Understanding the early formation of attitudes towards emerging technologies, such as quantum science & technology (QS&T), is essential for aligning the effect of science communication in...Show moreUnderstanding the early formation of attitudes towards emerging technologies, such as quantum science & technology (QS&T), is essential for aligning the effect of science communication in practice with its intentions. This study explores the application of natural language processing (NLP) techniques to investigate the influence of news articles on social media comments regarding QS&T. We curated a dataset of 217 articles and 14, 391 top-level comments from Reddit. Employing GPT-4, a semi-automated annotation method was developed to label comments for sentiment and engagement towards QS&T, achieving a Cohen’s kappa of 0.82 when pooling over multiple labelling repetitions and comparing with human annotations. We then used support-vector regression to determine if news article embeddings could be used to predict comment sentiment and engagement. After experimenting with various embedding strategies, including Sentence-BERT and RoBERTa models, no significant correlations were found between article content and comment sentiment or engagement towards QS&T. Further interdisciplinary work in empirical communication research and NLP is suggested to explore alternative representations for news articles and their ability to predict perceptions in news comments.Show less
Learning curves are important for decision making in supervised machine learning. They show how the performance of a machine learning model develops over a given resource. In this work, we consider...Show moreLearning curves are important for decision making in supervised machine learning. They show how the performance of a machine learning model develops over a given resource. In this work, we consider learning curves that model the performance of a machine learning model as a function of the number of data points used for training. For decision making, it is of- ten useful to extrapolate learning curves, which can be done, for example, by fitting a parametric model based on the observed values, or by using an extrapolation model trained on learning curves from similar datasets. We perform an analysis comparing these two techniques with different ob- servations and prediction objectives. When only a small number of initial segments of the learning curve have been observed we find that it is better to rely on learning curves from similar datasets. Once more observations have been made, a parametric model, or just the last observation, should be used. Moreover, we find that using a parametric model is mostly use- ful when the exact value of the learning curve itself is of interest. Lastly, we use this knowledge to improve machine learning on a particle physics dataset.Show less
This thesis research aims to improve traffic sign detection within dashcam footage by using temporal information. Essentially, a video is a set of images displayed at a fast rate. Temporal...Show moreThis thesis research aims to improve traffic sign detection within dashcam footage by using temporal information. Essentially, a video is a set of images displayed at a fast rate. Temporal information lies in the similarity across subsequent frames. However, current state-of-the-art object detection frameworks only use single images. To test whether temporal information can increase the performance of a Convolutional Neural Network (CNN), we train three models: YoloV5, a 3D CNN and a 4D CNN. YoloV5 is used to benchmark the other models against a state-of-the-art framework for object detection. Second, the existing architecture of YoloV5 is adopted as a basis for the 3D CNN. After tuning the hyperparameters for the 3D CNN, performance is compared to YoloV5. Third, the 3D CNN is changed into a 4D CNN that processes sets of frames. By combining the frames within a set, the information in each frame is fused together, including the temporal information across the frames. We call this temporal information fusion (TIF). Comparing the performance of the 3D CNN to those of the 4D CNN shows the effect of TIF. In this research, a balanced dataset containing 444 sets of frames containing traffic signs from dashcam videos is used to train and test the models. The objective is to correctly classify the traffic signs on the frames. The results show that TIF can increase the accuracy of a CNN model by 2\%, purely through the addition of TIF. The main drawback of using TIF is an increase in processing time. Instead of a single image, the network needs to process a set of images, which naturally will take longer. The results in this research can form a basis to explore TIF in object detection further.Show less
In recent years the use of network analysis has seen an increase across multiple fields. Using features derived from network analysis in a random forest model has already shown promising results in...Show moreIn recent years the use of network analysis has seen an increase across multiple fields. Using features derived from network analysis in a random forest model has already shown promising results in an experiment of the Netherlands Labour Inspection to predict risk scores of violating Dutch labour laws for Dutch companies. In order to see if this is the best way for the NLA to use network analysis this method has been compared with using node classification, a deep learning method that uses sampling and aggregation to label nodes of a network based on its neighbours and the network structure. In this study node classification has shown to perform better at predicting risk scores for Dutch companies than a random forest model with network features does. A simulation study was done on the node classification method to test its robustness and has shown that it is important that there are enough labels in the training set in order for the method to perform well and that the quality of these labels influences the extent to which the model overfits. If the data scientist of the NLA choose to use node classification in future projects it is important that they make an ethical selection of the variables to use for prediction and that they ensure that there are enough labels in the dataset for the node classification model to perform well.Show less
The recent popularization of machine learning as a new paradigm in computer science provides interesting opportunities for explaining phenomena of collective motion in living systems, as for...Show moreThe recent popularization of machine learning as a new paradigm in computer science provides interesting opportunities for explaining phenomena of collective motion in living systems, as for example flocks of birds or schools of fish. In this thesis we develop a model for collective motion using multi-agent reinforcement learning with orientation-based rewards, a new type of reward system that has not yet been found in literature. While the developed model is in principle generally applicable to all forms of collective motion observed in nature, we use the language of the flocking behaviour of birds as a particular example to frame our model. The birds have the option to either fly into an instinctive direction or act based on a Viscek-type of interaction with their neighbors, and are rewarded maximally when the resulting direction of movement is some predetermined prefered direction. The model distinghuishes between leaders that instinctively move towards this direction and followers that do not. We show that collective motion into this prefered direction emerges from this model, but only with a minimum of 1.23 encounters with neighbours on average, of which a minimal fraction of 0.2 should be leaders, which on average roughly corresponds to at least one encounter with a leader every four timesteps. These lower bounds are rudimentary estimates, as the present study serves mainly as a proof of concept that collective motion can emerge from this new type of model. Additionally it is suggested that, using deep reinforcement learning, this model can be viewed as a reinforcement learning extension of the Vicsek model.Show less
The inflationary hypothesis was introduced as a solution to the fine-tuning issue in the initial conditions of the Big Bang theory. In this Master’s research project, we introduce our work in the...Show moreThe inflationary hypothesis was introduced as a solution to the fine-tuning issue in the initial conditions of the Big Bang theory. In this Master’s research project, we introduce our work in the search for features in the Cosmic Microwave Background (CMB) power spectra that could result from reductions in the speed of sound of the inflaton. We study these features in the context of an effective single field theory of a multiple field scenario, due to the fact that a single field inflation approach is favoured by the current cosmological data, especially by the CMB. First, we present a brief review of the current cosmological model, the ΛCDM model, inflation and the possible extensions. Secondly, we review the physics of the CMB, the main theoretical cosmological codes and the needed data analysis tools from the point of view of Bayesian statistics. Finally, we update our current search for features using Planck 2015 temperature and polarization data introducing new parametrizations for the reduction of the speed of sound. In this search, we have recovered only some previous found modes, indicating the dependency of our results with respect to the parametrization we were using. For this reason, we have pointed out the necessity of reconstructing the reduction of the speed of sound, showing some preliminary results when Gaussian Processes are used as the reconstruction technique.Show less
Deep reinforcement learning has solved the game of Go, along with all other board games. Can it also be applied to real-world use cases? This research combines a literature study and experimental...Show moreDeep reinforcement learning has solved the game of Go, along with all other board games. Can it also be applied to real-world use cases? This research combines a literature study and experimental evaluation, focusing on the case of automation for tele-operated robotics. This is necessary because tele-operation of robots is slow and cumbersome. Classical robotics solutions are expensive, and limited in precision, but deep reinforcement learning provides an opportunity for learning visuomotor skills using partial information.Show less
