71 research outputs found
Supplementary Data from Prognostic Value of Residual Disease after Neoadjuvant Therapy in HER2-Positive Breast Cancer Evaluated by Residual Cancer Burden, Neoadjuvant Response Index, and Neo-Bioscore
No full textSupplementary data to Steenbruggen TG, van Seijen M, Janssen LM, van Ramshorst MS, van Werkhoven E, Vrancken Peeters MJTDF, et al. Prognostic value of residual disease after neoadjuvant therapy in HER2-positive breast cancer evaluated by Residual Cancer Burden, Neoadjuvant Response Index & Neo-Bioscore. Table S1, Table S2, Table S3, Table S4, Text and legend Figure S1</p
Bereiding van benzoëzuur door katalytische oxidatie van tolueen in de vloeistoffase
No full textDocument(en) uit de collectie Chemische ProcestechnologieDelftChemTechApplied Science
Automatic Segmentation of Ships in Digital Images: A Deep Learning Approach
No full textKnowledge on adversaries during military missions at sea heavily influences decision making, making identification of unknown vessels an important task. Identification of surrounding vessels based on visual data offers an alternative to AIS information (Automatic Identification System), the current standard in vessel identification, which can be spoofed. One visual approach employs human expertise and manually identifies vessels guided by a ship catalog. In order to minimize or potentially eliminate human error and performance limitations, there is strong interest in developing an automated vessel classification pipeline. One such pipeline is currently being developed at TNO, capable of classifying over 500 separate classes. A crucial part of the classification pipeline is retrieving an accurate contour of a vessel from a digital image. To address this important challenge, this thesis proposes an advanced deep learning pipeline to automatically segment the vessel image into background (e.g. sky and sea) and the object of interest (a vessel). Deep learning models based on Fully Convolutional Neural Networks (FCNs) have achieved high performance on the task of semantic segmentation. Several networks such as CRF-RNN, PSPNet, DeepLab and Mask R-CNN are employed to determine a baseline performance. We will focus on identifying the cause of poor or failing segmentations and aim to construct a robust network capable of handling these challenges. By sampling disturbances, caused by ship distance and camera noise, augmented data sets are built to tune networks to input from on-site images. Additionally, experiments are done to evaluate the influence of different levels of disturbances. Previous approaches implementing the CRF-RNN network achieved top 1 and top 5 classification accuracies of 31.1% and 44.0% respectively. Employing the DeepLab network, trained to convergence on artificial noise augmented data, we report top 1 and top 5 accuracy of 68.9% and 88.8% respectively. Additionally, implementing an ensemble of classifiers, performance is increased to 73.0% and 91.7% for top 1 and top 5 accuracy respectively. This best result is comparable to the classification results with human annotated ship silhouettes. The human performance accuracy is 73.4% on top 1, and 91.3% on top 5 classification performance. Finally, we show that training on a collection of different levels of image disturbances results in a network that is robust against increasing disturbance in images, while retaining performance on clean images.Mechanical Engineering | Systems and Contro
Wind speed measurements using distributed fiber optics: a wind tunnel study
No full textNear-surface wind speed is typically only measured by point observations. The so-called Actively Heated Fiber-Optic (AHFO) technique, however, has the potential to provide high-resolution distributed observations, allowing for better understanding of different processes. However, before it can be widely used, its performance needs to be tested in a range of settings. Therefore, in this work, experimental results on this novel observational wind-probing technique are presented. We utilized a controlled wind-tunnel setup to assess both the accuracy and the precision of AHFO as well as its potential for outdoor atmospheric operation. The technique allows for wind speed characterization with a spatial resolution of 0.3 m on a 1 s time scale. The flow in the wind tunnel is varied in a controlled manner, such that the mean wind, ranges between 1 and 17 m/s. Comparison of the AHFO measurements with observations from a sonic anemometer shows a high overall correlation, ranging from 0.94-0.99. Also, both precision and accuracy are greater than 95 %. As such, it is concluded that the AHFO has potential to be employed as an outdoor observational technique in addition to existing techniques. In particular, it allows for characterization of spatial varying fields of mean wind in complex terrain, such as in canopy flows or in sloping terrain. In the future the technique could be combined with regular Distributed Temperature Sensing (DTS) for turbulent heat flux estimation in micrometeorological/hydrological applications
Groundwater recharge in Myanmar - Estimations in the Chindwin catchment by base flow separation and SWAT
No full textFirst estimations for the groundwater recharge in the Chindwin basin in Myanmar are presented in this report. This estimations are based on base flow separation and a SWAT model. Multiple base flow separation methods are applied and these are compared with the base flow produced by the SWAT model. This first estimations show a range of 248-670 mm average groundwater recharge per year, which is roughly 11-30% of the average annual rain in the catchment. The upper limit, produced by SWAT, seems to be too high, as the total flow is overestimated due to a rate of evaporation which is too low compared with remote sensing based evaporation. In the Chindwin it appeared that it is rather difficult to separate the base flow due to the highly sensitive alpha parameter, which determines the response behaviour of the base flow, also there seem to be multiple groundwater components. This makes it hard to perform base flow separation, but also increases the difficulty of calibrating the SWAT model, especially the base flow component. However the SWAT model shows a clear spatial groundwater recharge pattern even though it needs further optimization. Optimizing a (hydraulic) model can be cumbersome and especially in a developing environment as Myanmar it is also interesting to look at other possibilities/models, like Water Accounting +, which are more flexible to adapt to changes as new dams or increase in groundwater irrigation.InternshipWater Managemen
Genetic Programming in Hydrology: Using genetic programming in conceptual modelling
