Talks – Machine Learning Lab

Dynamic Algorithm Configuration: Foundation of a New Meta-Algorithmic Framework

Video presentation by André Biedenkapp for our paper "Dynamic Algorithm Configuration: Foundation of a New Meta-Algorithmic Framework" which was presented at ECAI2020.

Abstract:
The performance of many algorithms ... in the fields of hard combinatorial problem solving, machine learning or AI in general depends on parameter tuning. Automated methods have been proposed to alleviate users from the tedious and error-prone task of manually searching for performance-optimized configurations across a set of problem instances. However, there is still a lot of untapped potential through adjusting an algorithm’s parameters online since different parameter values can be optimal at different stages of the algorithm. Prior work showed that reinforcement learning is an effective approach to learn policies for online adjustments of algorithm parameters in a data-driven way. We extend that approach by formulating the resulting dynamic algorithm configuration as a contextual MDP, such that RL not only learns a policy for a single instance, but across a set of instances. To lay the foundation for studying dynamic algorithm configuration with RL in a controlled setting, we propose white-box benchmarks covering major aspects that make dynamic algorithm configuration a hard problem in practice and study the performance of various types of configuration strategies for them. On these white-box benchmarks, we show that (i) RL is a robust candidate for learning configuration policies, outperforming standard parameter optimization approaches, such as classical algorithm configuration; (ii) based on function approximation, RL agents can learn to generalize to new types of instances; and (iii) self-paced learning can substantially improve the performance by selecting a useful sequence of training instances automatically.[+] Show More

Dynamic Algorithm Configuration: Foundation of a New Meta-Algorithmic Framework

Video presentation by André Biedenkapp for our paper "Dynamic [...]

Tutorial on Algorithm Configuration (PPSN'20)

Website: https://www.automl.org/events/tutorial_ac_2020/ Slides: [...]

Website: https://www.automl.org/events/tutorial_ac_2020/
Slides: https://www.automl.org/wp-content/uploads/2020/08/AC_Tutorial_2020.pdf
Colab: https://colab.research.google.com/drive/1v0ZH5S9Sfift30GxHAp96e0yZZUFS0Ah?usp=sharing

In particular for algorithms in AI (incl. machine learning, deep learning, evolutionary algorithms, and hard-combinatorial problem solvers, such as SAT solving), it is crucial to determine a ...good configuration of parameter settings to achieve peak performance. Since an in-depth understanding of these high-dimensional, abstract black-box problems is often not available, manual tuning of parameter configurations is often considered as a tedious, time-consuming and even error-prone task. Therefore, automated algorithm configuration (AC) systems were proposed that free users from this task, and they demonstrated the practical applicability of these methods by achieving substantial speedups in a wide range of applications (e.g., up to 3,000x in SAT, 50x in MIP, 300x in AI planning and 14x in ASP).

To achieve this, traditional AC systems first optimize the configuration on a set of problem instances wrt an arbitrary performance metric, and henceforth the found configurations will be applied to new instances. Although applying any kind of optimization approach (incl. evolutionary algorithms and model-based optimization) to AC might seem to be straightforward, AC systems are faced with many challenges in practice, incl. (i) configuration spaces might contain categorical and continuous parameters with hierarchical dependencies (ii) a single evaluation takes a lot of time, (iii) it is often infeasible to evaluate many configurations on all instances and thus (iv) we might only partially observe censored performance estimates, since we cannot afford to run all evaluations until the end. In the tutorial, we will give an overview on (i) how to address these problems, (ii) how recent advances in the field led to even more efficient systems than ever before, (iii) how to apply these to different applications, including applications to automated machine learning (AutoML) and SAT solving, and (iv) open challenges in the field.[+] Show More

Getting the most out of your planner(s): from static to dynamic algorithm configuration

Invited Talk at ICAPS 2020 by Frank Hutter

Learning Synthetic Environments for RL with Evolution Strategies (AAAI 2021 Meta-Learning Workshop)

This is the recording of our contributed talk at the AAAI 2021 [...]

HPOBench: A Collection of Reproducible Multi-Fidelity Benchmark Problems for HPO

https://arxiv.org/abs/2109.06716 To achieve peak predictive [...]