[dipl] Perustieteiden korkeakoulu / SCI

Permanent URI for this collectionhttps://aaltodoc.aalto.fi/handle/123456789/21

Browse

Now showing 1 - 20 of 5242

2D metal-organic framework of dicyanoanthracene on Au(111)
(2019-03-12) Pohjavirta, Ilona
Perustieteiden korkeakoulu | Master's thesis
Two-dimensional metal-organic frameworks (2D MOFs) are single-atom-thick porous lattices comprising coordinating metal atoms and molecular ligands on an atomistically flat surface. Every component in the system can be varied, resulting in diverse structures exhibiting numerous different lattice symmetries as well as a rich variety of physics. 2D MOFs have been predicted to host exotic quantum-mechanical phenomena, perhaps most interestingly topological phases which make them especially appealing for applications in e.g. spintronics and quantum computing. However, despite numerous theoretical predictions, the realization of topological electronic states in 2D MOFs remains elusive. This thesis focuses on a 2D MOF of gold centers and 9,10-dicyanoanthracene (DCA) molecules in a kagome lattice symmetry - a structure predicted to be a topological insulator. The synthesis of the Au-DCA lattice onto a Au(111) surface was demonstrated and the structure was characterized by scanning tunneling microsopy (STM) and scanning tunneling spectrosopy (STS). The results obtained point towards an extensive electronic system throughout the lattice, perhaps even band formation. However, challenges in synthesis resulting in limited sample size as well as the metal substrate used prevent from drawing further conclusions. After this first step, the future research towards realizing the topological phase in the Au-DCA lattice is outlined by enhancing the lattice growth as well as synthesis onto a weakly interacting substrate.
3-D Motion-Tracking for High-Intensity Focused Ultrasound Therapy
(2012) Suomi, Visa
Perustieteiden korkeakoulu | Master's thesis
Respiratory motion during high-intensity focused ultrasound (HIFU) therapy reduces the efficiency of the treatment in abdominal organs. Hence, for the purpose of motion-correction during HIFU therapy, two different ultrasound-based 3-D motion-tracking methods were presented and adapted to the existing Philips Sonalleve MR-HIFU platform. The displacement estimation accuracies of these two techniques were determined using a metal pin and an in vitro tissue sample as targets. The measurement data was collected on all movement directions using element clusters consisting of one, three and 32 transmitting transducer elements. Simulations of the acoustic fields were also performed in order to discuss the theoretical limitations of the motion-tracking with the existing HIFU system. The displacement estimation accuracies did not differ significantly for the two different techniques introduced but was rather dependent on the transmitting element cluster size and the orientation of the ultrasound beam axis. Using smaller number of elements and defining the beam axis angle accurately in the calculation algorithm yielded more stable results. The motion-tracking using the in vitro tissue sample was significantly more difficult to achieve than with the metal pin target. This was due to the incorrect time-shift values given by the cross-correlation algorithm. Hence, the failed channels had to be manually excluded from the calculations in order to yield the correct displacement estimation values. The introduced motion-tracking methods cannot be readily used as such during clinical HIFU therapy treatment, because the incorrect time-shift values had to be manually excluded from the calculation algorithm. This process could probably be automated by observing the peak count and the amplitude of the cross-correlation curves. However, this method was not verified in this thesis and hence requires further research about its feasibility.