No full textThis report introduces the use of Genetic Programming (GP) into hydrology by describing the results of GP using conceptual hydrological models as physical representation. First the possibilities of GP are tested on synthetic data, which results in a shortlist of good working objective functions and insight in the most important GP settings. The test on real data in the Belgium Ardennes showed that GP using the objective functions KG10, MM and Shafii performed better. Nevertheless all three models performed not well on simulating the low flows and high peaks. Furthermore GP using KG10 and MM both results in simple serial models which perform well overall, but bad on quick response runoff. Shafii resulted in parallel models which show quick response flow, however GP it is not able to capture the fast responses correctly (yet). GP has the potential to improve the understanding in the behaviour of catchments, however it still needs the human mind to observe, compare and analyse the modelling results. The main consideration with GP is to look for a balance between: model search space, objective function, randomness and (computational) time. The challenge is how to lead GP in an efficient way without removing the possibility of finding unknown patterns.Additional thesisWater Managemen
RACE:GP – a Generic Approach to Automatically Creating and Evaluating Hybrid Recommender Systems
No full textIn recent years, recommender systems have become a fundamental part of our online experience. Users rely on such systems in situations with many potential choices, such as watching a movie on a streaming service, reading a blog post, or listening to a song. Traditionally, these systems use techniques such as collaborative filtering and content-based recommendation. Both approaches have disadvantages, so to reduce those, recent research combines various techniques in different ways to create hybrid recommender systems. Creating a well-performing hybrid recommender system generally requires extensive knowledge of recommender systems, the domain on which one wants to provide recommendations, and trial and error. Automating this process makes recommender systems accessible for organizations that lack the resources to build these systems themselves. However, there is a lack of research regarding automating this process. This study aims to provide an initial exploration into this area by proposing RACE:GP, an end-to-end approach that automatically produces accurate non-trivial hybrid recommender systems with only a dataset and a definition of 'relevance'. RACE:GP automatically creates a programming language from a dataset in which any valid program is a recommender system on that dataset. By defining the relevant interaction, it can automatically evaluate the accuracy of these programs. It uses a search strategy based on genetic programming to find the best performing recommender systems in the language. To test our hypothesis, RACE:GP is used to produce recommender systems on three well-known datasets in recommender systems literature, and the results are compared to baselines based on collaborative filtering. Additionally, to verify its adaptability, we analyzed the produced recommender systems given different recommendation scenarios. The results showed that RACE:GP is able to produce recommender systems that outperform our chosen baselines by a significant margin. Furthermore, analysis of the produced recommender systems on different recommendation scenarios within a dataset shows that it can find systems that are especially accurate in situations with different densities or recommending specific interactions in datasets that contain different interactions. These results suggest that RACE:GP is a viable and generally applicable approach that makes creating hybrid recommender systems accessible for anyone with a dataset
FleXentral: Managing the contingent workforce
No full textThe market for contingent workers grows every year. Managing this workforce gets more and more challenging because of regulatory compliance, the difficulty of finding matching employees, and keeping track of expenses. FleXentral tries to be a platform that solves these issues. The goal of this project was to build a proof-of-concept of this concept, that can be used to show the potential to customers, but that will also be used as a basis for the final application. We created a set of requirements together with our client that such a system would require, and built the first version in a span of 10 weeks. This version includes a matching module that matches flexworkers to project-managers based on their competences, a negotiation module that simplifies the process of creating a contract and a reporting module where financial and compliance data can be inspected on an organizational-, project- and personal level. The final product we delivered during this project has basic versions of these systems implemented as a modern, responsive web-application. It is far from production ready, but it is demo-ready for potential customers, and of good quality to form a strong foundation for creating a production-ready platform. We recommend that for further development, the focus should be in four directions: the matching algorithm, the compliance module, user testing and non-functional requirements as maintainability and scalability. Each of these directions should be a cycle of continuous improvement.Electrical Engineering, Mathematics and Computer ScienceSoftware EngineeringComputer Scienc
Wind speed measurements using distributed fiber optics: a windtunnel study
No full textNear-surface wind speed is typically only measured by point observations. The Actively Heated Fiber-Optic (AHFO)
technique, however, has the potential to provide high-resolution distributed observations of wind speeds, allowing for better
characterization of fine-scale processes. Before AHFO can be widely used, its performance needs to be tested in a range of
settings. In this work, experimental results on this novel observational wind-probing technique are presented. We utilized a
controlled wind-tunnel setup to assess both the accuracy and the precision of AHFO under a range of operational conditions.
The technique allows for wind speed characterization with a spatial resolution of 0.3 m on a 1 s time scale. The flow in the wind
tunnel was varied in a controlled manner, such that the mean wind, ranged between 1 and 17 ms-1. The AHFO measurements
are compared to sonic anemometer measurements and show a high overall correlation (0.85-0.98). Both the precision and
accuracy of the AHFO measurements were also greater than 95%.We conclude that the AHFO has potential to be employed as an outdoor observational technique. It allows for characterization of spatially varying fields of mean wind in complex terrain,
such as in canopy flows or in sloping terrain. In the future, the technique could be combined with conventional Distributed
Temperature Sensing (DTS) for turbulent heat flux estimation in micrometeorological/hydrological applications
- …
