Evolutionary design of neural network architectures: a review of three decades of research

We present a comprehensive review of the evolutionary design of neural network architectures. This work is motivated by the fact that the success of an Artificial Neural Network (ANN) highly depends on its architecture and among many approaches Evolutionary Computation, which is a set of global-search methods inspired by biological evolution has been proved to be an efficient approach for optimizing neural network structures. Initial attempts for automating architecture design by applying evolutionary approaches start in the late 1980s and have attracted significant interest until today. In this context, we examined the historical progress and analyzed all relevant scientific papers with a special emphasis on how evolutionary computation techniques were adopted and various encoding strategies proposed. We summarized key aspects of methodology, discussed common challenges, and investigated the works in chronological order by dividing the entire timeframe into three periods. The first period covers early works focusing on the optimization of simple ANN architectures with a variety of solutions proposed on chromosome representation. In the second period, the rise of more powerful methods and hybrid approaches were surveyed. In parallel with the recent advances, the last period covers the Deep Learning Era, in which research direction is shifted towards configuring advanced models of deep neural networks. Finally, we propose open problems for future research in the field of neural architecture search and provide insights for fully automated machine learning. Our aim is to provide a complete reference of works in this subject and guide researchers towards promising directions.

     

Multifunctional Gadolinium‐Doped Mesoporous TiO2 Nanobeads: Photoluminescence,Enhanced Spin Relaxation,and Reactive Oxygen Species Photogeneration,Beneficial for Cancer Diagnosis and Treatment

Materials with controllable multifunctional abilities for optical imaging (OI) and magnetic resonant imaging (MRI) that also can be used in photodynamic therapy are very interesting for future applications. Mesoporous TiO₂ sub‐micrometer particles are doped with gadolinium to improve photoluminescence functionality and spin relaxation for MRI, with the added benefit of enhanced generation of reactive oxygen species (ROS). The Gd‐doped TiO₂ exhibits red emission at 637 nm that is beneficial for OI and significantly improves MRI relaxation times, with a beneficial decrease in spin–lattice and spin–spin relaxation times. Density functional theory calculations show that Gd³⁺ ions introduce impurity energy levels inside the bandgap of anatase TiO₂, and also create dipoles that are beneficial for charge separation and decreased electron–hole recombination in the doped lattice. The Gd‐doped TiO₂ nanobeads (NBs) show enhanced ability for ROS monitored via ^?OH radical photogeneration, in comparison with undoped TiO₂ nanobeads and TiO₂ P25, for Gd‐doping up to 10%. Cellular internalization and biocompatibility of TiO₂@x Gd NBs are tested in vitro on MG‐63 human osteosarcoma cells, showing full biocompatibility. After photoactivation of the particles, anticancer trace by means of ROS photogeneration is observed just after 3 min irradiation.     

Cross-property connections for copper–graphite composites

Copper–graphite composite materials in the range of 0–10 vol% of carbon phase were prepared from the mixture of copper and graphite powders by hot isostatic pressing. The microstructure, mechanical (tensile strength, elongation to fracture) and physical (electrical and thermal conductivity) properties of composite samples were investigated, and the cross-property connections were calculated. It was shown that electrical and thermal conductivity cross-property (Lorenz number) is almost constant and increases only slightly (no more than 10 % increase was observed). This implies that in the investigated composition range the Lorenz number of a copper–graphite composite system behaves according to Franz–Wiedemann law for pure metals at constant temperature. On the contrary, the conductivity to tensile strength cross-property connections showed significant linear increase (over 200 % in the investigated composition range) for both electrical conductivity and thermal conductivity of composite materials. The cross-property connections of conductivity to the elongation to fracture exhibit a nonlinear dependence on the volume fraction of graphite.     

Determination of the Optimal Fourier Number on the Dynamic Thermal Transmission

This article represents the result of experimental research on transient heat transfer in a multilayered (heterogeneous) wall. Our non-steady thermal transmission simulation is based on a finite-difference calculation method. The value of a Fourier number shows the similarity of thermal variation in conditional layers of an enclosure. Most scientists recommend using no more than a value of 0.5 for the Fourier number when performing calculations on dynamic (transient) heat transfer. The value of the Fourier number is determined in order to acquire reliable calculation results with optimal accuracy. To compare the results of simulation with experimental research, a transient heat transfer calculation spreadsheet was created. Our research has shown that a Fourier number of around 0.5 or even 0.32 is not sufficient (\({\approx }17\,\%\) of oscillation amplitude) for calculations of transient heat transfer in a multilayered wall. The least distorted calculation results were obtained when the multilayered enclosure was divided into conditional layers with almost equal Fourier number values and when the value of the Fourier number was around 1/6, i.e., approximately 0.17. Statistical deviation analysis using the Statistical Analysis System was applied to assess the accuracy of the spreadsheet calculation and was developed on the basis of our established methodology. The mean and median absolute error as well as their confidence intervals has been estimated by the two methods with optimal accuracy (\({F}_{\mathrm{oMDF}}= 0.177\) and \(F_{\mathrm{oEPS}}= 0.1633\) values).     

An adaptive method based on the improved LPA-ICI algorithm for MRI enhancement

Various diseases are diagnosed using medical imaging used for analysing internal anatomical structures. However, medical images are susceptible to noise introduced in both acquisition and transmission processes. We propose an adaptive data-driven image denoising algorithm based on an improvement of the intersection of confidence intervals (ICI), called relative ICI (RICI) algorithm. The 2D mask of the adaptive size and shape is calculated for each image pixel independently, and utilized in the design of the 2D local polynomial approximation (LPA) filters. Denoising performances, in terms of the PSNR, are compared to the original ICI-based method, as well as to the fixed sized filtering. The proposed adaptive RICI-based denoising outperformed the original ICI-based method by up to 1.32 dB, and the fixed size filtering by up to 6.48 dB. Furthermore, since the denoising of each image pixel is done locally and independently, the method is easy to parallelize.     

A SLIT THROUGH PHOTOGRAPHY

Post-war experimental and conceptual photography in former Yugoslavia has only rarely been the subject of detailed study and interpretation. In considering this period, it is necessary to take into account several factors, including the absence of permanent exhibition spaces for photography, the lack of magazines in which photographic themes were presented and discussed, the impossibility of studying the field of photography and, finally, the inadequate knowledge and application of contemporary criticism and theories of photography. Nevertheless, from the mid-1950s onwards it is possible to note a variety of innovations in the field, in terms of both form and subject-matter. This article considers rare instances of institutional support for progressive photography-related events and unique, intellectual-poetic works. After the break-up of Yugoslavia, there is a dominant tendency to nationalize art created in the former state, thus ignoring the specific Yugoslav cultural field as well as the European context. Based on a methodology which surpasses the national (but still acknowledges it) and searches for meaning within the broader socio-political space to which art is referring, the research aims to change the paradigm of the peripheral position and general ignorance of the circumstances under which this innovative practice emerges.     

Tunable electrochemistry of gold-silver alloy nanoshells

Nano Research - The widespread and increasing interest in enhancing biosensing technologies by increasing their sensitivities and lowering their costs has led to the exploration and application of...