3D Gaussian splatting theory and variance rendering extension
(2024-09-12) Taka, Veikka
Perustieteiden korkeakoulu | Master's thesis
Radiance field methods are a recent approach for novel-view synthesis. Radiance field methods reconstruct a representation of a scene from a set of images that capture the scene from multiple viewpoints. This scene representation can then be used to render photorealistic images from novel viewpoints. In 3D Gaussian Splatting (3DGS), the scene representation consists of hundreds of thousands of overlapping multivariate Gaussian distributions in three-dimensional space, centred on the discrete point samples of the underlying continuous density distribution. An image is rendered by projecting these Gaussians into the camera's image plane, sampling the 2D Gaussian projections at each pixel location, and compositing the sample values in depth order. 3DGS addresses the limitations of the preceding state-of-the-art method, namely Neural Radiance Fields, facilitating rapid reconstruction of the scene representation and real-time rendering. This thesis introduces a statistical formulation of the splatting equation, the image formation model of 3DGS, as a weighted arithmetic mean. This allows the compositing of generic point cloud quantities, such as depth, into a properly normalised image. Furthermore, it permits the rendering of a variance image, which describes the variance of the rendered Gaussian quantity at each pixel. The mathematical formulae and pseudocode are provided for the implementation of the variance image renderer and for the computation of analytical gradients through the variance image back to the scene parameters. Furthermore, a variance image renderer has been implemented on top of Splatfacto, an open-source implementation of 3DGS.
3D geometric correction in 2D multi-slice acquired MR images used in radiation therapy planning
(2020-03-16) Haukipuro, Eeva-Sofia
Perustieteiden korkeakoulu | Master's thesis
The role of MRI in the external radiation therapy treatment has increased in the past few decades significantly due to its superior soft-tissue contrast and absence of ionizing radiation. The two challenges concerning the lack of electron density values and limited geometric accuracy to use MR images in RT have been overcome. MR images can either be aligned with CT images or converted to pseudo-CT images for radiation dose planning, and the required geometric accuracy can be achieved by optimizing sequences and utilizing geometric correction algorithms. The purpose of this thesis was to improve the geometric accuracy of MR images by implementing 3D geometric correction to 2D multi-slice acquired images. Until now, this feature has not been available in Philips MR scanners. The advantages of 2DMS acquired images over 3D images are the reduced sensitivity to motion artifacts and the contrast that is preferred by radiologists. The geometric correction algorithm itself was not modified, since the existing algorithm for 3D images could be reused. The performance and the geometric accuracy achievable with the algorithm were evaluated with phantom imaging and the effect of the algorithm on human anatomy was evaluated with volunteer imaging. Altogether 21 subjects were imaged with Ingenia 1.5 T, Ingenia 3 T and 1.5 T MR scanner of Elekta Unity. The results reveal that the geometric distortion is significantly reduced when through-plane correction is applied alongside in-plane correction. The greatest effects are shown on the edges of the field-of-view where the effect of gradient non-linearities are the largest. It was also proven that the algorithm performs the geometric correction almost as well in 2DMS as in 3D images.
3D mapping and change detection for patrolling mobile robots
(2021-08-23) Göncz, Levente
Perustieteiden korkeakoulu | Master's thesis
The goal of this thesis was twofold: First, the capabilities of the ZED 2 stereo vision camera were to be assessed to determine whether the camera is suitable to use in real-time applications. Second, the thesis investigated the possibility of implementing a real-time change detection system using the camera. The camera came with very little information regarding its performance. Therefore, in this thesis, a large set of experiments were carried out to evaluate the performance of the camera with respect to depth accuracy and consistency, visual tracking and relocalization accuracy, and object detection performance. Through testing, it has been verified that the camera is able to provide fast and accurate depth measurements, 3D position information and object detection data. Thus, the ZED 2 stereo camera is suitable for real time applications. Furthermore, a 3D change detection system was designed which was built around the concept of semantic object maps using the ZED 2 stereo camera. The proposed algorithm is able to maintain a coherent semantic object map of the environment and detect object-level changes between consecutive patrol routes.
4D dose calculation in pencil beam scanning proton therapy
(2021-10-19) Hyytiäinen, Tero
Perustieteiden korkeakoulu | Master's thesis
Proton therapy using pencil beam scanning (PBS) has become more common in cancer treatment during the recent decades. Motion due to the patient's respiration still poses a problem for high-precision PBS treatments, and there is a need to quantify and predict the effects of respiratory motion with greater accuracy. A possible solution to this problem is the use of 4-dimensional (4D) dose calculation. 4D dose calculation aims to take into account the breathing motion of the patient, and it could be used to aid treatment planning and adaptations between fractions. In this thesis, the required background knowledge to understand 4D dose calculation is presented, after which a workflow for both retrospective 4D dose reconstruction using treatment log files, and for prospective 4D dose calculation utilizing a treatment simulation is introduced. The workflows are also carried out in two experiment cases using artificial 4D CT images and a breathing signal. The results indicate a significant difference to the reference dose in all cases, with an increase in movement amplitude corresponding to a larger difference. The effects of the interplay effect are also visible as dose hot and cold spots in all cases, but the magnitude of the interplay is not found to be affected by the motion amplitude. The results of prospective 4D dose calculations are also found to differ from the retrospective dose reconstruction, although the magnitude of the breathing motion's effects is found to be similar in both cases.
