Open semester project and master thesis opportunities for students:
Symbolic regression is a set of methods designed to discover symbolic, human-readable formulas that best fit a given dataset. It is a powerful tool for model discovery, enabling identification of underlying patterns or dynamics in an unknown system. The modern symbolic regression methods are highly applicable in fundamental science. The outcomes of symbolic regression methods are the models can take the form of lagebraic expressions as well as differential equations. However, non-linear symbolic regression methods often face significant challenges, including convergence to local minima and enforcing the sparcity in the resulting expression coefficients. Addressing these issues is crucial to improving the interpretability and robustness of the derived models. In this project, the student will have the opportunity to work on enhancing optimization strategies for previously proposed non-linear symbolic regression methods. This project could serve as an excellent entry point into the fields of symbolic regression and model discovery, with broad applications in advancing fundamental scientific understanding.
[1] Makke, N., & Chawla, S. (2024). Interpretable scientific discovery with symbolic regression: a review. Artificial Intelligence Review, 57(1), 2.
[2] Sahoo, S., Lampert, C., & Martius, G. (2018, July). Learning equations for extrapolation and control. In International Conference on Machine Learning (pp. 4442-4450). Pmlr.
You can apply by directly contacting Sergei Garmaev.
This project is part of a collaboration between the IMOS lab and Matterhorn Gotthard Bahn, a railway company operating in the Swiss Alps. The student will work on developing computer vision algorithms for automated visual inspection of retaining walls around railway tracks. Retaining (or supporting) walls are crucial infrastructure elements responsible for maintaining the structural integrity of terrains around railway tracks and ensure safe operation. They are subject to wear and damages including cracks, concrete cancer (i.e., alkali–silica reaction), displacements, erosion and water infiltration. Images of retaining walls have already been collected and labels are available. The goal will be to design algorithms to estimate the condition of a wall, with a focus on robustness, transfer learning, and explainability (XAI).
Keywords: Railway, Structural Health Monitoring, Computer Vision, Machine Learning, Transfer Learning, Domain Adaptation, XAI
Detailed description on SiROP
You can apply via SiROP or directly contact Florent Forest.
In this project the student will attempt to embed relevant physics priors in graph neural networks (GNNs) with the goal of developing an inverse modeling framework for the prediction of the underlying material model of a system, given the labeled input (load/excitation) and the output (mechanical response) data.
Keywords: Graph neural networks, Physics prior, finite element method
You can apply via SiROP or directly contact Vinay Sharma.
Time-series data is increasingly prevalent across various domains, including finance, healthcare, and environmental monitoring. The ability to extract meaningful information from time-series data is crucial for prediction, classification, and anomaly detection. This project focuses on exploring different time-series representations and their impact on machine learning tasks.
Objectives
-
Benchmarking Time-Series Representations: Investigate the performance of various time-series representations including Fast Fourier Transform (FFT), Recurrence Plot (RP), Gramian Angular Field (GAF), and Markov Transition Field (MTF) in the context of machine learning tasks.
-
Design of Fusion Method: Develop a novel method to fuse information from these representations, aiming to leverage their combined strengths.
-
Evaluation on Diverse Datasets: Assess the effectiveness of the proposed method on a variety of datasets pertaining to Time-Series Classification, Forecasting, and Anomaly Detection.
Methodology
-
Literature Review: Conduct a comprehensive review of existing literature on time-series representation methods and their applications.
-
Implementation of Representations: Implement and fine-tune FFT, RP, GAF, and MTF, ensuring a fair and consistent basis for comparison.
-
Benchmarking: Evaluate each representation method across multiple datasets, using standard metrics relevant to classification, forecasting, and anomaly detection.
-
Fusion Method Development: Design a fusion algorithm to integrate the different representations, potentially using techniques like feature concatenation, ensemble methods, or deep learning architectures.
-
Experimental Evaluation: Test the fusion method across the same datasets used for benchmarking, comparing its performance against the individual representation methods.
You can apply via SiROP or directly contact Hao Dong.
