Examensarbeten för masterexamen

Länka till denna samling:

https://odr.chalmers.se/handle/20.500.12380/44

Browse

Visar 1 - 5 av 427

Numerical Methods for mapping band-type resonance in insect flight
(2025) Zhang, Congxiao; Chalmers tekniska högskola / Institutionen för matematiska vetenskaper; Beilina, Larisa; Pons, Arion
Insect flight is a highly complex and energy-intensive process. Flapping-wing insects employ a unique muscle contraction mechanism that enables high-frequency wing beats, with metabolic rates reaching several times those at rest. Their remarkable endurance during flight highlights the importance of understanding the energy optimization involved. This thesis focuses on developing numerical methods to map band-type resonance, which serves as a benchmark for assessing whether a system achieves an energy-optimal state. We describe the mapping of band-type resonance as an optimization problem and propose two primary numerical methods: particle swarm optimization and numerical continuation. We evaluate the accuracy of the numerical solutions via the solution work loops and power waveforms and compare them with analytical approximations to the space of band-type resonant states. Our findings reveal that while the standalone particle swarm method can provide a relatively complete set of estimated solutions, the solution space lacks continuity. The numerical continuation method sacrifices some completeness in finding solution sets to ensure better continuity in the corresponding domain of the output solution set. After comparing these methods' performance in identifying potential solutions for simple cases, we improve them and propose a compound numerical method for solving more complex problems, such as higher harmonic and nonlinear oscillators. Notably, this compound algorithm performs well not only on simple linear cases with known analytical solutions but also on complex problems lacking analytical solutions, offering a valuable numerical tool for estimating the mapping zone of band-type resonance when analytical methods are not feasible. Comparing the results of mapping zones of band-type resonance with wingbeat frequency modulation behaviour observed in actual insect species suggests that such behaviour may be consistent with sustained resonant energy savings by exploiting band-type resonance. This report is written in English.
Large-Scale Transformer-Based Multi-Target Tracking
(2024) Spjuth, Oliver; Chalmers tekniska högskola / Institutionen för matematiska vetenskaper; Andersson, Adam; Andersson, Adam; Svedung Wettervik, Benjamin
In military surveillance, radar-based tracking of objects is essential. The growing use of small-scale drones, as seen in the Russia-Ukraine war, necessitates tracking at low speeds. At these speeds, birds are also detected and the number of false detections increases, making the already complex Multi-Target Tracking (MTT) problem more challenging. Recent advances in machine learning, particularly the transformer ar- chitecture, present new opportunities to address these challenges, making it valuable to explore their application in air surveillance contexts. Although transformers have shown promise in related fields such as automotive radar, adapting them to air surveillance presents specific hurdles. These include managing the quadratic scaling of attention as the number of detections increases, ensuring accurate state estimation across large continuous areas, and simultaneously estimating a large number of targets. To address these challenges, a four-module pipeline was developed. The first module reduced attention complexity by generating local contexts of detections for paral- lel processing. This was followed by a transformer-encoder-based filter designed to eliminate false detections (FDF). Next, the original problem was partitioned into independent subproblems using a graph-based clustering approach. One suggested implementation utilized the attention scores from the FDF to construct edges be- tween detections (nodes). The Leiden algorithm, a community detection algorithm, was then applied to identify clusters of related detections. These clusters were sub- sequently processed in parallel by the final transformer-based MTT module. This approach significantly reduced the initial memory demands of attention from approximately 320 GB to 1.6 GB while maintaining performance across the pipeline. The false detection filter achieved a balanced accuracy and F1 score of 99%, ef- fectively reducing the problem complexity. The attention-score-based partitioning method accurately identified subproblems that were predominantly optimal (single- target) or near-optimal. When evaluated using MTT metrics, the pipeline employing the attention-score- based partitioning method demonstrated promising results, with few missed or false detections and a total inference time of approximately 0.5 seconds for over 100,000 detections. The system scaled effectively with increased complexity and adapted well to varying conditions.
The Role of Wnt Signaling in an Embryoid Body Generation Protocol for Hematopoiesis Studies: Investigating the Emergence of Endothelial and Hematopoietic Stem and Progenitor Cells in hiPSC-Derived Embryoid Bodies through RT-qPCR and Flow Cytometry Analysis
(2024) Hilpold Berntsson, Elin; Chalmers tekniska högskola / Institutionen för matematiska vetenskaper; Kristiansson, Erik; Guibentif, Carolina
