The Simpson’s Paradox in the Offline Evaluation of Recommendation Systems

Recommendation systems are often evaluated based on user’s interactions that were collected from an existing, already deployed recommendation system. In this situation, users only provide feedback on the exposed items and they may not leave feedback on other items since they have not been exposed to them by the deployed system. As a result, the collected feedback dataset that is used to evaluate a new model is influenced by the deployed system, as a form of closed loop feedback. In this article, we show that the typical offline evaluation of recommender systems suffers from the so-called Simpson’s paradox. Simpson’s paradox is the name given to a phenomenon observed when a significant trend appears in several different sub-populations of observational data but disappears or is even reversed when these sub-populations are combined together. Our in-depth experiments based on stratified sampling reveal that a very small minority of items that are frequently exposed by the deployed system plays a confounding factor in the offline evaluation of recommendation systems. In addition, we propose a novel evaluation methodology that takes into account the confounder, i.e., the deployed system’s characteristics. Using the relative comparison of many recommendation models as in the typical offline evaluation of recommender systems, and based on the Kendall rank correlation coefficient, we show that our proposed evaluation methodology exhibits statistically significant improvements of 14% and 40% on the examined open loop datasets (Yahoo! and Coat), respectively, in reflecting the true ranking of systems with an open loop (randomised) evaluation in comparison to the standard evaluation.

Covariance matrix transformation method in all-earth integrated navigation considering coordinate frame conversion

In the exploration of polar region, navigation is one of the most important issues to be resolved. To avoid the limitations of single navigation coordinate frame, the navigation systems usually use different navigation coordinate frames in polar and nonpolar region, such as the north-oriented geographic frame and the grid frame. However, the error states and covariance matrix are related with the definition of navigation coordinate frame, since the coordinate frame conversion will cause the integrated navigation Kalman filter overshoot and error discontinuity. To solve this problem, the transformation relationship of error states defined in different frames is deduced, whereby the covariance matrix transformation relationship is also analyzed. On this basis, covariance transformation-based the open-loop and the closed-loop Kalman filter integrated navigation algorithms are proposed. The effectiveness of algorithms is verified by flight tests with rotational strapdown inertial navigation system (RSINS)/global navigation satellite system (GNSS) integrated navigation system.

Enhancement of dynamic performance of wave energy converter by introducing a flow control

The stabilization of power take-off (PTO) is imperative especially under circumstances of fluctuating input wave energy. In this paper, a flow control valve is introduced to optimize the transient process of the hydraulic PTO, which can contribute to a quicker adjustment and a stronger stability. Under variations of input power and load torque in transient process, an open-loop control method and a closed-loop control method are proposed as the opening law of the above valve, and the hydraulic motor speed, the pressure at the accumulator inlet and the generated power are chosen as indicators to examine the regulation performance. Then, the synergic effect of the flow control valve and the accumulator in the transient process is discussed. The effectiveness of the two presented control methods on the fluctuation suppression is respectively tested and compared in both regular wave and irregular wave situations via simulation. To validate the practical effectiveness of the proposed methods, field experiments are conducted. The results demonstrate that the open-loop control can only improve the damping ability of the hydraulic PTO in the speed raising stage, while the closed-loop control can improve the stability both in the speed raising stage and in the load increasing stage.

Red Fluorescent Protein Expression in Transgenic Founder of Angelfish (Pterophyllum sp) Driven by Zebrafish Myosin Light Chain 2 Promoter

Angelfish (Pterophyllum sp.) are attractive fish popular with aquarists. The introduction of fluorescent protein genes into angelfish has been reported, but specific techniques have not been revealed. This study aimed to develop a strategy to produce red fluorescent protein (RFP) transgenic angelfish driven by the myosin light chain 2 (mylz2) promoter from zebrafish. A 1999 bp Mylz2 promoter fragment was isolated from zebrafish muscle genomic DNA. This promoter fragment was then cloned into the plasmid pDsred2-1 open-loop at restriction enzyme SacI and AgeI sites to create the final transgene construct pMylz2-RFP. Angelfish embryos at one cell stage were microinjected with approximately 100 pg of the plasmid pMylz2-RFP. From 524 microinjected embryos, 16 successfully hatched, while 12 showed red fluorescence signals. Two larvae survived to 2 months of age. They showed significant red fluorescence expression in the muscles, suggesting that the angelfish could be used as potential transgenic founders to evaluate the next generation of stable red fluorescence expression transgenic fish.

