Aaltodoc

Aaltodoc is the institutional repository of Aalto University.

Aaltodoc has a new updated appearance. Instructions for searching and more information is available here.

Otakaari 1 grandhall. Photo: Esa Kapila
 

Communities in DSpace

Select a community to browse its collections.

Now showing 1 - 13 of 13

Recent Submissions

Velvet noise in audio processing
(2025) Fagerström, Jon
 School of Electrical Engineering | G5 Artikkeliväitöskirja
Noise plays a central role in various audio-processing applications, including artificial reverberation, audio decorrelation, acoustical measurements, sound synthesis, and speech processing. This dissertation focuses on applications of sparse noise, known as velvet noise. With its minimal density and smooth temporal envelope, velvet noise has been widely used in artificial reverberation algorithms, both as a core component as well as a decorrelating or diffusing element. The work within builds on this foundation by exploring novel variants and applications of velvet noise. The thesis introduces dark velvet noise as a low-passed variant, generalizes this to extended dark velvet noise for accurate modeling of non-exponential late-reverberation, and culminates in developing the binaural dark-velvet-noise reverberator. Additionally, short velvet noise filters are explored for decorrelation and variation filtering tasks, demonstrating their effectiveness in lowering inter channel correlation within feedback delay networks through velvet feedback matrices and generating realistic variations of sampled percussive sounds. The contributions of this thesis offer significant advances in applying velvet noise to audio processing, with particular emphasis on artificial reverberation and decorrelation.
Deep learning methods for modeling of spatiotemporal dynamical systems governed by partial differential equations
(2025) Iakovlev, Valerii
 School of Science | Doctoral dissertation (article-based)
This dissertation focuses on data-driven modeling of spatiotemporal dynamical systems, using observational data to develop models that approximate the underlying dynamical processes. Spatiotemporal modeling has a rich history with numerous successful applications. It has been continually advanced by technological and methodological improvements, evolving from early qualitative approaches to modern sophisticated deep learning methods. Despite recent progress enabled by deep learning—which has shown promise in modeling complex systems like weather patterns, traffic dynamics, and crowd flows—current deep learning-based spatiotemporal models face significant limitations. These include restricted applicability due to simplifying assumptions (such as fully observed states on fixed grids), data inefficiency requiring large datasets for good generalization, and long training times coupled with instabilities arising from complex loss landscapes. This dissertation addresses these challenges by developing novel deep learning-based models and techniques that enhance the flexibility, data efficiency, and stability of spatiotemporal systems modeling. To extend applicability of deep learning-based spatiotemporal models, a graph-based continuoustime model inspired by the method of lines is introduced, enabling modeling on irregular spatiotemporal grids. This is further extended to a space-time continuous model operating in latent space, allowing for learning dynamics from partially observed and noisy states. Finally, a model incorporating a spatiotemporal point process is developed to learn system dynamics from unstructured observations made at random times and locations. To improve data efficiency, the models leverage the locality bias inherent in PDE systems, achieving remarkable data efficiency and requiring significantly fewer training trajectories to generalize compared to previous methods. To enhance training stability and speed, an amortized Bayesian multiple shooting technique is proposed, extending classical multiple shooting to the Bayesian setting and modern computational regimes. This method stabilizes training and reduces training time by up to an order of magnitude. Additionally, a latent space interpolation technique is introduced to further accelerate training without compromising predictive accuracy. Overall, this dissertation advances the field of data-driven spatiotemporal modeling by introducing deep learning methods and techniques that are more widely applicable, data-efficient, and computationally efficient. These developments enable the modeling of a broader spectrum of complex dynamical systems under more realistic conditions than was previously possible.
Semiconductor nanowires on flexible plastic substrates
(2025) Khayrudinov, Vladislav
 School of Electrical Engineering | Doctoral dissertation (article-based)
Nanowires (NWs) offer exceptional potential for use in solar cells, lasers, LEDs, and photodetectors. In parallel, there's growing interest in flexible electronics due to their cost-effectiveness, lightweight design, mechanical resilience, and chemical stability. However, a major obstacle to integrating NWs with flexible electronics is the high temperatures typically required for metalorganic vapour phase epitaxy (MOVPE) growth, which makes NWs incompatible with most plastic substrates. This dissertation presents innovative techniques for the direct growth of III-V nanowires on flexible plastic substrates, along with methods to create functional nanowire-based flexible electronic devices. Firstly, this work establishes an isolated growth regime for self-catalysed InAs NWs, which enables their growth at record low temperatures. Extensive characterization reveals the ability to control crystal structure and NW density, which is particularly essential for achieving compatibility with lowtemperature processes necessary for flexible electronics. Next, the direct growth of III-V NWs on flexible plastic substrates is achieved. High-density, well-crystallized InAs and InP nanowires are grown on polyimide without