Search
Results
-
Insights on the interplay of network architectures and optimization algorithms in deep learning
(2022)In the recent decade, deep neural networks have solved ever more complex tasks across many fronts in science, industry, and society. Compared to classical machine learning methods, the tremendous success of deep neural networks builds upon their capacity to extract meaningful embeddings of input data that come in representations with low semantic content. The crucial step before deployment of such models is the "training" procedure, in ...Doctoral Thesis -
Live Aircraft Destination Prediction using Machine Learning & Adversarial Attacks
(2022)As a continuation of uncovering threats posed in aviation that are due to ADS-B not being secure, we looked into how a non-commercial aircraft user's privacy can be compromised in real-time with publicly available data. Since there are many ways to predict the destination of an airplane whilst in the air we investigated this matter using a long short-term memory and a gradient booster model. With the gradient booster model, depending on ...Bachelor Thesis -
MMP: An Object-Oriented Multi-Machine Parser Generator
(2022)When compared with the manual labor of crafting parsers, most parser generators currently available are not flexible and powerful enough. They prohibit the free combination of various parsing algorithms/machines, restrict the form of grammar rules, and provide limited support for semantic actions and nonterminals’ fields. In this thesis, we present a parser generator Multi-Machine Parsing (MMP) that solves these three problems. MMP ...Master Thesis -
Advancing Packet-Level Traffic Predictions with Transformers
(2022)Learning underlying network dynamics from packet-level data has been deemed an extremely difficult task, to the point that it is practically not attempted. While research has shown that machine learning (ML) models can be used to learn behaviour and improve on some specific tasks in the networking domain, these models do not generalize to any other tasks. However, a new ML model called the Transformer has shown massive generalization ...Master Thesis -
Designing a Low-Power Autonomous Embedded System for Aircraft Collision Prevention
(2022)Bachelor Thesis -
Practical data-centric optimization of C/C++ applications
(2022)C is a programming language used everywhere and in a big part of scientific codes. With the increase of highly parallel computing machines the code needs to be adapted to be able to use the full power available. In my thesis I will expand a compiler frontend to be able to transform C code into a data centric representation that will intern be able to identify data dependencies automatically and hence produce highly parallel code. ...Bachelor Thesis -
Low-rank Kronecker-factored Approximate Curvature for Training Deep Neural Networks
(2022)Second-order optimization methods applied to train deep neural net- works use the curvature information of a loss function to enable much faster convergence in the number of steps than first-order methods. However, since the second-order methods need the curvature matrix and its inverse, they have a huge computational and memory overhead that negates the faster convergence in terms of time. For this reason, first-order optimization methods ...Bachelor Thesis -
Development of a georeferencing system for a helicopter-borne radar system
(2022)The laboratory of Hydraulics, Hydrology and Glaciology of ETH Zurich developed an airborne ice radar system (AIR-ETH) for monitoring glacier bedrock. The system consists of a wooden rectangular platform, equipped with radar instruments. They are looking for a low-cost georeferencing system for AIR-ETH. This system should determine the position of the pivot point and the attitude of the platform with a sampling rate of up to 10 Hz. These ...Master Thesis -