Efficient and High-quality Recommendations via Momentum-incorporated Parallel Stochastic Gradient Descent-Based Learning

A recommender system (RS) relying on latent factor analysis usually adopts stochastic gradient descent (SGD) as its learning algorithm. However, owing to its serial mechanism, an SGD algorithm suffers from low efficiency and scalability when handling large-scale industrial problems. Aiming at addressing this issue, this study proposes a momentum-incorporated parallel stochastic gradient descent (MPSGD) algorithm, whose main idea is two-fold: a) implementing parallelization via a novel data-splitting strategy, and b) accelerating convergence rate by integrating momentum effects into its training process. With it, an MPSGD-based latent factor (MLF) model is achieved, which is capable of performing efficient and high-quality recommendations. Experimental results on four high-dimensional and sparse matrices generated by industrial RS indicate that owing to an MPSGD algorithm, an MLF model outperforms the existing state-of-the-art ones in both computational efficiency and scalability.      

An optimal deployable bandwidth aggregation system

The explosive increase in data demand coupled with the rapid deployment of various wireless access technologies have led to the increase of number of multi-homed or multi-interface enabled devices. Fully exploiting these interfaces has motivated researchers to propose numerous solutions that aggregate their available bandwidths to increase overall throughput and satisfy the end-user’s growing data demand. These solutions, however, do not utilize their interfaces to the maximum without network support, and more importantly, have faced a steep deployment barrier. In this paper, we propose an optimal deployable bandwidth aggregation system (DBAS) for multi-interface enabled devices. We present the DBAS architecture that does not introduce any intermediate hardware, modify current operating systems, modify socket implementations, nor require changes to current applications or legacy servers. The DBAS architecture is designed to automatically estimate the characteristics of applications and dynamically schedule various connections and/or packets to different interfaces. We also formulate our optimal scheduler as a mixed integer programming problem yielding an efficient solution. We evaluate DBAS via implementation on the Windows OS and further verify our results with simulations on NS2. Our evaluation shows that, with current Internet characteristics, DBAS reaches the throughput upper bound with no modifications to legacy servers. It also highlights the significant enhancements in the response time introduced by DBAS, which directly enhances the user experience.     

Design development of aggregates cooling systems for hot weather concreting

Using hot aggregates, in concrete production, results in a drop in compressive strength of the produced concrete. Various methods have been proposed for cooling concrete aggregates. This paper proposes new two designs for aggregates cooling systems for various production rate demands. Conveyor system for small to moderate production rates and rotating drum for high production rates. Simulation of the heat flow during the cooling process over the conveyor and through the drum are analyzed with the objective of understanding the effect of the various design parameters and achieving minimum cooling time with the least possible power. Finite element models for the new designs are proposed and discussed. The results of the finite element analysis of the new designs are presented for various initial conditions and cooling rates.     

Bioconversion of shrimp shell waste for the production of antioxidant and chitosan used as fruit juice clarifier

The pH, proteolytic activity, extent of demineralisation and deprotenisation of shrimp waste were studied during 7 days of fermentation using Pseudomonas aeruginosa A2. After 3 days, pH dropped from 7.0 to 4.4 and then remained constant. Simultaneously, a demineralisation of 92% was achieved. However, protease activity reached its highest level (1230 U mL^?1) after 1 day of incubation, and a protein removal of 90% was achieved. Chitin obtained was converted to chitosan. This chitosan, with 73% deacetylation, was tested for clarification of different fruit juices. It was observed that low concentrations of chitosan (below to 1%) greatly increase the clarity of juices without affecting the nutritional value. The antioxidant activity of the hydrolysates produced during fermentation was tested. Hydrolysate obtained after 3 days showed the strongest scavenging activity (90%), which was comparable to the positive control BHA; however, that obtained after 1 day exhibited the highest ferric‐reducing antioxidant power (OD 700 nm = 1.7).     

Techno‐economic optimization of hybrid power system using genetic algorithm

This paper presents an optimum sizing methodology to optimize the hybrid energy system (HES) configuration based on genetic algorithm. The proposed optimization model has been applied to evaluate the techno‐economic prospective of the HES to meet the load demand of a remote village in the northern part of Saudi Arabia. The optimum configuration is not achieved only by selecting the combination with the lowest cost but also by finding a suitable renewable energy fraction that satisfies load demand requirements with zero rejected loads. Moreover, the economic, technical and environmental characteristics of nine different HES configurations were investigated and weighed against their performance. The simulation results indicated that the optimum wind turbine (WT) selection is not affected only by the WT speed parameters or by the WT rated power but also by the desired renewable energy fraction. It was found that the rated speed of the WT has a significant effect on optimum WT selection, whereas the WT rated power has no consistent effect on optimal WT selection. Moreover, the results clearly indicated that the HES consisting of photovoltaics (PV), WT, battery bank (Batt) and diesel generator (DG) has superiority over all the nine systems studied here in terms of economical and environmental performance. The PV/Batt/DG hybrid system is only feasible when wind resource is very limited and solar energy density is high. On the other hand, the WT/Batt/DG hybrid system is only feasible at high wind speed and low solar energy density. It was also found that the inclusion of batteries reduced the required DG and hence reduced fuel consumption and operating and maintenance cost. Copyright © 2014 John Wiley & Sons, Ltd.     

An iterative learning approach for network contraction: Path finding problem in stochastic time‐varying networks

Path finding problem has a broad application in different fields of engineering. Travel time uncertainty is a critical factor affecting this problem and the route choice of transportation users. The major downside of the existing algorithms for the reliable path finding problem is their inefficiency in computational time. This study aims to develop a network contraction approach to reduce the network size of each specific origin and destination (OD) pair in stochastic time‐dependent networks. The network contraction is based on the comparison of optimistic and pessimistic solutions resulting from minimum and maximum travel time realizations of a Monte‐Carlo simulation (MCS)‐based approach. In this respect, the researchers propose a learning approach to utilize the information of the realizations in the initial iterations of the MCS approach. Implementation of this approach is in place for several OD pairs of two real‐world large‐scale applications. First, it is calibrated for the Chicago downtown network; the performance and accuracy of the proposed approach are investigated by comparing the results against that of the approach without any network contraction. In addition, the Salt Lake City network illustrates the transferability of the approach to other networks. The results demonstrate significant computational improvements, with an acceptable accuracy level relative to the approach without network contraction.     

Convolutional neural networks for the classification of chest X-rays in the IoT era

Multimedia Tools and Applications - Chest X-ray medical imaging technology allows the diagnosis of many lung diseases. It is known that this technology is frequently used in hospitals, and it is...