347 research outputs found
Erratum: The role of visual preferences in architecture views
No full textThe article “The role of visual preferences in architecture views” by Ali Akbar Amini, Bahman Adibzadeh, published on 24 September 2020 in the Journal of Architecture and Urbanism, 44(2), 122–127, https://doi.org/10.3846/jau.2020.12582 contained a following errors on:
122 p. The source is incorrectly cited in the text. The correct citation is:
(de la Fuente Suárez, 2016)
126 p. The references incorrectly indicate author name, lastname and title of article. The correct citation is:
de la Fuente Suárez, L. A. (2016). Towards experiential representation in architecture. Journal of Architecture and Urbanism, 40(1), 47–58. https://doi.org/10.3846/20297955.2016.1163243
Corrected version of the article is available online.
The publisher apologises for this error
Sedimentation Processes in the Tinto and Odiel Salt Marshes in Huelva, Spain
No full textGlobal warming is a key factor to take into account when a study is conducted on tidal
wetlands. Both Odiel and Tinto salt marshes are the major wetlands in Andalusia (Spain).
From the mid-1950s to date, the land use changes (LUC) have caused a great landscape
alteration that along with the effects of climatic variables and sea wave energy have given
rise to a hard impact on the environment. The advent of new image processing procedures and use of high-resolution images from satellites gave precise patterns of erosion.
In this work, a new method patented by the author is presented and used to obtain the
total cubic meters of eroded soil in both salt marshes. Moreover, the different factors that
begin this phenomenon as well as the influence of intertidal processes are discussed. The
results show how the greater integration of remote sensing and geographical information
systems (GIS) technologies, with regression model, was most useful to describe, analyze
and predict the volumetric change process in both salt marshes
LCDB 1.0: An Extensive Learning Curves Database for Classification Tasks
No full textThe use of learning curves for decision making in supervised machine learning is standard practice, yet understanding of their behavior is rather limited. To facilitate a deepening of our knowledge, we introduce the Learning Curve Database (LCDB), which contains empirical learning curves of 20 classification algorithms on 246 datasets. One of the LCDB’s unique strength is that it contains all (probabilistic) predictions, which allows for building learning curves of arbitrary metrics. Moreover, it unifies the properties of similar high quality databases in that it (i) defines clean splits between training, validation, and test data, (ii) provides training times, and (iii) provides an API for convenient access (pip install lcdb). We demonstrate the utility of LCDB by analyzing some learning curve phenomena, such as convexity, monotonicity, peaking, and curve shapes. Improving our understanding of these matters is essential for efficient use of learning curves for model selection, speeding up model training, and to determine the value of more training data.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.Pattern Recognition and Bioinformatic
Adversarially Robust Decision Tree Relabeling
No full textDecision trees are popular models for their interpretation properties and their success in ensemble models for structured data. However, common decision tree learning algorithms produce models that suffer from adversarial examples. Recent work on robust decision tree learning mitigates this issue by taking adversarial perturbations into account during training. While these methods generate robust shallow trees, their relative quality reduces when training deeper trees due the methods being greedy. In this work we propose robust relabeling, a post-learning procedure that optimally changes the prediction labels of decision tree leaves to maximize adversarial robustness. We show this can be achieved in polynomial time in terms of the number of samples and leaves. Our results on 10 datasets show a significant improvement in adversarial accuracy both for single decision trees and tree ensembles. Decision trees and random forests trained with a state-of-the-art robust learning algorithm also benefited from robust relabeling.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.Cyber Securit
Penalized FTRL with Time-Varying Constraints
No full textIn this paper we extend the classical Follow-The-Regularized-Leader (FTRL) algorithm to encompass time-varying constraints, through adaptive penalization. We establish sufficient conditions for the proposed Penalized FTRL algorithm to achieve O(t) regret and violation with respect to a strong benchmark X^tmax. Lacking prior knowledge of the constraints, this is probably the largest benchmark set that we can reasonably hope for. Our sufficient conditions are necessary in the sense that when they are violated there exist examples where O(t) regret and violation is not achieved. Compared to the best existing primal-dual algorithms, Penalized FTRL substantially extends the class of problems for which O(t) regret and violation performance is achievable.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.Networked System
SECLEDS: Sequence Clustering in Evolving Data Streams via Multiple Medoids and Medoid Voting
No full textSequence clustering in a streaming environment is challenging because it is computationally expensive, and the sequences may evolve over time. K-medoids or Partitioning Around Medoids (PAM) is commonly used to cluster sequences since it supports alignment-based distances, and the k-centers being actual data items helps with cluster interpretability. However, offline k-medoids has no support for concept drift, while also being prohibitively expensive for clustering data streams. We therefore propose SECLEDS, a streaming variant of the k-medoids algorithm with constant memory footprint. SECLEDS has two unique properties: i) it uses multiple medoids per cluster, producing stable highquality clusters, and ii) it handles concept drift using an intuitive Medoid Voting scheme for approximating cluster distances. Unlike existing adaptive algorithms that create new clusters for new concepts, SECLEDS follows a fundamentally different approach, where the clusters themselves evolve with an evolving stream. Using real and synthetic datasets, we empirically demonstrate that SECLEDS produces high-quality clusters regardless of drift, stream size, data dimensionality, and number of clusters. We compare against three popular stream and batch clustering algorithms. The state-of-the-art BanditPAM is used as an offline benchmark. SECLEDS achieves comparable F1 score to BanditPAM while reducing the number of required distance computations by 83.7%. Importantly, SECLEDS outperforms all baselines by 138.7% when the stream contains drift. We also cluster real network traffic, and provide evidence that SECLEDS can support network bandwidths of up to 1.08 Gbps while using the (expensive) dynamic time warping distance.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.Cyber Securit
Combined Effects of Saltwater and Water Flow on Deterioration of Concrete under Freeze–Thaw Cycles
No full textWireless Communication onboard Spacecraft: Draadloze Communicatie aan boord van Ruimtevaartuigen
