Machine learning helps to learn from enormous amounts of data gathered as a direct or indirect product of a process, whether it be an industrial product line, efficient farming for food production, prediction of bacterial growth, or any other industrial, healthcare-related, or technological endeavor. Similar to humans, machine learning extracts patterns from data if enough examples are provided. Reinforcement learning is an action-oriented type of learning in which an agent tries to maximize a given reward signal without being told explicitly which actions to take. This type of learning is often combined with other types of machine learning. For instance, deep learning is widely applied in scenarios with large state spaces.
The topic of machine learning leads to projects of varying nature such as theoretical (algorithm development, computational efficiency, convergence of machine learning algorithm, and formal representation), algorithm training for classification or regression (see some example projects below), predictive maintenance, and reinforcement learning.
Related Courses: Data Science
Available Project Proposals
If you are interested in the general topic of Machine Learning, or if have your own project idea related to the topic, please contact us directly. Alternatively, you can also work on one of the following concrete project proposals:
- Point cloud classification, segmentation, and visulization
Supervisors: Faizan Ahmed
Point clouds are sets of spatial data points captured by 3D scanning techniques such as lidar. These point clouds contain many million points of data, resulting in 3D representations of the railway environment. Point cloud data can be used to create machine learning models that can classify the object in rail infrastructure automatically.
The main research question is: How to reliably segment, classify, and visualize point clouds for railway catenary systems with scant computational power, memory, and ground truth labels?
However, the student can focus on a combination of topics given below:
- leverage existing trained models for point cloud labeling
- pre-process (down-sampling. Augmentation, filtering, voxelization etc.) point clouds for accurate model training
- train deep learning models for point cloud segmentation and object detection
- analyze the variation in object sizes with relation to the accuracy of the trained models.
- analyze the effect of point cloud density on accuracy
- determine the accuracy of the trained model objectively (or What are the excellent metric for such machine learning tasks?)
- connecting point clouds with existing 3D object libraries
- Optimising path planners using temporal specifications
Optimising path planners using temporal specifications
Contact person: Ernst Moritz Hahn, e.m.hahn@utwente.nl
Title: Optimising path planners using temporal specifications
Description: The master thesis is in the direction of optimising global path planners for autonomous driving vehicles. Global path planners synthesise vehicle trajectories to go from point A to point B while avoiding obstacles. Although this task seems simple, the complication occurs when the trajectories generated must follow rules of engagement and avoid obstacles (static and dynamic).
The goal is to utilise the previous work [1] on optimising robot trajectories published at IROS 2022 and adapt it for an autonomous driving application. Your tasks would be,
- to replicate the results from [1],
- extend Signal Temporal Logic specifications to work for one of the autonomous driving applications,
- setup machine learning based global path planner and,
- optimise path planner trajectories based on logical constraints.
[1] Dhonthi, Akshay, et al. "Optimizing demonstrated robot manipulation skills for temporal logic constraints." 2022 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). IEEE, 2022.
- Condition assessment of sewer and water pipe networks using data-driven models
Supervisors: Lisandro Jimenez Roa, Mariëlle Stoelinga
Assessing the condition of sewer and water pipe networks is vital for maintaining the reliability and safety of these essential infrastructure systems. Ageing and deteriorating pipes can cause various issues, such as leaks, breaks, and failures, leading to significant environmental and economic impacts. This project aims to investigate machine learning models for condition assessment in sewer and water pipe networks. These models will analyze historical inspection records to provide predictions of pipe conditions. Students will select and concentrate on one of the following research areas:
- Data Quality and Availability: This research area addresses the challenges arising from incomplete, inaccurate, and inconsistent data on sewer and water pipe conditions, which hinders the effectiveness of condition assessment and prediction models. The student will explore strategies for enhancing data quality and availability, such as data fusion techniques or crowdsourcing data. A tangible outcome will be a data quality improvement framework customized for the specific case study.
