465 research outputs found
Fundumental One-Dimensional Analysis of Photo-Diodes
No full textTitle: Fundumental One-Dimensional Analysis of Photo-Diodes, Author: T.B. Remple, Location: Thodethe program developed by A.M. Start, in his paper, Fundamental One-Dimentional Analysis of Transistors, Philips Research Report Supplements, #4, 1976, has been modified to handle high voltage, reversed biased p-i-n photo-diodes. The physical involved in the development of stark;s model is summarized and three different p-i-n diodes are analyzed. A Schottky barrier is also analyzed by assuming the metal contact is a very highly doped semiconductor material. A listing of the program is given in the appendices, as well as a description of the program and a user's guide. Te program is written in Fortran, was run on a CDC 6400 in double precision (giving 29 digits accuracy), requiring 45 k of memory and 300 to 1000 seconds run time.ThesisMaster of Engineering (ME
Development and Utilization of a Signal Averaging System
No full textTitle: Development and Utilization of a Signal Averaging System, Author: T. B. Remple, Location: ThodeA sampling oscilloscope was interfaced to a PDP11/10 minicomputer via one analog-to-digital converter (ADC) and one digital-to-analog converter (DAC), with the minicomputer being interfaced to a real time oscilloscope or plotter via two more DAC's. A software program that employed both assembler and Fortran was developed. This program controlled the signal averaging, did some data analysis, could output the data in normalized form onto a real time scope or a plotter, and also input and output the signal using punched type. The usefulness of the system was then demonstrated by measuring the rise time of a short laser pulse through various lengths of graded index fibre, and the fibre response was extracted as a function of fibre length.ThesisMaster of Engineering (ME
CHAINels: Journal
No full textA group of three students worked a couple of months at CHAINels for their computer science bachelor project. In these months a recommendation algorithm was designed and implemented in CHAINels. The recommendation algorithm recommends posts to a company and those posts are shown in the Journal which was also made during this project. In this report, every step of the design and implementation of the Journal and recommendation algorithm is explained.Software TechnologyElectrical Engineering, Mathematics and Computer Scienc
Solving the Plan Coordination Problem
No full textIn this thesis we focus on implicit coordination for multi-agent planning problems. In such problems, agents are not able or willing to cooperate with each other and hence we need to perform pre-planning coordination in order to ensure that merging all their plans always results in a feasible joint plan. More specifically, we are interested in finding a minimal cardinality set of constraints such that when add this set to the multi-agent planning problem, no infeasible joint plan can be constructed, whatever local plan each agent develops. Finding such a minimal cardinality set is known as the PLAN COORDINATION PROBLEM (PC) which has been proven to be ?p2 -complete [44]. Previous work has focussed on approximation and special cases for PC, however some scenarios require or allow for exact solutions. Also, smaller instances might be solvable in reasonable time. This thesis discusses several exact solving methods and combines them into one exact algorithm that is able to solve instances with task sizes up to 50 planarcs within the hour.Department of AlgorithmicsSoftware EngineeringElectrical Engineering, Mathematics and Computer Scienc
Project Scheduling: The Impact of Instance Structure on Heuristic Performance
No full textMany meta-heuristic approaches have been suggested for or applied to the Resource Constrained Project Scheduling Problem (RCPSP). The existence of a number of highly accessible standard benchmark sets has promoted a research focus on finding anything that improves average solution quality, without investigating what effect is responsible for the improvement, or what is responsible for holding us back. This work focuses instead on understanding the original constructive Schedule Generation heuristics and their interaction with a well known but poorly understood post-processing step called Forward-Backward Improvement that is known to almost always improve any generated RCPSP schedule. We follow an empirical investigation methodology by first observing the effect of FBI on a large generated testset. Based on these observations we explain why FBI works by means of hypotheses on its operation. These hypotheses generate predictions that we subsequently successfully test in a second round of experiments. In the process we are able to propose a novel priority rule heuristic based on the principles of FBI. We find that this new rule outperforms the current best priority rule heuristic.AlgorithmicsSoftware TechnologyElectrical Engineering, Mathematics and Computer Scienc
Using cross-model learnings for the Gram Vaani ASR Challenge 2022
No full textIn the diverse and multilingual land of India, Hindi is spoken as a first language by a majority of its population. Efforts are made to obtain data in terms of audio, transcriptions, dictionary, etc. to develop speech-technology applications in Hindi. Similarly, the Gram-Vaani ASR Challenge 2022 provides spontaneous telephone speech, with natural back-ground and regional variations in Hindi. The challenge provides: 100 hours of labeled train-set, 5 hours of labeled dev-set and 1000 hours of unlabeled data-set. For the 'Closed Challenge', we trained an End-to-End (E2E) Conformer model using speed perturbations, SpecAugment techniques and use VTLN to handle any unknown speaker groups in the blind evaluation set. On the dev-set, we achieved a 30.3% WER compared to the 34.8% WER by the Challenge E2E baseline. For the 'Self Supervised Closed Challenge', a semi-supervised learning approach is used. We generate pseudo-transcripts for the unlabeled data using a hybrid TDNN-3gram LM model and trained an E2E model. This is then used as a seed for retraining the E2E model with high confidence data. Cross-model learning and refining of the E2E model gave 25.3% WER on the dev-set compared to ∼33-35% WER by the Challenge baseline that use wav2vec models.Green Open Access added to TU Delft Institutional Repository 'You share, we take care!' - Taverne project https://www.openaccess.nl/en/you-share-we-take-care Otherwise as indicated in the copyright section: the publisher is the copyright holder of this work and the author uses the Dutch legislation to make this work public.Multimedia Computin
Regional gravity field modelling with radial basis functions
No full textPhysical and Space GeodesyAerospace Engineerin
Preconditioned conjugate gradient method enhanced by deflation of rigid body modes applied to composite materials.
No full textElectrical Engineering, Mathematics and Computer Scienc
Development of a solver for large algebraic system from structural mechanics
No full textElectrical Engineering, Mathematics and Computer Scienc
- …