5G microservice monitoring with Linux kernel
(2019-06-17) Oksanen, Ilkka
Perustieteiden korkeakoulu | Master's thesis
Software industry is adopting a scalable microservice architecture at increasing pace. At the advent of 5G, this introduces major changes for the architectures of telecommunication systems as well. The telecommunications software is moving towards virtualized solutions in form of virtual machines, and more recently, containers. New monitoring solutions have emerged, to efficiently monitor microservices. These tools however can not provide as detailed view to internal functions of the software than what is possible with tools provided by an operating system. Unfortunately, operating system level tracing tools are decreasingly available for the developers or system administrators. This is due to the fact that the virtualized cloud environment, working as a base for microservices, abstracts away the access to the runtime environment of the services. This thesis researches viability of using Linux kernel tooling in microservice monitoring. The viability is explored with a proof of concept container providing access to some of the Linux kernels network monitoring features. The main focus is evaluating the performance overhead caused by the monitor. It was found out that kernel tracing tools have a great potential for providing low overhead tracing data from microservices. However, the low overheads achieved in the networking context could not be reproduced reliably. In the benchmarks, the overhead of tracing rapidly increased as a function of the number of processors used. While the results cannot be generalized out of the networking context, the inconsistency in overhead makes Linux kernel monitoring tools less than ideal applications for a containerized microservice.
5G-enabled digital transformation in the Finnish forest industry
(2023-06-14) Laiho, Perttu
Perustieteiden korkeakoulu | Master's thesis
Digital transformation is a change currently happening in several domains in society including organizations, sectors, industries, and nations. For companies, digital technologies enable an opportunity to improve competitiveness by transforming, for example, business models and processes. 5G is a central part of this transformation as it’s the first wireless technology able to support the novel Industry 4.0 (I4.0) technologies in the forest industry. The goal of this study is to understand the key aspects of 5G adoption in the digital transformation of a forest industry enterprise’s production. To achieve this goal, the thesis adopted an inductive single-case study approach with five embedded units of analysis representing 5G-based use cases for a forest industry production plant. Semi-structured interviews were used as a primary data source to determine the key characteristics of each use case such as available benefits, adoption prerequisites, and barriers hindering the adoption. This data was used to establish a conceptual model for assessing 5G-based use cases. Within- and cross-case analysis was then applied to the use cases by following the conceptual model to draw the main findings. Finally, the results were evaluated against the existing literature. The adoption of 5G was found to be closely linked to its technological complements, I4.0 technologies. These complements induce prerequisites and barriers that must be considered during the adoption decision because covering only the 5G network’s similar aspects doesn’t reveal the overall situation. It was also found that 5G is likely not adopted if it can’t be linked to a novel benefit that has not been available with legacy technologies. This is because merely changing the network wasn’t found to improve the organization’s performance as the complements are responsible for the direct performance improvement whereas 5G connectivity only enables their implementation. The findings imply that forest industry enterprises should assess 5G adoption as a combination of 5G connectivity and necessary complements to implement a functioning use case. As 5G and its complements are not yet mature technologies, the adoption should first focus on the simpler and already available use cases to access the first benefits. At the same time, they should prepare to adopt more advanced solutions available in a few years by establishing partnerships, transforming their organization including ICT systems, establishing collaboration opportunities for employees, and acquiring necessary organizational knowledge. Finally, the results imply that the biggest productivity improvements lie in use cases increasing the level of automation or decreasing the equipment downtime as those bear the largest impact on the production process and thus, the financial upside of the adoption.
