MSc SS Thesis Presentation
- Thursday, 23 August 2018
- LB 01.170 Timmanzaal
Multidomain Graph Signal Processing: Learning and SamplingGuillermo Ortiz Jiménez
In this era of data deluge, we are overwhelmed with massive volumes of extremely complex datasets. Data generated today is complex because it lacks a clear geometric structure, comes in great volumes, and it often contains information from multiple domains. In this thesis, we address these issues and propose two theoretical frameworks to handle such multidomain datasets. To begin with, we extend the recently developed geometric deep learning framework to multidomain graph signals, e.g., time-varying signals, defining a new type of convolutional layer that will allow us to deal with graph signals defined on top of several domains, e.g., electroencephalograms or traffic networks. After discussing its properties and motivating its use, we show how this operation can be efficiently implemented to run on a GPU and demonstrate its generalization abilities on a synthetic dataset. Next, we consider the problem of designing sparse sampling strategies for multidomain signals, which can be represented using tensors. To keep the framework general, we do not restrict ourselves to multidomain signals defined on irregular domains. Nonetheless, this particularization to multidomain graph signals is also presented. To do so, we leverage the multidomain structure of tensor signals and propose to acquire samples using a Kronecker-structured sensing function, thereby circumventing the curse of dimensionality. For designing such sensing functions, we develop several low-complexity greedy algorithms based on submodular optimization methods that compute near-optimal sampling sets. To validate the developed theory, we present several numerical examples, ranging from multi-antenna communications to graph signal processing.Additional information ...
MSc SS Thesis Presentation
- Monday, 16 July 2018
- HB 17.150
Impedance-based bioassay for characterization of single malignant melanoma cancer cells usinG CMOS-MEA systemsMakrina Sekeri
Malignant Melanoma (MM) is the most aggressive type of skin-cancer. Current diagnostic tools for the detection of malignancies of the skin (MM cancer) include histological, optical, ultrasound, and impedance-based techniques. The inadequacies of the first three practices are overwhelmed by the Electrical Impedance Spectroscopy (EIS) method. EIS overcomes reported spatiotemporal tradeoffs as a label-free and optics-free analytical method. Yet, MM’s enhanced heterogeneity and metastatic potential still results in misdiagnosis, or late diagnosis leading to stages characterized by high mortality rates. Important biological information and processing ability on single-cell level is missing. Single-cell dynamics recorded with a high-throughput system, contain important biological information on the heterogeneous subpopulations which are responsible for the MM aggressiveness.
This project aims to investigate experimentally the possibility and capabilities of such a bioassay development, create working protocols and generate a fundamental basis for analysis and interpretation of the big-data-sets which derive from Impedance monitoring from a high-throughput transducer.
Experiments were performed, employing two diverse, human-derived, MM cancer cell-lines, and using a high-throughput HD-MEA system with a 1024-channel impedance readout unit developed at IMEC, in Belgium. The measurements were realized at 1kHz aiming to extract Rseal information. The main proposal presents an experimental protocol of mid-term and long-term experiments Temporal and spatial resolutions were enhanced (Control System Automation), allowing for implementation of an experimental set to test the assay’s capabilities and determine any necessary additions to make the assay more robust for research (i.e. Z-Map, templates and scripts for OriginLab and Matlab, statistical methods for validation of findings on the big-data sets, optimizations in the experimental process, etc).Additional information ...