- Predicting Water Pipe Failure: This research area aims to develop more accurate models for predicting water main breaks to reduce service disruptions and enhance overall infrastructure management. Students will investigate machine learning or stochastic models for water pipe failure prediction, including random forests, decision trees, support vector machines, or Markov chain models. A tangible outcome will be a predictive model that can accurately forecast water pipe failures.
- Robust Sewer Pipe Deterioration Models: This research area focuses on constructing (or improving existing) models for predicting sewer pipe conditions by developing data-driven models to predict the type and severity of damage in sewer pipe networks. Emphasis will be placed on model validation and prediction uncertainty, contributing to more robust condition assessment and maintenance decision-making. Students will examine techniques such as Bayesian networks, Gaussian processes, ensemble learning, neural networks, decision trees, Markov chains, or regression models. Tangible outcomes will include an enhanced prediction model with quantified uncertainty estimates and a validated sewer pipe deterioration model that can predict damage types and severities with a measurable level of confidence.
Case Study: The project will centre on a real-world case study focusing on either sewer or water networks. The case study will involve historical inspection data, including pipe covariates (material, geometry, location, etc.), damages, and severities. This practical approach will help validate the effectiveness of the developed models and ensure their applicability to real-world situations.
Some literature of interest:
- Hawari, Alaa, Firas Alkadour, Mohamed Elmasry, and Tarek Zayed. 2020. 'A state of the art review on condition assessment models developed for sewer pipelines', Engineering Applications of Artificial Intelligence, 93: 103721.
- Laakso, Tuija, Teemu Kokkonen, Ilkka Mellin, and Riku Vahala. 2018. 'Sewer condition prediction and analysis of explanatory factors', Water, 10: 1239.
- Nguyen, Lam Van, and Razak Seidu. 2022. 'Application of Regression-Based Machine Learning Algorithms in Sewer Condition Assessment for Ålesund City, Norway', Water, 14: 3993.
- Sousa, Vitor, José P Matos, and Natércia Matias. 2014. 'Evaluation of artificial intelligence tool performance and uncertainty for predicting sewer structural condition', Automation in Construction, 44: 84-91.
- Weeraddana, Dilusha, Bin Liang, Zhidong Li, Yang Wang, Fang Chen, Livia Bonazzi, Dean Phillips, and Nitin Saxena. 2020. 'Utilizing machine learning to prevent water main breaks by understanding pipeline failure drivers', arXiv preprint arXiv:2006.03385.
- Privacy-preserving machine learning: proven algorithms
Supervisors: Peter Lammich, Milan Lopuhaä-Zwakenberg
Differential Privacy (DP) is a metric to measure the privacy leakage of an algorithm handling sensitive data. Because of its strong privacy guarantees it is widely used in both industry and academia. An important application is privacy-preserving classifier training, where a data classifier must be created based on training data, without leaking too much information about the training data itself. Multiple privacy-preserving classifiers have been proposed; however, because differential privacy is hard to verify experimentally, many proposed classifiers have more privacy leakage than is desirable.
The aim of this project is to create algorithms whose privacy-preserving properties have been confirmed by the interactive theorem prover Isabelle. The first step is to formalize DP into Isabelle; then, the student can verify whether existing algorithms indeed satisfy DP, and study how they must be adapted otherwise.
- Privacy-preserving machine learning: verified algorithms
Supervisors: Petra van den Bos, Milan Lopuhaä-Zwakenberg
Differential Privacy (DP) is a metric to measure the privacy leakage of an algorithm handling sensitive data. Because of its strong privacy guarantees it is widely used in both industry and academia. An important application is private classifier training, where a data classifier must be created based on training data, without leaking too much information about the training data itself. In this project, the student investigates the formal verification of DP algorithms. More concretely:
- Implement DP classifiers into formal verification tools to verify their privacy;
- Study the difference, in implementation and verification guarantees, of static vs dynamic verification;
- See whether these tools have the functionality to verify a range of DP classifiers.
Contact
