An overview of the self-organizing map algorithm, on which the papers in this issue are based, is presented in this article.

The Self-Organizing Map

Matrix Factorization Techniques for Recommender Systems

1.Introduction 2.Fundamental Modeling and Control of Unmanned Small-Scale and Mi-niature Helicopters 3.Autonomous Control of a Mini Quadrotor Vehicle Using LQG Controllers 4.Modeling and Control of an Autonomous Quad-Tilt-Wing (QTW) UAV 5.Linearlization and Identification of Helicopter Model for Hierarchical Control Design 6.Analysis of the Autorotation Maneuver in Small-Scale Helicopters and Application for Emergency Landing 7.Autonomous Acrobatic Flight based on Feedforward Sequence Control for Small Unmanned Helicopter 8.Mathematical Modeling and Nonlinear Control of VTOL Aerial Vehicles 9.Formation Flight Control of Multiple Autonomous Helicopters Using Predictive Control 10.Guidance and Navigation Systems for Small Aerial Robots 11.Design and Implementation of a Low-Cost Attitude Quaternion Sensor 12.Vision-Based Navigation and Visual Servoing of Mini Flying Machines 13.Autonomous Indoor Flight and Precise Auto-Landing Using Infrared and Ultrasonic Sensors

Autonomous flying robots : unmanned aerial vehicles and micro aerial vehicles

A systematic review of blockchain scalability: Issues, solutions, analysis and future research

The utility model provides an unmanned aerial vehicle, the reciprocating impact tunnel drilling machine comprises a machine body, the fuselage is inside to be equipped with microcontroller and power, still includes the follow a plurality of wings that the fuselage peripherad extends, a plurality of wings are followed fuselage circumference evenly distributed, the end of every wing all is equipped with air multiplier, every all be equipped with a plurality of inlet ports on the wing, on inlet port and the same wing be connected with inlet channel between the air multiplier, the inside of wing is equipped with turbofan and first motor, the turbofan is located in the inlet channel, the turbofan with first motor links to each other, first motor with microcontroller links to each other. The utility model discloses simple structure, the modern design, convenient to use adopts the air multiplier to replace the mode of traditional screw in order to provide unmanned aerial vehicle flight thrust for unmanned aerial vehicle flight is reliable, the sound construction, and personnel can accidentally not injured to the flight in -process.

Unmanned aerial vehicle

Some neural network can be trained by transfer learning, which uses a pre-trained neural network as the source task, for a small target task’s dataset. The performance of the transfer learning depends on the knowledge (i.e., layers) selected from the pre-trained network. At present, this knowledge is usually chosen by humans. The transfer learning method PathNet automatically selects pre-trained modules or adjustable modules in a modular neural network. However, PathNet requires modular neural networks as the pre-trained networks, therefore non-modular pre-trained neural networks are currently unavailable. Consequently, PathNet limits the versatility of the network structure. To address this limitation, we propose Stepwise PathNet, which regards the layers of a non-modular pre-trained neural network as the module in PathNet and selects the layers automatically through training. In an experimental validation of transfer learning from InceptionV3 pre-trained on the ImageNet dataset to networks trained on three other datasets (CIFAR-100, SVHN and Food-101), Stepwise PathNet was up to 8% and 10% more accurate than finely tuned and from-scratch approaches, respectively. Also, some of the selected layers were not supported by the layer functions assumed in PathNet.

/pdf/stepwise-pathnet-a-layer-by-layer-knowledge-selection-based-58qfravwln.pdf

Stepwise PathNet: a layer-by-layer knowledge-selection-based transfer learning algorithm.

Blockchain – a distributed and public database of transactions – has become a platform for decentralized applications. Despite its increasing popularity, blockchain technology faces a scalability problem: the throughput does not scale with the increasing network size. Thus, in this paper, we propose a scalable blockchain protocol to solve the scalability problem. The proposed method was designed based on a proof of stake (PoS) consensus protocol and a sharding protocol. Instead of transactions being processed by the whole network, the sharding protocol is employed to divide unconfirmed transactions into transaction shards and to divide the network into network shards. The network shards process the transaction shards in parallel to produce middle blocks. Middle blocks are then combined into a final BLOCK in a timestamp recorded on the blockchain. Experiments were performed in a simulation network consisting of 100 Amazon EC2 instances. The latency of the proposed method was approximately 27 s and the maximum throughput achieved was 36 transactions per second for a network containing 100 nodes. The results of the experiments indicate that the throughput of the proposed protocol increases with the network size. This confirms the scalability of the proposed protocol.

