Automated Configuration of Mixed Integer Programming Solvers

By Frank Hutter, Holger H. Hoos, Kevin Leyton-Brown. 2010

209 citations recorded by the source corpus.

Subfields: Industrial and Manufacturing Engineering

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Industrial and Manufacturing Engineering

Citations in the corpus

Taking the Human Out of the Loop: A Review of Bayesian Optimization (2016, methodology): “This concept has been applied to e.g., tuning mixed integer solvers [78], [159], and tuning approximate nearest neighbour algorithms [109].”
Exponential Regret Bounds for Gaussian Process Bandits with Deterministic Observations (2012, methodology): “An example of this is the configuration of CPLEX parameters in mixed-integer programming (Hutter et al., 2010).”
Parallel Algorithm Configuration (2012, methodology): “In our research on PARAMILS, we have adopted the policy of performing 10 to 25 parallel independent runs, returning the configuration found by the run with best training performance [1, 4, 2, 3], which can be formalized as follows:”
Algorithm runtime prediction: Methods & evaluation (2012, methodology): “For the commercial MIP solver IBM ILOG CPLEX, we used the same configuration space with 76 parameters as in previous work [52].”
Hydra-MIP : Automated Algorithm Configuration and Selection for Mixed Integer Programming (2011, methodology): “(Please see [13] for the description of each underlying set.”
Surrogate Benchmarks for Hyperparameter Optimization (2014, methodology): “In particular, we plan to support optimization across a set of problem instances, each of which can be described by a fixed-length vector of characteristics, and to assess the resulting surrogates for several problems that algorithm configuration has tackled successfully, such as propositional satis”
UvA-DARE (Digital Academic Repository) Bayesian Optimization in High Dimensions via Random Embeddings (2013, unknown): “…methods for algorithm conﬁgura-tion have recently demonstrated that substantial performance gains of state-of-the-art algorithms can be achieved in a fully automated fashion [ Mo ˇ ckus et al. , 1999; Hutter et al. , 2010; Vallati et al. , 2011; Bergstra et al. , 2011; Wang and de Freitas, 2011].”
Bayesian Optimization in a Billion Dimensions via Random Embeddings (2013, methodology): “…1 / 4 standard deviation conﬁdence intervals of the optimality gap across 50 trials. strated that substantial performance gains of state-of-the-art algorithms can be achieved in a fully automated fashion (Moˇckus et al., 1999; Hutter et al., 2010; Bergstra et al., 2011; Wang & de Freitas, 2011).”
Identifying Key Algorithm Parameters and Instance Features Using Forward Selection (2013, methodology): “We used the same configuration space with 76 parameters as in previous work [1], excluding all CPLEX settings that change the problem formulation (e.”
Programming by optimization (2012, methodology): “9 · 10(47) configurations of the solver), resulting in up to 50-fold speedups over the extensively optimized default configuration.(21) SMBO procedures for algorithm configuration are based on the idea of using the information gained from parameter configurations evaluated so far to build and mainta”
Automated Algorithm Configuration and Parameter Tuning (2012, methodology): “(Details on other, minor differences between TB-SPO and SPO+ can be found in the original article by Hutter et al, 2010b.)”
An Efficient Approach for Assessing Hyperparameter Importance (2014, background): “, 2011), mixed integer programming (Hutter et al., 2010), AI planning (Vallati et al.”
Efficient Parameter Importance Analysis via Ablation with Surrogates (2017, methodology): “Examples include the configuration of state-of-the-art solvers from propositional satisfiability solving (Hutter et al. 2017), AI planning (Fawcett et al. 2011) and mixed integerprogramming (Hutter, Hoos, and Leyton-Brown 2010).”
Learning to Search in Branch and Bound Algorithms (2014, methodology): “We have presented a novel approach to learn an adaptive node searching order for different classes of problems in branch-and-bound algorithms.”