Symbol Kurs

2021-10-04 - 2021-10-08 CAS AML Module 3 - Deep Learning Research

Deep Neural Networks - Advanced Topics (M3 CAS AML)

Reiter

Bern Autumn School on Machine Learning
Module 3 of the CAS Advanced Machine Learning

About 
Due to increased access to data and compute capacity Machine Learning and Artificial Intelligence have become useful in many areas. In research and industry it is applied in various fields. Image recognition, online discrimination, natural language treatment, robotics, omics ... you name it. With this school we target people already familiar with machine learning who would like learn about this year's topics (see below).  

In this school you attend lectures and tutorial sessions over four mornings. This happens in a great hotel on Mallorca, a great resort for hiking and water sport, far away from La Palma and Ballermann. Please fill out the doodle sent out to you to inform us of your in-person participation on Mallorca at the hotel Es Blau des Nord: https://www.esblaudesnord.com/en/. All lectures are held in hybrid mode and you can access on-the-spot or online. 

This year's topics
  • Autoencoders (day 1)
  • TBC (day 2)
  • Interpretable ML (day 3)
  • Data Augmentation Techniques (day 4)
Target group
  • Enrolled people for CAS Advanced Machine Learning
  • Other interested people with the required prerequisites
Prerequisites
  • You must bring your own laptop
  • Mathematics and statistics at the level of an introductionary course on university level 
  • Python knowledge ABSOLUTELY required
  • Data analysis and Machine Learning experiences
  • The training is as language independent as possible, but examples and practical work is in Python
Methods
  • Lectures based on Jupyter notebooks, evening talks, tutorials (with Jupyter notebooks), project work with presentation.
Certificate 
  • A certificate will be delivered to participants who have attended the whole training and presented their project work successfully. The school yields 2 ECTS points. 
Coaches
  • The coaches are local and external experts
Time : 2021-10-04 - 2021-10-08 (afternoons for work, swimming, hiking, wellness or whatever) 
Location Mallorca: https://www.esblaudesnord.com/en/
Location University of Bern: Sidlerstrasse 5, room 228, if closed, call housekeeper on 031 631 4040.
URL: Login to Ilias to see the link. 

Course Fee : For CAS AML participants there is no fee and full pension hotel (three meals) is included in the CAS fee, For others the fee is 600 CHF or 900 CHF with project and 2 ECTS points. Others also have to pay the hotel. All participants arrange and pay their travels themselves.     

Language: English
Participants : Max 24
Registration : Mandatory
Responsible : PD Dr. Sigve Haug
Monday 2021-10-04 (Arrival)
17:00 - 19:00 Apero at the bar or on the terrace
19:00 - 21:00 Dinner at the hotel

Tuesday 2021-10-05
08:00 - 08:15 Welcome and Module Introduction (Sigve)
08:15 - 09:00 Lecture 1 on Interpretable Machine Learning (Sigve)
09:15 - 09:45 Tutorial
10:00 - 10:30 Coffee break
10:15 - 11:30 Lecture 2 on Interpretable Machine Learning (Sigve)
11:45 - 12:30 Tutorial
12:15 - 17:00 Beach, sport, hiking, work or whatever
17:00 - 19:00 ML Exercise (Sigve)
19:00 - 20:30 Dinner at hotel
20:30 - 21:30 Evening Industry Lecture (Enes Deumic): Predicting Disengagement from the Events Data. Read this before the lecture.
Abstract: The patterns in user behavior can be a very useful early indicator of user's disengagement from the product and finally, churn. Predicting disengagement on time can allow us to act early by motivating users to stay with us, for example, by offering a discount. We will show how to use transformer models on heterogeneous event data to get predictions of user disengagement, but also embeddings for categorical data that can then be used on the other downstream tasks.
Wednesday 2021-10-06
08:00 - 08:45 Lecture 1 on Autoencoders (Mykhailo)
09:00 - 09:45 Tutorial
09:45 - 10:15 Coffee break
10:15 - 11:30 Lecture 2 on Autoencoders (Mykhailo)
11:45 - 12:30 Tutorial
17:00 - 19:00 ML Exercise (Mykhailo)
20:00 - 20:00 Dinner at the hotel
20:00 - 20:30 TBC

Thursday 2021-10-07
08:00 - 08:45 Lecture 3 on Interpretable Machine Learning (Aris/Mykhailo)
09:00 - 09:45 Tutorial
10:00 - 10:30 Coffee break
10:15 - 11:30 Lecture 4 on Interpretable Machine Learning (Aris/Mykhailo)
11:45 - 12:30 Tutorial
12:15 - 17:00 Beach, sport, hiking, work or whatever
17:00 - 19:00 ML Exercise
20:00 - 20:00 Dinner at hotel

Friday 2021-10-08 
08:00 - 09:00 Tutorial on TensorBoard
09:15 - 09:30 Projects (M3 and CAS)
10:00 - 10:30 Coffee break
10:30 - 12:00 Projects
12:00 - 12:30 Wrap up (all)
12:30         End of school

Friday 2021-11-xx Presentation Day 
09:00 - 12:30 Presentations
To pass Module 3 you need to peform a small (30h) project and present it on December 4th, from 09:00 to 12:30. The URL is the same as for the course (log in). Onsite room will be announced here. 

Expected effort: 30 hours

Result: 10 minutes presentation of your project notebook. Pleaes upload your notebook to Ilias before presenting (naming convention: surname-firstname-projectname.ipynb)

Teamwork:You are encouraged to work and present in teams of two (or three). You may also work alone.

Expectations: Training and study of an autoencoder on a dataset not used in class, an application of one method from day one and one method from day three, i.e. some interpretation, either of your encoder model or some other model.

Assessment: You will get feedback right after your presentation. If you have given it a good try (~30h) your project will pass. There is no further grading. The project together with school attendance yield 2 ECTS credit points.

Schedule (25 min per team):
Registration: Here on Ilias

More information about the hotel in Mallorca: https://www.esblaudesnord.com/en/ (included in the CAS fee and part of the Bern Autumn School CAS AML). We have pre-booked the single rooms already. You are free to bring familiy and friends of course (not participating in the school) and book extended days on own cost, if there are bigger rooms free.
PD Dr. Sigve Haug (overview, school responsible)

Sigve studied physics in Germany, Spain and Norway. He has been involved in neutrino physics experiments and high energy frontier experiments, often with main focus on the computing challenges related to the large and distributed data from these experiments. Today he is working for the Albert Einstein Center for fundamental Physics and the Mathematical Institute of the University of Bern where is leading the Science IT Support unit. Beyond science he likes philosophical conversations in the evening. 

Dr. Aris Marcolongo 

Aris is an ML expert by training, PhD in computer science, and experience in various research enterprises. Currently he is pursuing a research project investigating ML methods for understatnding extreme climate events with compound drivers. He carries the rare combination of friendliness and deep technical knowledge. 

Dr. Mykhailo Vladymyrov

Mykhailo is the best tutor we ever had. He is a trained physicist who learned to know at the Albert Einstein Institute of Fundamental physics (and beyond) with many years of experience with big data, machine learning and GPU computing. This year he is working for the Theodor Kocher Institute at the University of Bern. Mykhailo has a high level humor and view upon the human strive.