Abstract
An unsupervised method for convolutional neural network (CNN) architecture design is proposed. The method relies on a variable neighborhood search-based approach for finding CNN architectures and hyperparameter values that improve classification performance. For this purpose, t-Distributed Stochastic Neighbor Embedding (t-SNE) is applied to effectively represent the solution space in 2D. Then, k-Means clustering divides this representation space having in account the relative distance between neighbors. The algorithm is tested in the CIFAR-10 image dataset. The obtained solution improves the CNN validation loss by over 15% and the respective accuracy by 5%. Moreover, the network shows higher predictive power and robustness, validating our method for the optimization of CNN design.

Abstract
It is commonly accepted that one of the most important factors for assuring the high performance of an electrical network is the surveillance and the respective preventive maintenance. From a long time ago that TSOs and DSOs incorporate in their maintenance plans the surveillance of the grid, where is included the aerial power lines inspection. Those inspections started by human patrol, including structure climbing when needed and later were substituted by helicopters with powerful sensors and specialised technicians. More recently the Unmanned Aerial Vehicles (UAV) technology has been used, taking advantage of its numerous advantages. This paper addresses the problem of improving the real-time perception capabilities of UAVs for endowing them with capabilities for safe and robust autonomous and semi-autonomous operations. It presents a new vision based power line detection algorithm denoted by PLineD, able to improve the detection robustness even in the presence of image with background noise. The algorithm is tested in real outdoor images of a dataset with multiple backgrounds and weather conditions. The experimental results demonstrate that the proposed approach is effective and able to implemented in real-time image processing pipeline.

Abstract

Abstract
The present study assesses the perceptions of female entrepreneurs in Tanzania regarding the access to bank loans and the difficulties experienced in the process of financing their businesses. Focusing on small-scale businesses, face-to-face interviews were conducted with 75 female entrepreneurs from the Dar es Salaam area. Resorting to double-hurdle estimation models, we conclude that: 1) women who perceive higher discrimination and/or inequality in accessing bank loans, but who also recognize that female entrepreneurs often lack relevant business skills, tend to apply more often for bank loans; 2) women running larger business, operating in the tailoring industry, face fewer difficulties; 3) although highly educated female entrepreneurs apply less for bank loans, formal education acts as a shield to the difficulties faced by women when applying to bank loans; 4) more autonomous and money seeking female entrepreneurs are less likely to report difficulties during the bank loan application process.

Abstract
In the present paper a novel cone shaped nano-fiber (CSNF) pH sensor using multi-mode fiber (MMF) has been fabricated and demonstrated. Three different pH indicators, chlorophenol red, bromothymol blue and cresol red with precursor tetraethyl orthosilicate (TEOS) have been used for fabrication of pH sensing layer. A significant enhancement in sensing properties of pH sensor with TiO2 thin film has been observed. The pH sensor with TiO2 thin film shows the quite high sensitivity (1.16 dBm/pH) as compared to sensor with simple pH coating (0.81 dBm/pH) at 1550 nm with a good linear response. Moreover, the sensor with TiO2 film exhibits fast response time of ~ 25 s for pH values ranging from 4 to 11 with excellent stability and durability. © 2016 Elsevier B.V.

Abstract