Natural policy gradient methods and the covariance matrix adaptation evolution strategy, two variable metric methods proposed for solving reinforcement learning tasks, are contrasted to point out their conceptual similarities and differences. Experiments on the cart pole benchmark are conducted as a first attempt to compare their performance.

An approach based on a (μ+1)-ES and three simple tournament rules is proposed to solve global optimization problems. The proposed approach does not use a penalty function and does not require any extra parameters other than the original parameters of an evolution strategy. This approach is validated with respect to the state-of-the-art techniques in evolutionary constrained optimization using a...

We employ local meta-models to enhance the efficiency of evolution strategies in the optimization of computationally expensive problems. The method involves the combination of second order local regression meta-models with the Covariance Matrix Adaptation Evolution Strategy. Experiments on benchmark problems demonstrate that the proposed meta-models have the potential to reliably account for th...

AbstractIn this paper, we propose the use of a Simple Evolution Strategy (SES) (i.e., a -ES with self-adaptation that uses three tournament rules based on feasibility) coupled with a diversity mechanism to solve constrained optimization problems. The proposed mechanism is based on multiobjective optimization concepts taken from an approach called the Niched-Pareto Genetic Algorithm (NPGA). The ...

This paper describes a method for rendering search coordinate system independent, Adaptive Encoding. Adaptive Encoding is applicable to any iterative search algorithm and employs incremental changes of the representation of solutions. One attractive way to change the representation in the continuous domain is derived from Covariance Matrix Adaptation (CMA). In this case, adaptive encoding recov...

Learning emotion assessment is a non-negligible step in analyzing learners’ cognitive processing. Data are the basis of learning assessment. However, existing models cannot balance model accuracy and interpretability well due to influence uncertainty process data collection parameter errors. Given above problems, new based on evidence reasoning belief rule base (E-BRB) proposed this paper. Firs...

According to a theorem by Astete-Morales, Cauwet, and Teytaud, “simple Evolution Strategies (ES)” that optimize quadratic functions disturbed by additive Gaussian noise of constant variance can only reach a simple regret log-log convergence slope ≥ −1/2 (lower bound). In this paper a population size controlled ES is presented that is able to perform better than the −1/2 limit. It is shown exper...

In this paper we show how to modify a large class of evolution strategies (ES) to rigorously achieve a form of global convergence, meaning convergence to stationary points independently of the starting point. The type of ES under consideration recombine the parents by means of a weighted sum, around which the offsprings are computed by random generation. One relevant instance of such ES is CMA-...

The diversity of floral forms in nature can be explained largely as adaptations to the diversity of biotic and abiotic selective agents with which different plant species interact. Ecological genetics is the study of the process of adaptation, and therefore is an ideal approach to understanding floral adaptations. Here I review work on selection and genetic variance and covariance of floral tra...