In this project, you will focus on the task of learning/simulating rigid objects dynamics from videos, with the end-goal of predicting future or alternative trajectories for the objects in the scene. This task includes the decomposition of a visual scene into multiple blocks (identifying individual objects), the modeling of their evolution and interactions in the scene (positions, velocities, collisions, …), and the prediction of future (or alternative) trajectories (also called rollout).
Keywords: Computer Vision, Mechanics, Scene Understanding, Object-centric Learning
You can apply via IS-Academia – project #14080 or directly contact Amaury Wei.
The Industrial Internet of Things (IIoT) generates vast quantities of data from interconnected sensors and devices. To utilize the full potential of this data, it’s essential to go beyond individual sensor readings and understand the complex relationships between them. These relationships can be diverse: sensors may monitor specific equipment, while actuators directly control or operate parts of a process. Graph Neural Networks (GNNs) are powerful tools to model complex sensor networks, leveraging relationship information typically lost in traditional machine learning approaches. Analyzing IIoT data without considering sensor relationships can lead to missed insights and suboptimal decision-making. By inferring a graph structure from an IIoT sensor network, we create a structure capable of representing these relationships. GNNs, by learning directly from this graph, can model and distinguish between different relationship types (“monitors,” “operates,” etc.), leading to more accurate and interpretable predictions, fault/anomaly detections, and diagnostics. This thesis aims to bridge this gap, developing methods to capture sensor relationships and leverage GNNs for their analysis.
You can apply via SiROP or directly contact Mengjie Zhao.
Modern buildings are equipped with a large number of HVAC (Heating, Ventilation, and Air Conditioning) devices, comprising sensors, actuators, control systems mounted in very different locations. The integration of these components into the overarching building control system necessitates a meticulous configuration process known as commissioning. This undertaking represents the predominant portion of the interaction time between the device and the human operator, excluding routine operational tasks such as temperature readings from room sensors. Consequently, commissioning emerges as the most time-consuming, expensive, and error-prone phase in the establishment of a seamlessly operating building, as any errors during commissioning can significantly impact the overall efficiency and functionality of the building.
In the context of this Master Thesis, the objective is to investigate the feasibility of developing an algorithm capable of predicting and verifying parameter values during the commissioning process, aiming to mitigate the occurrence of human errors.
You can apply via SiROP or directly contact Leandro von Krannichfeldt.
Thermal cameras have grown increasingly popular due to their competitive pricing, robustness in adverse weather, and the unique modality of capturing surface thermal distributions. These techniques have been widely applied in robust object detection and tracking, infrastructure diagnostics, and large-scale urban analysis. This project will explore the integration of thermal imaging with cutting-edge 3D reconstruction methods. The student will develop techniques to incorporate thermal images into 3D models, enabling the visualization of high-fidelity thermal distributions across diverse objects and scenes.
You can apply via SiROP or directly contact Chenghao Xu.
Most deep models are trained under a close-world assumption and hence do not possess knowledge of what they do not know, leading to over-confident and inaccurate predictions for unknown objects. Out-of-distribution (OOD) detection, which aims to identify the unknown objects, has attracted much attention as it plays a vital role in a variety of applications such as industrial inspection, medical diagnosis, and autonomous driving. One of the primary challenges of effective OOD lies in the wide-existing domain shift in the real world. The testing dataset might have a significant distribution from the training data of deep learning models. Thus, it is even more difficult for the model to distinguish between the domain difference and the OOD objects. In this task, the student is expected to work on the OOD detection task with domain shift. Specifically, we focus on test-time domain adaptation, which focuses on OOD detection during inference time online. The potential application case can be found here: https://www.epfl.ch/labs/cvlab/data/road-anomaly/. Overall, the goal of this project is to improve the OOD performance on new domains through the use of test-time domain adaptation techniques. The student is expected to undertake the following tasks in this project: – explore current test-time domain adaptation methods – improve those methods and modify them for OOD detection.
You can apply via SiROP or directly contact Han Sun.
