objects or structures result in a negation or otherwise other classification compared to when these (negating) objects are not present. To be more specific, a variant of the XOR problem is present in a small amount of objects in...

Cross-Modal Re-identification of Persons between RGB and Depth

Cross-modal person re-identification is the task to re-identify a person which was sensedin a first modality, like in visible light (RGB), in a second modality, like depth. Therefore, the challenge is to sense between inputs from separate modalities, without information from both modalities at the same time step. Lately, the scientific challenge...

An analysis of Generative Adversarial Networks on lower dimensional spaces

Generative Adversarial Networks (GANs) provide a new way of generating data. In this thesis, a strictly controlled parameter space is introduced from which a sample space with known underlying distributions can be generated. Having exact knowledge of the underlying distributions of the parameter space, makes that we can evaluate the quality of...

Controlled Modification of Generated (Style)GAN Latent Vectors

StyleGAN is a neural network architecture that is able to generate photo-realistic images. The diversity of generated images are ensured by latent vectors. These latent vectors encodes important features of generated images. They provide us insight-full information about properties of image generation in StyleGAN, which may also occur similarly...

Comparing Quantitative Metrics for Generative Adversarial Neural Networks

Since their conception in 2014, a large number of Generative Adversarial Networks (GANs) [2] has been pro- posed and developed. GANs have achieved great results in realistic image generation, among other fields. Recently, stunning images have been produced. The theory and application of GANs has received much attention. However, the evaluation...

Domain Adaptation in Acoustic Rainfall Sensors

Rainfall is increasing in frequency and intensity due to climate change. Hydrological models exist that can report bottlenecks in urban infrastructures. However, these require accurate rainfall estimations with high temporal and spatial resolution. The fulfillment of these requirements is challenged due to high costs. This can be solved with...