Gaussian Process Optimization with Mutual Information
By Emile Contal, Vianney Perchet, Nicolas Vayatis. 2013
87 citations recorded by the source corpus.
In this paper, we analyze a generic algorithm scheme for sequential global optimization using Gaussian processes. The upper bounds we derive on the cumulative regret for this generic algorithm improve by an exponential factor the previously known bounds for algorithms like GP-UCB. We also introduce the novel Gaussian Process Mutual Information algorithm (GP-MI), which significantly improves further these upper bounds for the cumulative regret. We confirm the efficiency of this algorithm on synthetic and real tasks against the natural competitor, GP-UCB, and also the Expected Improvement heuristic.
Subfields: Artificial Intelligence, Computational Theory and Mathematics, Management Science and Operations Research
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Artificial Intelligence
- Frank Hutter
- Sergey Levine
- Quoc V. Le
- Ameet Talwalkar
- Nando de Freitas
- Ziyun Wang
- Pieter Abbeel
- Aaron Klein
In Computational Theory and Mathematics
- Frank Hutter
- Ziyun Wang
- Nando de Freitas
- Ameet Talwalkar
- Kevin G. Jamieson
- Bobak Shahriari
- Kevin Swersky
- Afshin Rostamizadeh
In Management Science and Operations Research
- Frank Hutter
- Benjamin Van Roy
- Daniel Russo
- Afshin Rostamizadeh
- Ziyun Wang
- Ameet Talwalkar
- Lisha Li
- Giulia DeSalvo
Citations in the corpus
- Hyperband: A Novel Bandit-Based Approach to Hyperparameter Optimization (2016, background): “linear regret bounds (Srinivas et al., 2010; Grunew alder et al., 2010). Wang et al. (2016) improved upon GP-UCB by removing the need to tune a parameter that controls exploration and exploitation. Contal et al. (2014) derived a tighter regret bound than that for GP-UCB by using a mutual information”
- Taking the Human Out of the Loop: A Review of Bayesian Optimization (2016, background): “bandits [89]; Bayes regret bounds for TS [85], [127]; bounds for high-dimensional problems with an underlying low-rank structure [46]; bounds for parallel Bayesian optimization [45]; and improved regret bounds using mutual information [41].”
- Learning to Optimize via Information-Directed Sampling (2014, methodology): “[22] study problems with Gaussian process priors and a method that guides exploration using the mutual information between the objective function and the next observation.”