Increase the Efficiency of an Induction Motor Feed from Inverter for Low Frequencies by Combining Design and Control Improvements

Speed-controlled induction motors have the most significant potential for energy savings. The greatest problems with obtaining high efficiency occur in motors with a wide range of rotational speed regulation, as in the motors for driving industrial washing machines under consideration. While for the highest speeds, the dominant phenomenon is at field weakening. The problem is obtaining the optimal size of the magnetic flux for low rotation speed to prevent excessive saturation increasing current, and reduction of efficiency. This problem is usually solved by selecting the appropriate control for an already built machine. The authors propose a combination of activities when designing the motor structure with the selection of proper control, which allows for high efficiency. Since the drive does not require precise speed control or obtaining the required dynamics, it was possible to use an inexpensive control in an open loop, avoiding the cost of transmitters. Furthermore, the number of design parameters that are subject to change is significantly limited by technological factors and the available space in the washing machine. Proper parameter selection was made using a peripheral method assisted by field-circuit simulations. The proposed approach can be used in designing structures and selecting motors controls for other applications.

Range and Accuracy Improvement of Artillery Rocket Using Fixed Canards Trajectory Correction Fuze

This paper presents a two-phase guidance and control algorithm to extend the range and improve the impact point accuracy of a 122-mm rocket using a fixed canards trajectory correction fuze. The guidance algorithm consists of a unique glide and correction phase of the rocket trajectory that is activated after the flight’s apex. The glide phase operates in an open-loop configuration where guidance commands are generated to increase the range of the rocket. In contrast, the correction phase operates in a closed-loop configuration where the Impact Point Prediction method based on Modified Projectile Linear Theory is used as a feedback channel to correct the range and drift errors. The proposed fixed canards trajectory correction fuze has a simple and reliable single channel roll-orientation control configuration. The rocket trajectory model consists of a 7-DOF non-linear dynamic model of a dual-spin rocket configuration with a fixed canards correction fuze mounted at the nose. A Monte Carlo simulation of the rocket’s inertial and launch point perturbations show that the fixed canards fuze with the proposed guidance algorithm can double the range of the rocket without changing the rocket motor thrust-time curve. At the same time, the rocket’s accuracy can also be improved beyond the results of an unguided rocket.

Vector Control and Modeling of Modular Multilevel High Voltage Converter for Rail Vehicle Traction

Rail transit plays an important role in the social and economic life of China and even all countries in the world, especially some populous countries or regions. The traction drive system of rail vehicle provides three-phase AC with adjustable voltage and frequency for the traction motor, controls the speed and torque of the traction motor, and then controls the operation of the vehicle. The modular multilevel converter has the advantages of low harmonic, good power quality of output waveform, high reliability, no input filtering and power compensation, and is suitable in the field of frequency conversion. In this work, the open-loop scalar control and vector closed-loop control of modular multi-level high-voltage inverter were adopted. It was found that driven by modular multi-level variable frequency vector control system, asynchronous motor not only has less harmonic content of voltage and current waveform, but also its speed regulation characteristics have been improved.

Manipulation of free-floating objects using Faraday flows and deep reinforcement learning

The ability to remotely control a free-floating object through surface flows on a fluid medium can facilitate numerous applications. Current studies on this problem have been limited to uni-directional motion control due to the challenging nature of the control problem. Analytical modelling of the object dynamics is difficult due to the high-dimensionality and mixing of the surface flows while the control problem is hard due to the nonlinear slow dynamics of the fluid medium, underactuation, and chaotic regions. This study presents a methodology for manipulation of free-floating objects using large-scale physical experimentation and recent advances in deep reinforcement learning. We demonstrate our methodology through the open-loop control of a free-floating object in water using a robotic arm. Our learned control policy is relatively quick to obtain, highly data efficient, and easily scalable to a higher-dimensional parameter space and/or experimental scenarios. Our results show the potential of data-driven approaches for solving and analyzing highly complex nonlinear control problems.

Suboptimal control synthesis for state and input delayed quadratic systems

In this article a suboptimal linear-state feedback controller for multi-delay quadratic system is investigated. Optimal state and input coefficients resulting from the expansion over a hybrid basis of block pulse and Legendre polynomials are first obtained by formulating a nonlinear programming problem. Afterwards, suboptimal control gains are found by solving a least square problem constructed with optimal coefficients of the open loop study. A sufficient condition for the exponential stability of the closed loop is obtained from generalized Grönwall–Bellman lemma. The Van de Vusse chemical reactor case is handled allowing to validate the proposed technique.