Meta‐heuristic multi‐ and many‐objective optimization techniques for solution of machine learning problems

Recently, multi‐ and many‐objective meta‐heuristic algorithms have received considerable attention due to their capability to solve optimization problems that require more than one fitness function. This paper presents a comprehensive study of these techniques applied in the context of machine learning problems. Three different topics are reviewed in this work: (a) feature extraction and selection, (b) hyper‐parameter optimization and model selection in the context of supervised learning, and (c) clustering or unsupervised learning. The survey also highlights future research towards related areas.     

Data Parallel Neural Dynamics Model for Integrated Design of Large Steel Structures

We have two objectives in creating novel design theories and computational models: automation and optimization. These two aspects are particularly important in design of complex and large engineering structures. In this article, a robust data parallel neural dynamics model is presented for discrete optimization of large steel structures based on the AISC ASD or LRFD specifications. The computational model has been implemented on a CM–5 supercomputer and applied to integrated minimum–weight design of two steel high–rise building structures. The largest example is a 144–story modified tube–in–tube super–high–rise building structure with 20,096 members. Optimization of such a large structure subjected to the highly nonlinear constraints of actual design codes, such as the AISC LRFD code, where nonlinear second–order effects have to be taken into account, has never been attempted before. The computational model developed in this research finds the minimum–weight design for this very large structure subjected to multiple dead, live, and wind loadings in three different directions automatically. This research demonstrates how a new level in design automation is achieved through the ingenious use of a novel computational paradigm and new high–performance computer architecture.     

Tuned Mass Dampers

A review of representative research on tuned massed dampers (TMD) reported in journals in recent years is presented. TMDs are divided into four categories: conventional TMDs, pendulum TMDs (PTMDs), bi-directional TMDs (BTMDs), and tuned liquid column dampers (TLCDs).     

Long-term behavior of nonionic surfactant-added PDMS for self-driven microchips

The outstanding characteristics of polydimethylsiloxane (PDMS) owe its extensive use to the fact that it is a base material for the microfluidic devices manufacturers’. In spite of favorable physical and chemical properties, the hydrophobic surface of PDMS is a handicap when pumping aqueous solutions through microchannels using only capillary forces. There are several techniques to achieve a hydrophilic behavior of PDMS, but most of them face the problem of hydrophobic recovery after a short period of time while most commercial microdevices require long storage and distribution times. The use of surfactant-added PDMS provides a novel method to overcome hydrophobicity and to control the hydrophobic recovery over a long period of time. There are many different types of surfactants and not a deep methodology to choose one in terms of efficiency, clearance and duration of the hydrophilic behavior. This paper has compared three non-ionic surfactants with different critical micelle concentration and chemical composition: Triton X-100, Brij 35 and Tween 20. Short and long-term studies were done using 5-μL deionized water droplet on the surface of the prepared surfactant-added PDMS. The experiments demonstrated that Triton X-100 is more efficient than Brij 35 and Tween 20 since with less concentration it achieves a maximum contact angle of around 23.5°. In terms of hydrophobic recovery, the experiments showed that using surfactants and controlling humidity of samples, hydrophobic recovery of the surfactant-added PDMS was negligible after 2 months. According to these results, the use of PDMS with Triton X-100 and Brij 35 provides a good potential for building capillary driven devices without the need of tedious preprocessing techniques.     

Many-objective control optimization of high-rise building structures using replicator dynamics and neural dynamics model

Recently the authors presented a single-agent Centralized Replicator Controller (CRC) and a decentralized Multi-Agent Replicator Controller (MARC) for vibration control of high-rise building structures. It was shown that the use of agents and a decentralized approach enhances the vibration control system. Two key parameters in the proposed control methodologies using replicator dynamics are the total population (total available resources or the sum of actuators forces) and the growth rate. In the previous research, a sensitivity analysis was performed to determine the appropriate values for the population size and growth rate. In this paper, the aforementioned control methodologies are integrated with a multi-objective optimization algorithm in order to find Pareto optimal values for growth rates with the goal of achieving maximum structural performance with minimum energy consumption. A modified neural dynamic model of Adeli and Park is used in this research to solve the many-objective optimization problem where the Normal Boundary Intersection method is employed to find Pareto optimality. Sample results are presented using a 20-story steel benchmark structure subjected to historical and artificial accelerograms.     

Carbon nanotubes-graft-polyglycerol: Biocompatible hybrid materials for nanomedicine

New biocompatible and water soluble hybrid materials containing multi-wall carbon nanotubes (MWCNTs) as core and hyperbranched polyglycerol (PG) as shell were synthesized successfully. In this work, pristine MWCNTs were opened and functionalized through treatment with acid and polyglycerol was covalently grafted onto their surface by the “grafting from” approach based on in-situ ring-opening polymerization of glycidol. Some short-term In vitro cytotoxicity and hemocompatibility tests were conducted on HT1080 cell line (human Fibrosarcoma), because this epithelial cell line can be one of the first route of entry of the exogenous materials to the vascular system and therefore subsequent interactions with the whole body, in order to investigate their potential application in nanomedicine and to understand the limitation and capability of these material as nanoexcipients in biological systems.     

Scalable Production of Nanographene and Doping via Nondestructive Covalent Functionalization

A new method for top‐down, one‐pot, gram‐scale production of high quality nanographene by incubating graphite in a dilute sodium hypochlorite solution at only 40 °C is reported here. The produced sheets have only 4 at% oxygen content, comparable with nanographene grown by chemical vapor deposition. The nanographene sheets are covalently functionalized using a nondestructive nitrene [2+1] cycloaddition reaction that preserves their π‐conjugated system. Statistical analyses of Raman spectroscopy and X‐ray photoelectron spectroscopy indicate a low number of sp³ carbon atoms on the order of 2% before and 4% after covalent functionalization. The nanographene sheets are significantly more conductive than conventionally prepared nanographene oxide, and conductivity further increases after covalent functionalization. The observed doping effects and theoretical studies suggest sp² hybridization for the carbon atoms involved in the [2+1] cycloaddition reaction leading to preservation of the π‐conjugated system and enhancing conductivity via n‐type doping through the bridging N‐atom. These methods are easily scalable, which opens the door to a mild and efficient process to produce high quality nanographenes and covalently functionalize them while retaining or improving their physicochemical properties.