Search Strategy by Means of Knowledge Extraction

Improvement of the Search Strategy by Means of Knowledge Extraction 

While most combinatorial researches reported up to now involve the use of GA, using the traditional crossover and mutation operators (e.g. WGS 1), it has also been proposed to design new operators for each specific application, to improve search efficiency by means of knowledge extraction [32]. Hence, new methods that combine ES with a knowledge extraction engine have been reported recently within the field of heterogeneous catalysis, such as mining association rules [12, 18, 30, 33] and neural networks [19, 29, 34]. 

In the mid-1990s few papers related the use ANN for catalysts development by predicting performances from their elemental compositions [35-38] but without clear efficiency due to limited data sets. Today, hundreds of catalysts can typically be screened in a week, which opens up the prospect of using ANNs as pre-screening tools for speeding up the optimisation process. This concept was recently reported for modeling quantitatively the performances of ODHE catalysts as a function of elemental composition [39]. 

For the present case study, a first attempt to predict quantitatively performances of WGS catalysts by ANN regression technique led to a rather poor correlation between predicted and experimental CO conversion values (Fig. 10.13). This suggests that, in addition to noisy data, the used descriptors, which were restricted to the single elemental composition of the catalysts, do not contain per se sufficient 

This simplified ANN methodology, which turns out to be efficient even when applied to noisy data sets, may therefore be used as a first pass filter for selecting the best performing systems, without testing the bad ones, during an optimisation process [12, 19]. 

---