Afsaneh Mahanipour

The fast development of Internet-of-Things (IoT) devices and applications has led to vast data collection, potentially containing irrelevant, noisy, or redundant features that degrade learning model performance. These collected data can be processed on either end-user devices (clients) or edge/cloud server. Feature construction is a pre-processing technique that can generate discriminative features and reveal hidden relationships between original features within a dataset, leading to improved performance and reduced computational complexity of learning models. Moreover, the communication cost between clients and edge/cloud server can be minimized in situations where a dataset needs to be transmitted for further processing. In this paper, the first federated feature construction (FFC) method called multi-modal multiple FFC (MMFFC) is proposed by using multimodal optimization and gravitational search programming algorithm. This is a collaborative method for constructing multiple high-level features without sharing clients' datasets to enhance the trade-off between accuracy of the trained model and overall communication cost of the system, while also reducing computational complexity of the learning model. We analyze and compare the accuracy-cost trade-off of two scenarios, namely, 1) MMFFC federated learning (FL), using vanilla FL with pre-processed datasets on clients and 2) MMFFC centralized learning, transferring pre-processed datasets to an edge server and using centralized learning model. The results on three datasets for the first scenario and eight datasets for the second one demonstrate that the proposed method can reduce the size of datasets for about 60%, thereby reducing communication cost and improving accuracy of the learning models tested on almost all datasets.

Recently, a significant interest has been attracted by the potential use of aluminum nanostructures as plasmonic color filters to be great alternatives to the commercial color filters based on dye films or pigments. These color filters offer potential applications in LCDs, LEDs, color printing, CMOS image sensors, and multi-spectral imaging. However, engineering the optical characteristics of these nanostructures to design a color filter with a desired pass-band spectrum and high color purity requires accurate optimization techniques. In this work, an optimization procedure integrating genetic algorithm with FDTD Solutions software is utilized to design plasmonic color filters automatically. Our proposed aluminum nanohole arrays have been realized successfully to achieve additive (red, green, and blue) color filters using the automated optimization procedure. Despite all the considerations for fabrication simplicity, the designed filters feature transmission efficacies of 45-50% with a FWHM of 40 nm, 50 nm, and 80 nm for the red, green, and blue filters, respectively. The obtained results prove an efficient integration of genetic algorithm and FDTD Solutions revealing the potential application of the proposed method for the automated design of similar nanostructures.

One of the most important issues to tackle in data classification has been the existence of non-informative features in feature sets. Therefore, feature construction (FC) is an important pre-processing task to construct discriminating features from the original ones. Gravitational search algorithm (GSA) is a powerful swarm-based metaheuristic algorithm, which has been improved and adapted to represent multiple new constructed features. Most of the swarm-based algorithms entail a population of individuals while one individual would be returned as an optimal solution at the end of the process. In this paper, the solutions' structure of GSA have been changed in a way that each individual can be considered as a part of the solution, and the final result consists of the whole population. Consequently, each individual is a constructed feature aiming to achieve a population of good features. In other words, the proposed method is a novel multiple feature construction (MFC) method based on the GSA which is called in brief GSAMFC. The experimental results on thirteen standard data sets demonstrate that the proposed GSAMFC is highly beneficial for providing suitable and small feature subsets as well as improving the classifier accuracy. The obtained results of GSAMFC and those of the competing algorithms prove the proficiency of the proposed method.

Abstract—Finding the optimal size in VLSI circuits such as
voltage controlled oscillator is one of the challenging tasks for IC
designers. VCO is one of the most widely used blocks in both
digital and analog circuits. In this paper, a new approach has
been presented to find the optimum size of transistors in order to
optimize phase noise and power in ring VCOs. In this work, the
gravitational search algorithm (GSA) is employed to optimize the
size of transistors for the purpose of reaching the best phase noise
and power. Tested differential ring VCOs are simulated in 65nm
Cadence Virtuoso by considering the size of transistors extracted
from GSA in MATLAB environment.

Abstract
Recently, automatic programming approaches have attracted great deal of interest aiming to utilize search techniques to find out
optimal programs in various problems. Genetic programming is the most commonly explored automatic programming technique
which uses genetic algorithm to evolve and discover programs with the tree structure. Herein, we focus on a new gravitational
search algorithm (GSA)-based technique to create computer programs, automatically. This method is called gravitational search
programming (GSP). Using GSA, the approach of generating the tree structure and insertion of internal nodes has been explained
in detail. The GSP has been employed to the symbolic regression (SR) and the problem of feature construction (FC) that are
widely used as a mathematical expression fitting to a given set of data points, and a data preprocessing technique for classification,
respectively. The proficiency of the proposed algorithm has been evaluated and compared with the well-known automatic
programming algorithms as well as C4.5 decision tree classifier. The results have been obtained over ten typical functions and 13
diverse datasets. The obtained results prove the effectiveness of the proposed method in achieving improved accuracy values in
comparison to those of competing algorithms.

— Feature construction can improve the classifier's performance by constructing powerful and distinctive features. Genetic programming algorithm is one the automatic programming methods which provides the possibility of constructing mathematical expressions without any predefined format. As we know, all features of a data set are not suitable; therefore, we believe that if all features are used for feature construction, inappropriate and ineffective features may be constructed. Hence, the main purpose of this paper is firstly, selecting the suitable features, before the construction process, and then constructing a new feature using these selected features. To do so, a fuzzy rough quick feature selection technique is employed. For assessment, the proposed method along with 5 other feature construction methods are applied on 6 standard data sets. The obtained results indicate that the proposed method has more ability in constructing more distinctive features compared to competing approaches. Introduction Nowadays, with the improvement in collecting and saving the data in different sciences, we have confronted a large amount of information. From one hand, the growing increase of information all around the world and from the other hand, the need for fast accomplishment of tasks have resulted in developing software methods to reduce the computational and time complexities.

—Utilization of a plasmonic nanohole array as a color filter, proposes important advantages like the compatibility with CMOS processes. A color filter is an important component for applications, such as LCDs, LEDs, CMOS image sensors, etc. In this article, a set of primary color filters (red, green and blue) are designed by an optimization procedure, employing genetic algorithm integrated with Lumerical FDTD software. The filters consist of a square lattice of nanoholes in an aluminum film on a silicon dioxide substrate. They are suitable for using in CMOS image sensors. Despite the practical restrictions to simplify the fabrication, the optical response of the filters have shown a transmission peak of 30-43 percent with a FWHM of 40 (nm), 50 (nm) and 80 (nm) in accurate resonant wavelengths of red, green, and blue filters, respectively. These results demonstrate the efficacy of the proposed optimization method.

—Feature construction (FC) can improve the classification performance by creating powerful features from the original ones. Particle Swarm Optimization (PSO) is a global search technique that can construct features directly. We believe that using raw features may lead the PSO-based FC method to an inefficient feature, so in this paper, the aim is to select the prominent features before applying PSO-based FC method. The Forward Feature Selection (FFS) method is used for selecting more informative feature subset from original set and constructing feature by the selected ones. Experimental results show that the proposed method can increase the accuracy by constructing a new powerful feature.