Improved Search Technique Research Articles

Improving convergence in swarm algorithms by controlling range of random movement

Swarm intelligence algorithms are stochastic algorithms, i.e. they perform some random movement. This random movement imparts the algorithms with exploration capabilities and allows them to escape local optima. Exploration at the start of execution helps with thorough inspection of the search/solution space. However, as the algorithm progresses, the focus should ideally shift from exploration to exploitation. This shift would help the algorithm to enhance existing solutions and improve its convergence capabilities. Hence if the range of random movement is not kept in check, it may limit an algorithm’s convergence capabilities and overall efficiency. To ensure that the convergence of an algorithm is not compromised, an improved search technique to reduce range of uniform random movement was recently proposed for bat algorithm. Uniform distribution and levy distribution are the most commonly used random distributions in swarm algorithms. In this paper, the applicability of the improved search technique over different swarm algorithms employing uniform and levy distributions, as well as Cauchy distribution has been studied. The selected algorithms are firefly algorithm, cuckoo search algorithm, moth search algorithm, whale optimization algorithm, earthworm optimization algorithm and elephant herding optimization algorithm. The resultant variants of each of these algorithms show improvement upon inclusion of the improved search technique. Hence results establish that the improved search technique has positive influence over swarm algorithms employing different random distributions.

New insights on modeling news dissemination on nuclear issues

Using a modified epidemiological model, the dissemination of news by media agents after the occurrence of large scale disasters was studied. A modified compartmented model was developed in a previous paper presented at INAC 2007 in which, the Chernobyl nuclear accident (1986) and the Concorde airplane crash (2000) were used as a base for the study. Now the model has been applied to a larger and more diverse group of events – nuclear, non-nuclear and naturally caused disasters. To be inclusive, old and recent events from various regions of the world were selected. A more robust news repository was used, and improved search techniques were developed to ensure that the scripts would not contain false positive news. The same model was used but with improved non-linear embedded simulation optimization algorithms to generate the parameters of interest for our model. Individual parameters and some specific combination of these allow a number of interesting perceptions on how the nature of the accident/disaster gives rise to different profiles of growth and decay of the news. In our studies, events involving nuclear causes generate news repercussion with more explosive/robust surge profiles and longer decaying tails than those of other nature. As a consequence of these differences, public opinion and policy makers are also much more sensitive to some issues than to others. The model, through its epidemiological parameters, shows in quantitative manner how “nervous” the media content generators are with respect to nuclear installations and how resilient this negative feelings toward nuclear can be.

Protein Design by Directed Evolution

While nature evolved polypeptides over billions of years, protein design by evolutionary mimicry is progressing at a far more rapid pace. The mutation, selection, and amplification steps of the evolutionary cycle may be imitated in the laboratory using existing proteins, or molecules created de novo from random sequence space, as starting templates. However, the astronomically large number of possible polypeptide sequences remains an obstacle to identifying and isolating functionally interesting variants. Intelligently designed libraries and improved search techniques are consequently important for future advances. In this regard, combining experimental and computational methods holds particular promise for the creation of tailored protein receptors and catalysts for tasks unimagined by nature.

Motion Compensated Interframe Image Prediction

Interframe motion estimation of subblocks based on improved search techniques is developed. These techniques are based on minimizing the mean difference between the subblock in question in the present frame and the displaced subblock in the previous frame. The performance of the motion compensated prediction developed here is investigated for various block sizes and is compared to other techniques.