Scalable Blockchain Protocol Based on Proof of Stake and Sharding

The performance of transfer learning in convolutional neural networks depends on the selection of which layers are to be learned again and which are not. Generally, layers selection is performed manually; however, as the number of layers increases, the layers selection process becomes increasingly difficult. Thus, we propose a method to select an effective layers in transfer learning automatically using a genetic algorithm. In the proposed method, a genotype representing which layers’ weights are updated or fixed in transfer learning is considered, and we achieve efficient layers selection in the way that a genotype with high validation accuracy is survived during genotype selection. Experiments are performed using the InceptionV3 network that pre-trained ImageNet as source images with transfer learning to CIFAR-100 as target images. Experimental results demonstrate that the test data accuracy in an ensemble of models whose layers are selected by the genetic algorithm is 15% and 12% greater than that of models trained by from-scratch and fine-tuning, respectively. In general transfer learning approach, layers on the output side are selected as adjustable layers; however, it is found that the distribution of the selected layers as an effective adjustable layers obtained by the genetic algorithm extends to the entire network. Transfer learning using a genetic algorithm may successfully capture the characteristics of a convolutional neural network’s structure.

Transfer Learning Layer Selection Using Genetic Algorithm

We designed a connected drone that can be connected to surrounding drones by electromagnets. However, since the magnetic field was disturbed by the magnetic force of the electromagnets, the magnetic sensors could not be used. Thus, since magnetic sensors play an essential role in the attitude estimation of drones, the problem was that the drone could not perform high-precision autonomous flying control. As a result, to substitute the use of magnetic sensors, this study proposes a method of integrating the IMU and multiple GPSs using the Kalman filter. The proposed method can perform highly accurate attitude estimation by adding the velocity vector to the observation vector. First, a comparison experiment shows that the Kalman filter is the most suitable method for the attitude estimation of connected drones among other methods, including the Q-Method, Madgwick filter, and Kalman filter. Next, we show that the addition of the velocity vector improved the roll and pitch angle estimation accuracy by about 10% and the yaw angle estimation accuracy by about 5% in comparison with the conventional method.

Attitude Estimation by Kalman Filter Based on the Integration of IMU and Multiple GPSs and Its Application to Connected Drones

A principal purpose of recommender systems is to induce positive user actions such as clicks and purchases. The performance of a recommender is typically evaluated in terms of prediction accuracy; a recommender is considered to be better if a larger number of its recommended items are purchased. However, accurate prediction is not enough for increasing user actions. The items might have been purchased even without recommendations, that is, the purchases are not triggered by the recommendations. Conversely, a user might be reluctant to purchase any items and a mere recommendation alone cannot motivate the user to purchase an item even if the recommended item is the best among all the items. In this work, we pursue recommender systems for triggering user actions. For this purpose, we tackle two issues: 1) uplift optimization and 2) persuasive explanation. For the first issue, we propose uplift-based evaluation and optimization methods for recommenders. Uplift, which is defined as an increase in user actions caused by recommendations, cannot be observed directly. We apply a causal inference framework to estimate the average uplift for the offline evaluation of recommenders. For optimization, we derive positive and negative samples specific to uplift and construct pointwise and pairwise optimization methods. Through experiments with three public datasets, we demonstrate the effectiveness of our optimization methods in improving uplift. For the second issue, we propose a new explanation style using context. The context style explanation presents contexts suitable for consuming the recommended items. The exhibited contexts make users imagine situations of items' usage and motivate them to purchase the items. Via a crowdsourcing-based user study, we confirm that the persuasiveness of our explanation style is better than conventional styles. The hybridization of context style with other styles further improves the persuasiveness.

Shin Kawai

Papers

Stepwise PathNet: a layer-by-layer knowledge-selection-based transfer learning algorithm.

Scalable Blockchain Protocol Based on Proof of Stake and Sharding

Transfer Learning Layer Selection Using Genetic Algorithm

Attitude Estimation by Kalman Filter Based on the Integration of IMU and Multiple GPSs and Its Application to Connected Drones

Action-Triggering Recommenders: Uplift Optimization and Persuasive Explanation