Generative Adversarial Networks (GANs) are most popular generative frameworks that have achieved compelling performance. They follow an adversarial approach where two deep models generator and discriminator compete with each other. They have been used for many applications especially for image synthesis because of their capability to generate high quality images. In past few years, different variants of GAN have proposed and they produced high quality results for image generation. This paper conducts an analysis of working and architecture of GAN and its popular variants for image generation in detail. In addition, we summarize and compare these models according to different parameters such as architecture, training method, learning type, benefits and performance metrics. Finally, we apply all these methods on a benchmark MNIST dataset, which contains handwritten digits and compare qualitative and quantitative results. The evaluation is based on quality of generated images, classification accuracy, discriminator loss, generator loss and computational time of these models. The aim of this study is to provide a comprehensive information about GAN and its various models in the field of image synthesis. Our main contribution in this work is critical comparison of popular GAN variants for image generation on MNIST dataset. Moreover, this paper gives insights regarding existing limitations and challenges faced by GAN and discusses associated future research work.
Neural Computing and Applications - Aim of this research is to explore the strength of evolutionary and swarm intelligence techniques for parameter identification of control autoregressive moving... 相似文献
New approaches for the determination of the radiation characteristics of planar arrays are proposed. Conventionally, the best compromise directivity-side lobes level is achieved using Dolf-Chebyshev amplitude distribution. However, classical distribution forms present disadvantages in calculation time, possibility in analyzing larger arrays and the array factor is only optimum in the two principle planes. To overcome these problems, the excitation amplitudes are written under the modal form. The obtained radiated fields compared with measurement results validate the technique. It is also shown that square arrays provide improved performances from point of view of directivity compared to those of rectangular arrays with comparable size. However, above specific array size, the directivity saturates. For this, another modification of the array factor of Chebyshev square array is proposed. The modification preserves the radiation pattern configuration (same number of null radiation directions and sidelobes) and improves both directivity and side lobe levels. Analytical expressions of the array factor, the directivity, maximum half-power beamwidth, and the maximum spacing between elements have been derived. It has been shown that the use of this new technique results in an improved directivity with respect to the classical Chebyshev for a square arrays identical in size, element spacing, direction of the main lobe and side lobes level. 相似文献
The fatty acid-based imidazolium gemini surfactants containing hydroxy-substituted spacers were synthesized and characterized by different spectral techniques. The micellization behavior of product compounds was studied by determining their critical micelle concentration (cmc) values from steady state fluorescence and conductivity measurements. The thermodynamic parameters of micellization of synthesized gemini surfactants confirmed that the micelle formation is a thermodynamically favorable, endothermic and an entropy driven phenomenon. The surfactants were further analyzed for their cytotoxic nature by treating them upon human embryonic kidney cell (HEK-293) line and were found to be significantly low cytotoxic. The interesting features like low cmc values, non-cytotoxic and eco-friendly nature of the gemini surfactants may offer new and more efficient substitutes to the conventional drug delivery systems. Thus the gemini surfactants may serve as promising candidates in the biomedical and pharmaceutical fields. 相似文献
Conventional 3D woven T-shaped preforms were modified for improving joint/ peel off strength of associated T-shaped polymeric composites. Preforms were modified at weaving level by preferred yarns orientations for better performance in associated composites. Major modifications studied are; the addition of supporting layer, single or double-crossing in joint layers and crossing along with supporting layer (in a single sample). Novel derivatives were compared with conventionally used (on-loom and off-loom) T-shaped 3D woven composites. Microscopic analysis was also carried out to analyze the orientation of yarns and analysis of samples from failure point after testing. By combined variation of crossing and supporting layers improvement of 47.37% in peel off strength and 70.37% in impact strength was observed. Modified T shapes can serve to be a good replacement for conventional T-shaped stiffener (used to avoid folding under loads) and in joints. 相似文献
Lifetime management of underground pipelines is mandatory for safe hydrocarbon transmission and distribution systems. The use of high-density polyethylene tubes subjected to internal pressure, external loading and environmental variations requires a reliability study in order to define the service limits and the optimal operating conditions. In service, the time-dependent phenomena, especially creep, take place during the pipe lifetime, leading to significant strength reduction. In this work, the reliability-based assessment of pipe lifetime models is carried out, in order to propose a probabilistic methodology for lifetime model selection and to determine the pipe safety levels as well as the most important parameters for pipeline reliability. This study is enhanced by parametric analysis on pipe configuration, gas pressure and operating temperature. 相似文献