dplbnDE: Discriminative Parameter Learning of Bayesian Networks by...

In dplbnDE: Discriminative Parameter Learning of Bayesian Networks by Differential Evolution

Discriminative Parameter Learning of Bayesian Networks by Differential Evolution

Description

Implements Differential Evolution (DE) to train parameters of Bayesian Networks (BN) for optimizing the Conditional Log-Likelihood (Discriminative Learning) instead of the log-likelihood (Generative Learning). Any given BN structure encodes assumptions about conditional independencies among the attributes and will result in error if they do not hold in the data. Such an error includes the classification dimension. The main goal of Discriminative learning is minimize this type of error.

Details

DE variants: Based on different strategies followed by the operators of DE, there are different variants, which define the way in which the mutant and trial vectors are generated. The most popular variant is called DE/rand/1/bin, where “DE” means Differential Evolution, the word “rand” indicates that the so-called base vector is randomly chosen, “1” is the number of vector pairs (i.e., vector differences to be calculated) chosen, and finally “bin” means that a binomial recombination is chosen. The following is a list with the available variants:

• DErand: Implements DE/rand/ variant with 1 or 2 pairs of vector differences, and binomial or exponential recombination. (Price and Storn, 1996)

• DEbest: Implements DE/best/ variant with 1 or 2 pairs of vector differences and binomial or exponential recombination. (Price and Storn, 1996)

• jade: A variant that includes some mechanisms to decrease the dependence to its parameter values such as F and CR. JADE uses a mutation strategy called DE/current-to-pbest, where p \in (0 , 1]. Base vectors are selected from the best 100p maintaining diversity, uses an optional external archive. (Zhang and Sanderson, 2009)

• lshade: An improved version of JADE, LSHADE incorporates a Linear Population Size Reduction (LPSR) in order to improve the performance. (Tanabe and Fukunaga, 2014)

References

Price K and Storn R (1996), Minimizing the real functions of the icec’96 contest by differential evolution. In Proc. of IEEE C. Evol. Computat., pp. 842–844.

Zhang J and Sanderson A (2009). Jade: adaptive differential evolution with optional external archive. IEEE Trans. Evol. Comput., pp. 945–958.

Tanabe R and Fukunaga A (2014). Improving the search performance of shade using linear population size reduction. In Proc. of IEEE C. Evol. Computat., pp. 1658–1665.

dplbnDE documentation built on Aug. 19, 2023, 1:07 a.m.