COSEAL Workshop 2019

Photo credit: HPI/K. Herschelmann

Overview

The seventh COSEAL Workshop is a forum for discussing the most recent advances in the automated configuration and selection of algorithms. It will take place on August 26 and 27, 2019 in Potsdam, Germany and is organized by Holger Hoos and Marius Lindauer as co-chairs and Tobias Friedrich and Torsten Schaub as local chairs.

The workshop will consist of posters and talks about late-breaking research and useful tools, discussions regarding intra- and international cooperation, and many opportunities to interact with other attendees. Note also that this COSEAL workshop will be co-located with the FOGA conference,  which runs from 27-29th of August.

Contact

All inquiries should be done by email to lindauer@cs.uni-freiburg.de

Scope of the Workshop

The scope of the workshop includes, but is not limited to:

  • Algorithm selection
  • Algorithm configuration 
  • Algorithm portfolios
  • Performance predictions and empirical performance models
  • Bayesian optimization
  • Hyperparameter optimization
  • Automated machine learning (AutoML)
  • Neural architecture search
  • Meta-learning
  • Algorithm and parameter control
  • Explorative landscape analysis
  • Programming by optimization
  • Hyper-heuristics

Important Dates

  • Registration opens: April 4th
  • Registration deadline: July 31st
  • Talk/Poster abstract deadline: July 31st
  • Workshop: August 26th and 27th

Registration 

Registration for presenting your work and attending all talks is free. Registration for the joint FOGA/COSEAL social dinner on Tuesday evening is optional and costs 100€.  We hope to cover the 100€ from external funding (see below  “Individual COST support”).

If you would like to present some work at the conference, either as a talk or as a poster (format A1), we would appreciate it if you submit a proposal for your presentation as soon as possible but not later than by July 31st. If you only want to attend and participate in discussions, please also register early, as the number of seats is limited.

To register we have a 3 step process:

  1. Registration for the workshop (mandatory)
  2. Registration for the social dinner on Tuesday evening (optional)
  3. Registration to present a poster or to give a talk (optional)

Please note that the deadline for all three registrations is July 31st 2019.
For the first and third registration, there is no immediate confirmation email.

Deadline extension for social dinner: 27th September

Catering Cost Support

Every event gets better with more coffee and cookies. For exactly this, we get so-called ‘local organizer support’ from COST, depending on the number of COST-registered participants. Unfortunately, COST only counts participants who (a) have created a COST profile, (b) have told us until August 20 which email address they used at e-COST, and (c) will sign the list of participants with a blue pen at the event. This is not required, but highly appreciated as it helps the event. Remember: this improves your own coffee!

Individual COST support

 If you have difficulties to get your trip reimbursed from your home institution, you can get a flat rate fee of 100€ (for one hotel night) from COST after the event. This is completely optional and independent of the aforementioned local organizer support. If this is relevant for you, tick the appropriate box here.

Location

Hasso-Plattner-Institut für Digital Engineering gGmbH
Prof.-Dr.-Helmert-Str. 2-3
D-14482 Potsdam

  • Talks: HS2 (in HS Building)
  • Poster session: D-Space (H-Building)
  • Breakout session: TBA
Map for finding the workshop location (–> HS)

How to reach the venue

Check details here

Accommodation recommendations

Social Program

Monday evening: informal get-together: Beer Garden of the “201” , Rudolf-Breitscheid-Straße 201, 14482 Potsdam; everyone has to pay for herself/himself

Tuesday afternoon: Boat trip + Dinner (registration is mandatory; all drinks included)

Schedule

  • Monday (August 26th)
    • 09:15 – 09:30 Opening (slides)
    • 09:30 – 10:00 Talk I: “Automated Instance Generation via a Constraint Modelling Pipeline” by N. Dang et al.
    • 10:00 – 10:30 Coffee break
    • 10:30 – 11:30 Poster session I
    • 11:30 – 12:00 Talk II
      • “Automating AutoML: A Simple Meta-Learning Approach for Choosing the Model Selection Strategy” by M. Feurer et al.
      • “TPOT-SH: a Faster Optimization Algorithm to Solve the AutoML Problem on Large Datasets” by P. Laurent et al.
    • 12:00 – 13:30 Lunch break
    • 13:30 – 14:00 Talk III
      • “Sparkle: Towards Automated Algorithm Configuration for Everyone” by K. van der Blom et al.
      • “EPP: interpretable score of model predictive power” by A. Gosiewska et al.
    • 14:00 – 15:00 Poster session II
    • 15:00 – 15:30 Coffee break
    • 15:30 – 16:00 Talk IV: “A rigorous analysis of self-adaptation in discrete evolutionary algorithms” by B. Case and P. Lehre
    • 16:00 – 18:00 Breakout session
      • Please propose some discussion topics here
    • 18:00 Informal Get-Together: Beer Garden
  • Tuesday (August 27th)
    • 09:15 – 09:45 Spotlight talks
      • “Darwin & Goliath: Recommendations-As-a-Service (RaaS) with Automated Algorithm-Selection and White-Labels” by Joeran Beel et al.
      • “AI for Materials Science: Tuning Laser-Induced Graphene Production” by L Kotthoff et al.
      • “What algorithms to select in sports timetabling?” by D. Van Bulck and Dries Goossens
      • “Analysis of the Impact of Performance Metrics on the Solution Quality of Algorithm Configurators” by George T. Hall et al.
    • 09:45 – 10:45 Poster session III
    • 10:45 – 11:15 Coffee break
    • 11:15 – 11:45 Invited talk by Johannes Lengler: “What is FOGA?
    • 11:45 – 12:00 “Closing and Invitation to COSEAL’20”
    • 12:00 – 12:15 Group picture
    • 12:15 – 13:30 Lunch break COSEAL + FOGA
    • 13:30 – 14:30 FOGA opening and invited talk by Kevin Leyton-Brown
    • 14:30 – 15:00 Coffee break
    • 15:00 – 17:15 FOGA talks
    • 18:00 – 22:00 Social event: Boat trip + dinner