No full textThis dissertation focuses on intra-spacecraft wireless communication as a solution for reducing the spacecraft onboard harness. Despite outstanding advances in aerospace industry, the cost of accessing space is still very high and the amount of engineering work required for spacecraft design and development is enormous. The key elements which increase the development and launch cost of a spacecraft are size, mass, and the necessity of a tailored design for each mission. Studies show that the contribution of onboard harness to spacecraft mass is about 6% to 10%. Any effort to reduce harness can directly result in reducing the launch cost and arriving to a more modular and flexible design.This thesis aims to answer the following questions:1. What are the problems of onboard wired standards and what are the benefits and characteristics of wireless network onboard spacecraft?2. Which spacecraft subsystems could benefit most from a wireless onboard communication paradigm?3. What is the major challenge regarding employing a wireless standard onboard a spacecraft?4. How can we solve the identified system level design challenge?To answer these questions, this dissertation reviews the existing wired spacecraft data bus standards and major commercial off the shelf (COTS) wireless communication solutions to identify and characterize their architectures. These wireless standards are Wi-Fi, Bluetooth and ZigBee. Categorizing different onboard data types aids to identify a suitable COTS wireless communication solution for each application category. Specifically, sensors of attitude determination and control system (ADCS) can greatly benefit from a low power and low data rate wireless communication solution such as ZigBee, however, the major challenge is conserving energy on the sensors to enable a wireless architecture and achieve an adequate battery life without compromising the system performance. This dissertation proposes two onboard energy managers based on sensor scheduling schemes to tackle the energy conservation challenge. These solutions are tailored to ADCS and aim to reduce the overall ADCS energy consumption without affecting the required accuracy of attitude determination. Both energy managers use similar design elements and decision making algorithms while one of them presents a centralized scheme and the other one employs a decentralized architecture. A unique characteristic of these designs is that the energy management solution is fully integrated with the onboard attitude determination system of the spacecraft. Simulation results show that enabling the energy managers result in total energy saving between 20.9% to 51% (depending on the scenario) without compromoising accuracy of attitude determination
Rate-constrained multi-microphone noise reduction for hearing aid devices
No full textMany people around the world suffer from hearing problems (In the Netherlands, around 11%of the population is considered hearing-impaired). To overcome their hearing problems, advanced technologies like hearing aid devices can be used. Hearing aids are meant to assist the hearing-impaired to improve the speech intelligibility and the quality of sounds that they intend to hear. Usually these include processors which are mainly designed to enhance the sound signals originating from the source of interest by reducing the environmental noise. Binaural hearing aids, on the other hand, can also help to preserve some spatial information from the acoustic scene, which can help the hearing aid user to hear the sounds from the correct locations. To construct the binaural hearing aid system, two hearing aids are needed to be placed in the left and the right ears, which can potentially communicate through a wireless link. In addition, one can think of additional assisting devices with microphones placed in the environment. One common way to reduce the noise is to use advanced binaural multi-microphone noise reduction algorithms, which aim at estimating some desired sources while reducing the power of the undesired sources. One typical method is to use spatial filtering, which aims at estimating the target signal by shaping the beam towards the location of the desired source while canceling/suppressing the other sources. To perform binaural noise reduction, while assuming centralized processing, the signals recorded at remote microphones (for example from additional assisting devices or in the binaural hearing aid setup, the sound signals from the contralateral hearing aid) need to be transmitted to the central processor. Due to the power and bandwidth limitations, the data needs to be compressed before transmission. Therefore, the main question would be, at which rate the data should be compressed to have reasonably good noise reduction performance. This links the noise reduction problem to the data compression problem. Generally, the higher the data rate, the better the noise reduction performance. Therefore, there is a trade-off between the performance of the noise reduction algorithm and the data-rate at which the information is compressed. This problem is closely connected to the rate-distortion problem from an information-theoretic viewpoint. Studying the effect of data compression on the performance of noise reduction problems would be of great interest to reduce the power consumption of hearing assistive devices. Oneway to incorporate data compression into the noise reduction problem is to perform quantization, which leads to a rate-constrained noise reduction problem. In the rate-constrained noise reduction, the goal is to estimate the desired sources based on the imperfect data. The observations from remote sensors are quantized and transmitted to the fusion center. The main challenge in the binaural rate-constrained noise reduction is to find the best quantization rates for the different sensors at different frequencies, given the physical constraints like bitrate and power constraints. Another aspect of the rate-constrained noise reduction is to expand the network to receive more information on the acoustic scene using additional assistive devices. Target source estimation using information from such assistive devices (rather than only binaural hearing aids) is shown to result in better noise reduction performance. Now the question is how to allocate the bitrates to the assistive devices as well. These assistive devices can be thought of as the remote embedded microphones on the cell-phones (mobile) or wearable microphones placed at the users’ bodies. The binaural hearing aid system can thus be generalized to allow other assistive devices to contribute to noise reduction. In this dissertation, we study and propose different rate-constrained multi- microphone noise reduction algorithms. We try to expand the notion of the binaural rate constrained noise reduction to multi-microphone rate-constrained noise reduction for general wireless acoustic sensor networks (WASNs). The WASN in this case can include the binaural setup along with other assistive devices. We propose different algorithms to cover the main objectives of rate-constrained noise reduction problems. These objectives mainly include good target estimation (less environmental noise power) given the compressed data, good rate allocation strategies in WASNs, and preferably preserved spatial information of the sources in the acoustic scene to get the correct impression of the acoustic scene.Signal Processing System
- …