A data flow management methodology for component based web application
(2018-06-18) Ruan, Yulong
Perustieteiden korkeakoulu | Master's thesis
Nowadays, the codebase of client web applications is becoming larger and more complex. New applications are built based on the web technologies and legacy applications are migrated from old systems to web platform. A modern client web application handles much more logics compared with the conventional web application which is rendered on the server side. Due to the increasing complexity of client web application, traditional ways of designing and modeling UI applications such as MVC pattern, are not applicable for developing large and scalable modern client web applications. The goal of this thesis is to propose an approach to manage the increasing complexity of client application and implement a web application development framework. I will introduce the concept of state machine and apply it to the web application development. Around the state management, I will review the relevant technologies such as Virtual DOM, Immutable Data Structure and One-Way Dataflow and use them as the basis of the framework. The target of the framework is to simplify the application state management in large system and decrease the complexity when adding/removing/modifying features from an existing codebase.
A/B testing and usability assessment methods in small companies
(2014) Luzik, Maksim
Perustieteiden korkeakoulu | Master's thesis
Usability assessment methods have been considered for long as useful tools to improve usability of the product or service. Many small companies are still reluctant to practice usability assessment methods due to many reasons including limited resources and management bias. A/B testing is a new movement in a corporate world. Multiple companies are considering to begin practicing the method or are practicing it already. Moreover, some people assume that A/B testing can actually cover all the usability assessment methods. This work provides answers to what A/B testing and usability assessment methods are and how they can be practiced in a small company (10-100 employees). Additionally the strengths and weaknesses of both methods are presented and compared. Finally suitability of the methods will be discussed. The target of optimization in this paper is an application of A/B test into a complex enterprise system. Only common usability assessment methods will be evaluated that are used today. These include standard usability testing, heuristic evaluation and cognitive walkthrough. No other usability assessment methods are included in tnis paper. Both practices: usability assessment methods and A/B testing have a place in a project. Despite their different approach they often complement each other and give answers to different questions. Usability assessment methods tend to answer to question why while A/B test usually answers to question which or how many. Usability assessment methods work well as an initial evaluation while A/B test can be used as a follow up method to fine-tune product or service even further.
Aalto Data Repository: Research data management, sharing and publishing in the world of data intensive science
(2016-01-18) Nurmela, Miro
Perustieteiden korkeakoulu | Master's thesis
All fields of science are becoming data intensive. The decrease of computing price and the evolution of data collection methods have created novel research opportunities. This new data intensive paradigm puts a new kind of premium to research data, since it more than ever before forms the lifeblood of research. As a result the demand for publishing research data has increased from both funding bodies and the research community. These factors combined present novel challenges for research data management and publication. This thesis sheds light on the current status of research data management, sharing and publishing. The primary contribution of the thesis is the examination of existing technical solution to these research data challenges. In addition requirements for successful research data solutions are proposed. The secondary contribution of the thesis is the research of the cultural atmosphere surrounding research data management, sharing and publishing. Technical solutions for the three research data challenges were found mainly from within the open source community. Solutions like Dataverse, Invenio, Hydra Project and CKAN offer platforms for sharing and publishing data. Solutions like iRODS can be used to manage research data. These solutions serve their purpose, but there is no good integrated solution that would solve all three research data challenges. The lack of holistic solutions combined with the lack of culture and knowledge about research data management result in limited research data publishing and sharing. Future work should, in addition to building an integrated solution for sharing, publishing and managing research data, aim to make the culture around research data management more open.