COSEAL attendees can also attend the Tuesday afternoon session of FOGA (free of charge).

Posters

All talks will also eventually be presented in the poster session immediately following their oral presentation, for a deeper interaction with the audience.

Poster Format: Portrait DIN A1

Poster Session I

  • Title: Multi-Criteria AutoML by Florian Pfisterer, Stefan Coors, Janek Thomas, Bernd Bischl
  • Capping strategies based on performance envelope for the automatic design of meta-heuristics by Marcelo de Souza, Marcus Ritt
  • Towards Assessing the Impact of Bayesian Optimization’s Own Hyperparameters by Marius Lindauer, Matthias Feurer, Katharina Eggensperger, André Biedenkapp and Frank Hutter
  • Hyperparameter Optimisation in Reinforcement Learning by Baohe Zhang, Raghu Rajan, Kurtland Chua, Roberto Calandra, Frank Hutter
  • Algorithm Selection for Image Quality Assessment by Markus Wagner, Hanhe Lin, Shujun Li, Dietmar Saupe
  • Bayesian Optimization over Sets by Jungtaek Kim, Michael McCourt, Tackgeun You, Saehoon Kim, Seungjin Choi
  • Automating Multi-Label Classification Extending ML-Plan by Marcel Wever, Felix Mohr, Alexander Tornede, Eyke Hüllermeier
  • A Method for Realtime gray box algorithm configuration by Dimitri Weiß, Kevin Tierney
  • Utilizing Software Features for Algorithm Selection by Damir Pulatov, Lars Kotthoff
  • Neural Networks for Predicting Algorithm Runtime Distributions by Katharina Eggensperger, Marius Lindauer, Frank Hutter

Poster Session II

  • Towards White-box Benchmarks for Algorithm Control by André Biedenkapp, Furkan Bozkurt, Marius Lindauer, Frank Hutter
  • Algorithm configuration for smart coupling in multiphysics simulations by Hamid Arjmandi, Deepan Chakravarthi Padmanabhan
  • Expressiveness and Robustness of Landscape Features by Quentin Renau, Johann Dreo, Carola Doerr and Benjamin Doerr
  • Understanding and Robustifying Differentiable Architecture Search by Arber Zela, Thomas Elsken, Tonmoy Saikia, Yassine Marrakchi, Thomas Brox, Frank Hutter
  • Neural Large Neighborhood Search for the Vehicle Routing Problem by André Hottung, Kevin Tierney
  • Weighted sampling for combined model selection and hyperparameter tuning by Dimitrios Sarigiannis, Thomas Parnell, Haris Pozidis
  • ADENINE: Automated Deep learning approaches to predict EsseNtIal geNEs by Karthik Azhagesan, Anand Rajasekar, Karthik Raman, Balaraman Ravindran
  • Algorithm Selection as Recommendation: From Collaborative Filtering to Dyad Ranking by Alexander Tornede, Marcel Wever, Felix Mohr, Eyke Hüllermeier
  • Does Algorithm Selection Matter? Exploiting Extreme Parallelism in Portfolio Solving by Jesse Evans, Lars Kotthoff

Poster Session III

  • Darwin & Goliath: Recommendations-As-a-Service (RaaS) with Automated Algorithm-Selection and White-Labels by Joeran Beel, Alan Griffin, Conor O`Shea
  • AI for Materials Science: Tuning Laser-Induced Graphene Production by Lars Kotthoff, Vivek Jain, Alexander Tyrrell, Hud Wahab, Patrick Johnson
  • What algorithms to select in sports timetabling? by David Van Bulck, Dries Goossens
  • Analysis of the Impact of Performance Metrics on the Solution Quality of Algorithm Configurators by George T. Hall, Pietro S. Oliveto, Dirk Sudholt
  • SAPIAS concept: towards an independent self-adaptive per-instance algorithm selection for metaheuristics by Mohamed Amine EL MAJDOULI
  • An Open Source AutoML Benchmark by Pieter Gijsbers, Erin LeDell, Janek Thomas, Sebastien Poirier, Bernd Bischl, Joaquin Vanschoren
  • Adaptive Landscape Analysis by Anja Jankovic, Carola Doerr
  • Interactive Visualizations for ASlib.net by Katherine Chawla, Lars Kotthoff
  • Automated Machine Learning  for Dynamic Energy Management by Can Wang, Holger Hoos, Thomas Bäck
  • Online Selection of CMA-ES Variants by Diederick Vermetten, Sander van Rijn, Thomas Bäck, Carola Doerr