Research Area

Over the last 4-5 decades, evolutionary computation researchers have proposed several alternative approaches to construct evolutionary algorithms (EAs). Some such alternatives are one-point / two-points / uniform crossover operators, tournament / ranking / stochastic uniform sampling selection methods, clearing / crowding / sharing based niching algorithms, adaptive penalty / epsilon / superiority of feasible constraint handling approaches and so on. Clearly, there are several alternative approaches at every step of an EA and users will have to perform numerous simulations and pick the best approaches. In addition, each approach may require users to fine tune associated parameters. Furthermore, at different stages of evolution, different strategies and different parameter values can be more appropriate. Therefore, the trial and error approach to module selection and associated parameter tuning approach is not efficient. Recently, an ensemble strategy was proposed to benefit from both the availability of diverse approaches and the need to tune the associated parameters. Our research has shown the general applicability of the ensemble strategy in solving diverse problems by using different populated optimization algorithms. Further details can be found in our publications listed below. Codes of some of the publications are available on request.

In the context of time series forecasting, the above work shows that the direct links in RVFL improve the performance in a statistically significant manner. The direct links can be compared to the time delayed line in the finite impulse response filter (FIR). Randomization range can also be tuned to improve performance. RVFL with direct input-output links slightly outperformed SVR also. 18 time series datasets were used in this study.

Promising Future Research Direction: There is an increase in research activities using randomized neural networks. Researchers are encouraged to further investigate the role of direct links in RVFL and other randomized neural network configurations.

The below 3 papers on NN are identified as HOT papers and/or highly cited papers by Science Citation Index (or Web of Science) as of 15^th Feb 2019.