Abstract Data Visualisation in Mobile VR Platforms
(2018-12-10) Christaki, Kyriaki
Perustieteiden korkeakoulu | Master's thesis
Data visualisation, as a key tool in data understanding, is widely used in science and everyday life. In order data visualisation to be effective, perceptual factors and the characteristics of the display interface play a crucial role. Virtual Reality is nowadays accepted as a valid medium for scientific visualisation, because of its inherent characteristics of real-world emulation and intuitive interaction. However, the use of VR in abstract data visualisation is still limited. In this research, I investigate the use and suitability of mobile phone-based Virtual Reality as a medium for abstract data visualisation. I develop a prototype VR Android application and visualise data using the Scatterplot and Parallel Coordinates methods. After that, I conduct a user study to compare the effectiveness of the mobile VR application compared to a similar screen-based one by implementing some data exploration scenarios. The study results, while not being statistically significant, show improved accuracy and speed in the mobile VR visualisation application. The main conclusions are two-fold: Virtual Reality is beneficial for abstract data visualisation, even in the case of limited processing power and display resolution. Mobile VR, an affordable alternative to expensive desktop VR set-ups can be utilized as a data visualisation platform.
Academic Dissemination and Exploitation of a Clean-slate Internetworking Architecture: The Publish-Subscribe Internet Routing Paradigm
(2010) Ain, Mark
School of Science | Master's thesis
Accelerated Real-Time Volumetric Cloud Rendering with Distance Maps
(2023-05-15) Gröndahl, Janne
Perustieteiden korkeakoulu | Master's thesis
Accelerating Bayesian Optimization Structure Search with Transfer Learning
(2021-01-25) Sten, Nuutti
Perustieteiden korkeakoulu | Master's thesis
In this thesis I studied the use of transfer learning for reducing the computational cost of Bayesian optimization structure search (BOSS). BOSS combines Gaussian process surrogate models and first principle simulations to model the potential energy surface of atomistic structures. The aim is to find the global minimum of the surface at highest possible accuracy while minimizing consumed resources. BOSS can provide accurate results with limited number of simulations. However, significant number of simulations are still required for studying many-atom structures. I studied whether transfer learning can reduce the number of expensive, high fidelity simulations in BOSS without compromising the accuracy of the global minimum estimate. In transfer learning, output from a machine learning problem is used to initialize another one. The idea in this work was to take low fidelity simulation data and use it to initialize BOSS on high fidelity search. I studied a particular transfer learning solution for Gaussian processes implemented in BOSS in my earlier work. In this thesis I refined and evaluated the application of the method in BOSS with a series of computational tests. I found that the method can significantly reduce the computational costs of finding the potential energy minimum of a high fidelity task.
Accelerating Convolutional Neural Network Inference on Digital Signal Processor
(2022-03-22) Ollila, Uula
Perustieteiden korkeakoulu | Master's thesis
The ongoing deployment of 5G NR is to bring a completely new wave of technology and revolutionize wireless data transfer. The new standard will provide improved data rates, lower latency, better energy efficiency and larger capacity for connected devices. New performance requirements and increased flexibility challenge conventional physical layer (L1) solutions and boost research of new algorithmic approaches. One proposed new approach is deep learning (DL), which in recent studies have shown to provide a feasible alternative for existing algorithms in terms of computational load and accuracy. So far, neural networks (NNs) based solutions are not available in any commercial 5G physical layer product. They employ highly specialized hardware, which poses challenges when efficient NN inference implementations are considered. Convolutional neural networks (CNNs) are an essential subtype of NNs and have recently been introduced to replace several L1 processing blocks. This work aims to develop a comprehensive framework that enables accelerating CNN inferences on CEVA-XC4500 DSP, a current state of the art in L1 processing. The basic idea of the framework is presented earlier for multilayer perceptrons (MLPs), and this work elaborates the idea for 1-D CNNs. Essential parts of CNNs are covered, including convolution layers, pooling layers and some activation functions. The DSP implementation is optimized with the single instruction, multiple data (SIMD) intrinsics that the DSP offers broad support. Since the DSP is designed for fixed-point arithmetic, the framework includes a procedure for quantizing the network parameters as a part of the framework's offline preprocessing part. Implementation's real-time performance is evaluated by recording the processor cycle counts of running inference with five different CNN models. The results are recorded in a software simulator that simulates the operation of the DSP in a cycle-accurate manner. The implementation's accuracy and quantization procedure are also evaluated. The results show that the target DSP can accelerate optimized CNN inferences effectively. Compared to the unoptimized reference implementation, speedups ranging from 6.5 to 26.8 were observed. Most performance gains originate from optimizing the convolution layers, typically computationally the heaviest part of CNNs. Clear benefits can also be observed from optimizing pooling layers and activation functions.