Denoising Diffusion Probabilistic Models (DDPMs) are a highly popular class of deep generative models that have been successfully applied to various problems, including image and video generation. Recently, there have been attempts to apply these models to time-series data. However, they are not enough for modeling spatial-temporal interactions. They primarily focus on temporal dependencies within individual sensors, neglecting the spatial correlations between different nodes. In reality, objects are spatially correlated with each other. This project aims to explore a graph-based denoising diffusion probabilistic model to jointly capture both temporal dependencies and spatial interactions.
- This project seeks to merge the capabilities of DDPM with Spatial-Temporal Graph Neural Networks, potentially unlocking new possibilities for academic research and practical applications in the Industrial Internet of Things (IIoT). The objectives of this project include:
- Develop a graph-based denoising diffusion probabilistic model that effectively captures both temporal dependencies and spatial interactions in IIoT data. • Conduct comprehensive experiments to compare the performance (e.g., forecasting, imputation, etc.) of the proposed model against existing state-of-the-art methods.
You can apply via SiROP or directly contact Keivan Faghih.
Detecting out-of-distribution (OOD) samples is important for deploying machine learning models in safety-critical applications such as autonomous driving and robot-assisted surgery. Existing research has mainly focused on unimodal scenarios on image data. However, real-world applications are inherently multimodal, which makes it essential to leverage information from multiple modalities to enhance the efficacy of OOD detection. A recent work [1] has introduced the first-of-its-kind benchmark for Multimodal Out-of-Distribution Detection characterized by diverse dataset sizes and varying modality combinations. Some work [2] also explores vision-language models for unimodal OOD detection on images. In this project, we will explore the rich information in Foundation Models to improve the Multimodal OOD Detection performances.
References
[1] MultiOOD: Scaling Out-of-Distribution Detection for Multiple Modalities, https://arxiv.org/pdf/2405.17419
[2] Delving into Out-of-Distribution Detection with Vision-Language Representations, https://arxiv.org/pdf/2211.13445
You can apply via SiROP or directly contact Hao Dong [email protected]
In many real-world complex systems, the output is influenced by both the input, such as operating conditions, and the system’s health status. To accurately predict the dynamics of such systems, it’s essential to not only understand the relationship between input and output based on various health conditions, but also to model the system’s degradation over time. In this project, our objective is to combine physical knowledge to develop a neural ordinary differential equation (ODE) to model this degradation, thereby enabling precise predictions of the system’s dynamics over time. Here, we will mainly focus on two systems, the Lithium-ion battery and CMAPSS jet engine.
Key words: Neural ODE, Degradation modeling, Physics-informed machine learning.
Goal:
- Conduct a comprehensive literature review to gain a deep understanding of the fundamental physics underlying the system’s behavior.
- Based on the learned physical knowledge, design an optimal neural ODE architecture that effectively captures the dynamics of degradation and system outputs.
- Investigate various approaches for inferring the initial conditions of the neural ODE model, i.e. the initial guess of degradation based on initial context, and perform an extensive and systematic comparison of these methods.
- Quantitatively evaluate the proposed neural ODE model’s accuracy in predicting the system dynamics, and compare its performance with other existing algorithms.
You can apply via SiROP or directly contact Zhan Ma [email protected]
- Conduct a comprehensive literature review about Building Thermal Modeling techniques and Physics-Inspired GNN for time series.
- Design a novel Physics-Inspired GNN based on an existing GNN for time series and Reduced Order Model
- Enforce principal laws of thermodynamics in the message passing algorithm of the Physics-Inspired GNN
Prospective candidates are expected to possess a foundational understanding of statistics, optimization, machine learning, and practical experience in programming such as Python. Moreover, fundamental knowledge in thermodynamics is a prequisite. Pior experience with Graph Neural Networks or Graph Signal Processing is highly recommended. Interested students are asked to submit an updated CV along with transcripts of academic records.
You can apply via SiROP or directly contact Leandro Von Krannichfeldt [email protected]