Research indicates that childhood leukemia originates in utero during hematopoietic development, emphasizing the importance of understanding embryonic hematopoiesis to identify disruptions leading to diseases such as leukemia. However, studying human embryonic hematopoiesis in vivo faces significant ethical and technical challenges. As an alternative, embryoid bodies (EBs) derived from human induced pluripotent stem cells (hiPSCs) offer a promising in vitro model for studying human developmental processes and hematological diseases. This thesis evaluates an established EB generation protocol for its efficacy in differentiating hiPSCs into hematopoietic stem and progenitor cells (HSPCs). Through the application of real-time quantitative PCR, this study investigates the activation of the Wnt signaling pathway, marked by RSPO3 gene expression, assessing its impact on cell fate decisions within the EBs. Key findings reveal that activation of the Wnt pathway not only induces the emergence of endothelial cells but also facilitates the specification of HSPCs. The robust upregulation of genes such as SOX17, CDH5, RUNX1, GATA2, and HOXA9 over time validates the progressive development of endothelial and hematopoietic lineages. Flow cytometry analysis results also confirm the emergence of endothelial and blood cell populations. However, the response of differentiating hiPSCs to Wnt signaling varies between independent experiments, suggesting that further optimization of the EB generation protocol can be implemented. Overall, the results underscore the role of Wnt signaling in lineage specification, offering valuable insights into hematopoietic differentiation and development in a controlled environment.
Using Transformer-based Neural Networks for classifying cellular states in Glioblastoma
(2024) Hedberg, Ronja; Chalmers tekniska högskola / Institutionen för matematiska vetenskaper; Jörnsten, Rebecka; Jörnsten, Rebecka; Lozada Cortés, Alejandro
By taking inspiration from the progress made in Natural Language Processing with the use of Transformer-based Neural Networks, similar approaches have been proposed for single-cell RNA-sequencing data in hope of capturing complex gene-to-gene interactions. One such approach is the pre-trained single-cell bidirectional encoder (scBERT), whose architecture and pre-training follows its Natural Language counterpart, BERT. Unlike BERT, scBERT was pre-trained for masked gene expression prediction using single-cell datasets comprising over 1.5 million single-cell RNAsequencing profiles. This thesis performs an initial assessment of the use of scBERT with novel single-cell data. In classifying annotated cellular states of Glioblastoma, the inclusion of scBERT showed overall limited advantages compared to using the gene expression directly. However, through the simulation of different scenarios, this thesis provides preliminary evidence in favor of the use of scBERT in the lack of ample signal (low number of expressed genes, and scarce number of training examples). This showcases the potential benefits of using the gene representations of massive single-cell Transformer-based models, especially when little information is available, which is frequently the case when working with in-house data or heavily underrepresented cellular states.
Machine Learning-Enhanced Column Generation for the Crew Pairing Problem: Leveraging Gated Recurrent Units to Generate the Initial Restricted Master Problem
(2024) Eriksson, Wilmer; Chalmers tekniska högskola / Institutionen för matematiska vetenskaper; Jonasson, Johan; Grover, Divya; Jonasson, Johan
The crew scheduling problem is a critical challenge for airlines as crew costs represent the second largest operating cost. It consists of determining the optimal assignment of crew members to each flight leg in an airline’s schedule. The problem is broken into two separate problems that are solved sequentially. This research focuses on the initial problem, referred to as the crew pairing problem, where sequences of flight legs, known as pairings, are created. Traditionally, the crew pairing problem is solved by Integer Column Generation frameworks such as the branch-and-price algorithm. This work proposes the integration of Gated Recurrent Units (GRU) into the stateof- the-art industry algorithms used to solve the crew pairing problem. This method, named the GRU-Enhanced Pairing Initializer (GEPI), generates an initial potential partial solution using supervised learning before starting the traditional column generation process. For large-scale optimization problems, a fundamental trade-off between optimality and computational time exists. Sometimes, problems are not solved to complete optimality; instead, satisfactory or near-optimal solutions are found. Therefore, by improving the starting point of the algorithm, it should be possible to find better solutions. The results shows that GEPI can generate high-quality pairings that are used in the final solution for 59 out of the 192 test cases. These GEPI-generated pairings reduced the flight schedule cost in 34 test cases, while two cases saw a cost increase. The introduction of GEPI in the optimization process led to an increase of execution time for 37 test cases. These outcomes suggest that there is some potential of improving the starting point of the column generation process using a neural network. However, the decrease in scheduling cost appears to be associated with longer execution times. To gain a more comprehensive understanding of GEPI’s potential and limitations, further tests and analysis are needed.

Browse

Senast publicerade