9th Munich Earth Skience School
Berghotel Sudelfeld
www.berghotel-sudelfeld.de
17-22 February, 2019
|
This winter school will focus on the use of machine learning in Earth Sciences with specific applications to seismology. As usual we will combine theory with practicals using the Python language and Jupyter notebooks. The machine learning part will make use of the scikit-learn (scikit-learn.org) tools and/or tensorflow (tensorflow.org). In terms of methodology and applications we will touch on classification, regression, clustering, feature extraction problems.As always we start with a day of Python/ObsPy introduction with a focus on new features. A video on the winter school 2019 can be seen here. All course material can ce accessed here. Lecturers/Tutors: Lion Krischer (ETH Zurich), Men-Andrin Meier (CalTech, Pasadena), Polina Polunina (IPG Paris), Michaela Wenner (ETH Zurich) Organization: Heiner Igel (LMU Munich) All workshop material (presentations, Jupyter notebooks) now available here. Video on the 9th Skience School available here. |
Design study (mid eighties) for the legendary CM-1 parallel computer by Thinking Machines Co. anticipating its use for artificial intelligence. Logo invented by and graphic courtesy of Tamiko Thiel (tamikothiel.com). |
Practical information: Lectures, practicals will be given from 8-11 and after 3pm (open end). Between 11am and 3pm (depending on conditions) winter sports facilities can be used (www.sudelfeld.de). More info on the winter sport programme can be found here.
Philosophy: This workshop combines the thrill of learning, studying, and practicing seismology and new methodologies, with the thrill of wintersports and being out in nature. Beware that the hotel is isolated in the mountains, no shopping, little to no public transport to anywhere. WLAN is available free of charge at the hotel. Note that we provide half pension only, in the ski area (within walking distance) there are restaurants where lunch is available.
Costs: 500 € (incl. 5 days half pension, conference facilities, tuition, software). Arrival Sunday February 17 leaving Fri February 22nd (lunch time). Note that the costs do not cover: lunch, drinks during meals, winter sport facilities. The number of participants is limited to ~40. First come, first serve basis. The area can be reached by train from Munich Airport (terminal station: Bayrischzell). Please note that there are only a few single rooms. You are encouraged to (or might be forced to) sharing a double room.
Registration: Click here to access registration page. Deadline January 20, 2019.
Logistics: Please make your way to Bayrischzell Station on Sunday February 17 afternoon, unless you come with your own vehicle. We will provide a shuttle service from the train station to the hotel. More information shortly before the workshop.
|
Morning 8-11 |
Afternoon 15-18 |
Evening 20-open end | |
|
Sun |
Arrival - Icebreaker at 18:00 | Opening, software installation | |
|
Mon |
Python/ObsPy Introduction |
Python/ObsPy - Intro to Workshop Topic |
|
|
Tue
|
Machine Learning - General Introduction Classical” supervised regression & classification methods (linear/logistic regression, SVM, naive Bayes) |
Shallow networks including activation functions, back-propagation, gradient descent, Unsupervised learning (k-means, t-sne) Unsupervised learning application to large data volumes |
Late Night special |
|
Wed |
Random Forests & Application to Hangslides |
Free Afternoon - Parallel slalom |
Bogner Ski night (movies) |
|
Thu |
Intro to deep learning (why the hype, why does it suddenly work, vanishing gradient, GPUs, network architectures, etc)
|
Production level deep learning Other types of networks (recurrent networks, bayesian/stochastic networks) Generative networks |
Mulled wine |
|
Fri |
Applications (short presentations by participants) |