Due to the covid19 pandemic, the first Doctoral School and the Midterm meeting will be held online during the first week of February 2021. Despite its name, the Midterm meeting with the REA Project Officer is held on February 4, 2021 as an one day event. Both events will be held on Zoom, links are emailed to the participants. Please join the meeting using your full name!
Schedule of the Doctoral School I
The First Doctoral School focuses on machine learning for shapes. There will be introductory short courses and presentations by GRAPES members and other invited high-profile speakers. Confirmed speakers include P. Alliez (Inria & GF), Y. Avrithis (Inria & Athena RC), M. Bronstein (USI and ICL), M. Salamó (UB), and A. Leutgeb from industrial partner RISC-Software. A pdf version of the schedule is also available.
Monday, February 1st, 2021(Chair: I. Emiris) 11:00 Tutorial [Part A]: Pierre Alliez (Inria & GF): Clustering algorithms and introduction to persistent homology
Abstract: This tutorial will first offer an introduction to clustering, both discrete and continuous, and persistent homology: K-means, Lloyd iteration, hierarchical clustering, single linkage algorithm, dendograms, mode seeking clustering, persistent homology. I will also provide a basic introduction to machine learning, neural networks, and related data structures. 12:30 Lunch break 14:00Tutorial [Part B]: Pierre Alliez (Inria & GF): Machine Learning 101 and Neural networks 15:30 Coffee break 17:00 Invited talk:Thierry Chevalier (Airbus): Industrial interest in learning, processing & optimising shapes 18:00 End of the day
Tuesday, February 2nd, 2021(Chair: M. Salamó) 11:00 Tutorial [Part A]: Yannis Avrithis (Inria & Athena RC): Deep learning and computer vision
Abstract: This tutorial will comprise of the following parts: (a) Visual representation: Data-driven representation learning. Neuroscience and receptive fields. Visual descriptors. Hierarchical representations. (b) Machine learning: Linear classification, binary and multi-class. Multiple layers, neural networks. (c) Convolution: Definition, convolutional networks. Pooling. Network architectures. (d) Optimization: Gradient computation. Optimizers, initialization, normalization. Deeper architectures. (e) Retrieval: Spatial pooling. Metric learning and image retrieval. Unsupervised and semi-supervised learning. Graph-based methods. 12:30 Lunch break 14:00 Tutorial [Part B]: Yannis Avrithis (Inria & Athena RC): Deep learning and computer vision 15:30 Coffee break 16:00 Supervisory Board meeting (restricted) 17:00 End of the day
Wednesday, February 3rd, 2021(Chair: A. Mantzaflaris) 11:00 Tutorial [Part A]: Michael Bronstein (USI & Imperial College London) 12:30 Coffee break 13:00 Tutorial [Part B]: Michael Bronstein (USI & Imperial College London) 14:30 Lunch break 16:00 “Social” activities: Virtual tours of the Acropolis: #1, #2, and VR 360ᵒ video tour of the Acropolis Museum.
Friday, February 5th, 2021(Chair: G. Muntingh) 11:00 Tutorial [Part A]: Maria Salamó (UB): Point-cloud analysis and classical Machine Learning techniques
Abstract: The analysis of point cloud has been done in different ways in the field of machine learning. This tutorial is devoted to introducing the main machine learning algorithms that can be used for point-cloud analysis, among other applications. It has been divided in two parts. First part is focused on the theoretical aspects of classical machine learning algorithms, while the second part is devoted to deep learning algorithms. In both parts, the theoretical foundations of every type of algorithm will be detailed. We will see practical examples of 3D shapes analysis on every part. 12:30 Lunch break 14:00 Tutorial [Part B]: Simone Balocco (UB): Deep learning approaches 15:30 Coffee break 16:00 Educational Committee meeting (restricted) 16:30 End of the day
Monday, February 8th, 2021 11:00 Tutorial [Part A]: Alexander Leutgeb (RISC-Software): Dynamic Cutting Force Prediction (During Simulation of Machining Processes) (Chair: A. Fabri)
Abstract: The problem domain is to predict the cutting forces during simulation of metal cutting processes. There already exist several force computation models. These models have in common, that for different material combinations (cutting tool and workpiece) the determination of their coefficients is a time consuming task by performing different metal cutting patterns with the real machine and measuring the resulting cutting forces over the time. The novel approach of our research project should learn this coefficients from regular machining processes. That means we need on the one hand the data from the simulation (machining parameters and geometric representation of cutter workpiece engagement) and on the other hand the measured forces of the sensors from the real machine. Because usually the real machines do not come with sensors providing force measurements we will use an Internet-of-Things sensor from Pro Micron, which measures the forces directly on the tool side. So from a Machine Learning standpoint we have to first correlate the time series data from the simulation with that of the sensor. Later on we perform model fitting/regression (linear & non linear) to learn the coefficients for the force computation model.
After the model is trained, the simulation of the machining process can predict the cutting forces over the time. The application of these forces is relevant for machine vibration analysis, feed rate optimization, increase of energy efficiency, geometric errors in finished workpieces caused by tool cutter deflections, and so on. 12:00 Lunch break 14:00 Tutorial [Part B]: Christoph Hofer (RISC-Software): Dynamic Cutting Force Prediction (During Simulation of Machining Processes) (Chair: A. Fabri) 15:00 Coffee break 17:00 Seminar:Managing and sharing research outputs: all you need to know — SLIDES & VIDEO Presenters: Elli Papadopoulou, Athena RC/ OpenAIRE, Iryna Kuchma, Electronic Information for Libraries – eifl / OpenAIRE
Abstract: This is an introductory session to the Open Science model. It aims to ensure that all ESRs have a comprehensive understanding of the basics of Open Science and are able to apply them in their everyday work using available resources. Presentations will highlight good practices in scientific publishing and data management activities that increase researchers recognition and at the same time achieve compliance with European requirements (e.g. grants received under H2020, HorizonEurope and ERC). A “first-aid toolkit” will be provided so that researchers can practice Open Science skills at their own time and pace. 18:15 End of the Doctoral School
Schedule of the Midterm Meeting (February 4th, 2021)
The purpose of the Midterm meeting between the REA Project Officer (Filippo Galiardi) and the Consortium is as to discuss the implementation of the project, make sure that all ESRs have been recruited and review the rules of the Marie Curie programme to see if there are any issues that need to be addressed before the beginning of the scientific activities. All beneficiaries and ESRs should participate the meeting. Our Partner Organisations are strongly encouraged to participate.
You can find a pdf version of the Midterm meeting program here. All times are Greek local time (EET=UTC+2=GMT+2).
10:00 Introduction by the Project Coordinator and the REA Project Officer 10:05 Tour de table: All scientists-in-charge briefly present their research team and describe their role within the network. Introduction of the Partner Organisations 10:45 REA Project officer presentation 11:05 Coordinator’s report: Presentation of the Network and the Progress report 11:35 Lunch break 13:00 Fellows’ individual presentations 14:00 Coffee break 14:15 Fellows’ individual presentations 15:00 Restricted Meeting between the fellows and the Project Officer 16:00 Coffee break 16:15 Restricted session: Meeting between site-leaders and Project Officer 16:30 Feedback and open discussion 17:00 End of the meeting