Combining Feature and Algorithm Hyperparameter Selection using some Metalearning Methods.

Miguel Cachada Salisu Mamman Abdulrahman Pavel Brazdil

Published in: AutoML@PKDD/ECML (2017)

Keyphrases

computational cost
significant improvement
preprocessing
dynamic programming
synthetic and real images
detection algorithm
learning algorithm
simulated annealing

selection algorithm
test images
machine learning
heuristic methods
convergence rate
matching algorithm
optimization algorithm
k means

data mining problems
feature maps
exhaustive search
probabilistic model
cross validation
clustering method
worst case
expectation maximization

classification method
cost function
model selection
feature subset
parameter settings
computational complexity
optimal solution
neural network