Accelerating Digital Transformations
(2021-08-25) Satharasi, Rakesh
Perustieteiden korkeakoulu | Master's thesis
Digital transformations happening and planned across businesses are the key to unlock the digital potential across the globe. According to (Jacques Bughin et al., 2016), there is a lot of digital potential yet to be realized across the globe. For example, Europe has realized only 12% of its digital potential. Europe’s economy is experiencing the early impact of digitization, with some correlation between productivity growth and digital intensity across sectors. The impact on the labor market is mixed with widespread dislocation of workers but a proliferation of digital tools that offer new ways of working, matching skills, and acquiring skills. Another important fact stated by (Rogers, 2016) is that 84% of digital transformations fail. While there are various reasons for this, in this study, we examine the different digital transformation cases and arrive at a list of dos and don’ts, which will help organizations increase the success chances of their digital transformation and at the same time help them in accelerating the benefits expected from these transformations. As part of this work, we interviewed industry experts and practitioners who are playing a pivotal role in their organization’s transformation journey to gather key practical insights from both the macro and micro lens of digital transformation. Some of the key findings in this study underlines the importance of having a clear strategy and end to end vision, in order to succeed in digital transformations. The study also highlights that these initiatives must be people centric in order to have better adoption and better success. All the findings from the study are analyzed using scientific methods explained later in the document and conclusions are inferred through established and well know scientific methods.
Accelerating simultaneous localization and mapping
(2020-08-18) Karppinen, Petri
Perustieteiden korkeakoulu | Master's thesis
In simultaneous localization and mapping (SLAM), maps of environment are used to localize vehicle accurately while the estimated position is simultaneously utilized to build and update the maps. SLAM can be used to obtain position estimates for autonomous mobile robots in areas where other localization methods, such as global navigation satellite system, do not give reliable positioning estimates. This thesis concentrated on a SLAM implementation that uses normally distributed transforms occupancy map (NDT-OM) framework to build the maps. The SLAM implementation was not working in real-time in outdoor environments. Thus, the goal of this thesis was to accelerate the SLAM implementation to achieve this requirement. In this work, the slowest parts of the SLAM were identified and accelerated. These parts were map building and updating, and a registration process for iteratively obtaining improved position estimates by matching maps together. Program optimization and parallel computing methods were utilized to accelerate these algorithms. By using the proposed methods, real-time performance was achieved, with approximately three-fold speedup in comparison with the current SLAM implementation.
Acceleration of Innovation Processes in the ICT Sector
(2010) Taajamaa, Mauno
School of Science | Master's thesis
The goal of this thesis is to study the acceleration and success factors of business model innovation processes within a network of companies in the ICT sector. In the literature review general perspectives to the field and to the context are presented. Then the factors regarding the speed and the success of innovation processes are studied in more detail. A case study was used to evaluate, verify and deepen the findings from the literature. The subject of the case study was a network based business innovation development project in the ICT sector. The case study was conducted by interviewing the participants and relevant persons outside the project. A linear and systematic innovation process, in the context of network development, gives only small input to minimise the throughput-time or to maximise the success of the innovation. Instead, parallel activities and the use of lead users in development work has a positive impact, especially on the speed of the process. Formal agreement among the participants is an essential factor not only to the success of the innovation, but to the lifespan of the network-based innovation project in the first place. Agreements should be done at the earliest possible phase. In a technology innovation, where the business model is based on services run by third parties with high end-user penetration, the commitment of the main third parties is crucial for the success of the innovation. The key third parties should be identified and involved in the development process as early as possible.
The acceptance of and capabilities for value-based exchange in manufacturing industry
(2022-01-26) Jaskari, Julius
Perustieteiden korkeakoulu | Master's thesis

Browse

Browsing [dipl] Perustieteiden korkeakoulu / SCI by Title