In recent years the t-distributed Stochastic Neighbor Embedding (t-SNE) algorithm has become one of the most used and insightful techniques for exploratory data analysis of high-dimensional data.... Show moreIn recent years the t-distributed Stochastic Neighbor Embedding (t-SNE) algorithm has become one of the most used and insightful techniques for exploratory data analysis of high-dimensional data. It reveals clusters of high-dimensional data points at different scales while only requiring minimal tuning of its parameters. However, the computational complexity of the algorithm limits its application to relatively small datasets. To address this problem, several evolutions of t-SNE have been developed in recent years, mainly focusing on the scalability of the similarity computations between data points. However, these contributions are insufficient to achieve interactive rates when visualizing the evolution of the t-SNE embedding for large datasets. In this work, we present a novel approach to the minimization of the t-SNE objective function that heavily relies on graphics hardware and has linear computational complexity. Our technique decreases the computational cost of running t-SNE on datasets by orders of magnitude and retains or improves on the accuracy of past approximated techniques. We propose to approximate the repulsive forces between data points by splatting kernel textures for each data point. This approximation allows us to reformulate the t-SNE minimization problem as a series of tensor operations that can be efficiently executed on the graphics card. An efficient implementation of our technique is integrated and available for use in the widely used Google TensorFlow.js, and an open-source C++ library. Show less
Yilmaz, D.; Heijden, A.C. van der; Thijssen, J.; Schalij, M.J.; Erven, L. van 2017
Although both ICD and CRT have proven to be an effective treatment strategy for selected patients in large clinical trials, many issues of the effects of defibrillator treatment in routine clinical... Show moreAlthough both ICD and CRT have proven to be an effective treatment strategy for selected patients in large clinical trials, many issues of the effects of defibrillator treatment in routine clinical practice remain unclear. In the current thesis, some of these unresolved questions are clarified based on data from a large cohort of ICD and CRT-D patients with long-term follow-up in routine clinical practice. In the first part the thesis was specifically focused on clinical aspects such as pocket related complications, the mode of death, ventricular arrhythmias and the suitability for subcutaneous device implantation in ICD recipients. In the second part the thesis was focused on socio-economic implications of ICD therapy such as the cost-effectiveness of ICD therapy, device longevity and an evidence-based approach regarding driving restrictions in ICD patients. In addition, the last chapter focused on the development and implementation of criteria that allow better identification of high risk patients and to limit the number of defibrillator implants in patients who will not benefit Show less