pre-treatment, demonstrating strong mid-infrared emission for InAs and near-infrared emission for InP. Significantly, the electrical properties of these NWs allow for the fabrication of flexible nanowire-based p-n junction devices on plastic in just two fabrication steps. Additionally, this research marks the first successful growth of InSb nanowires on flexible plastic substrates. These NWs display high material quality, room temperature photoluminescence, and remarkable flexibility, showing promise for future flexible optoelectronics. Finally, a method for fabricating GaAs NW LEDs is presented. GaAs NWs are grown directly on flexible plastic substrates within the MOVPE reactor, showing zinc blende crystal structure, room temperature photoluminescence emission, and desirable optical properties. These findings open up possibilities for roll-to-roll compatible LED production using GaAs NWs, a key step in integrating these materials into a wide range of flexible optoelectronic devices. These advancements push the boundaries of flexible optoelectronics by enabling the direct growth of III-V nanowires on plastic substrates, while also lowering fabrication complexity and temperature requirements. The methods presented here pave the way for more versatile, durable, and efficient devices across a range of applications
Intelligence enablement and orchestration
(2025) Ramos, Edgar
 School of Electrical Engineering | Doctoral dissertation (article-based)
This work focuses on the enablement and distribution of intelligence all the way to devices and automation. It proposes multiple architectural frameworks, tools, and system solutions to address the needs analysed to facilitate intelligence adoption in devices. By the analysis of such requirements, the decoupling of intelligent services from the application layer and promoting interoperability between them utilizing semantic tools and models facilitates the management and operation of such intelligence, as well as promotes its development. The work conceives an intermediate stratum between the applications and the intelligence services named "intelligence layer" and the analysis of the required functionality and possible implementation mostly focuses on the intelligence enablement at the far edge (devices). Also, the interaction between multiple "intelligent agents" that focus on realizing their own task with their capabilities is developed towards a concept named "intelligence orchestration". A proposal for the architecture and main functions to realize such an endeavour is produced and the supported tools and limited testing and validation are documented. The systematization of the solution by analysis of requirements, state-of-the-art, and objectives to achieve taking into consideration the current ecosystems and applications was the main methodology to design the propositions presented. Also, some prototyping and analysis of the implementations guided how to improve and clarify the concepts and refine the ideas further.
Exploring behavioral economics in archival auditing research
(2025) Nickpour, Mohammadali
 School of Business | Doctoral dissertation (article-based)
This doctoral dissertation comprises an introduction and three self-contained essays that, using archival data, examine the role of innate personality traits, reputation concerns, and compensation incentives in shaping the judgment and behavior of individual auditors. The first essay investigates the relationship between the inherent characteristics of audit partners and audit quality. Drawing on political psychology literature, it argues that individuals' political activism reflects their extraversion and openness to experience. Using data on political campaign contributions in the United States, the study finds a consistent positive association between audit partners’ extraversion and openness to experience, as indicated by their political contributions, and various aspects of audit quality. These findings underscore the significance of considering individual differences in the recruitment and training processes within audit firms. The second essay explores the influence of reputation concerns on auditor behavior. It hypothesizes that the identification of individual engagement partners in the United States, as mandated by PCAOB’s Rule 3211, increases their reputation concerns. The results indicate that, among clients with existing or potential internal control material weaknesses, there is an increased likelihood of adverse internal control opinions in the post-Rule 3211 period. In addition, in the post-Rule 3211 period, clients with high internal control risk experience a significantly greater increase in audit fees and audit report delays compared to clients with effective internal controls. Furthermore, the study shows that adverse opinions increase the likelihood of early audit partner rotations within the same audit firm. These findings are valuable for regulators and audit firms seeking to understand how auditors’ reporting on internal controls has evolved in response to engagement partner identification and heightened reputation concerns. The third essay examines how individual audit partners respond to horizontal pay disparities among partners within audit firms. The study utilizes Belgian data, a setting in which most audit partners establish limited liability management entities through which they perform audit services and bill the audit firm. This structure allows for observing the actual partner compensations. The findings reveal that greater pay disparity among partners is associated with lower audit quality, with partners earning below the firm's average delivering lower-quality audits as overall compensation disparity increases. The results are more pronounced among male and early-career audit partners. Further analyses dismiss the alternative explanation that these findings are driven by auditors with lower ability or quality. This study provides practical insights for audit firms in designing effective compensation systems for partners.