Showing papers on "Feedback loop published in 2020"

Feedback Loop and Bias Amplification in Recommender Systems

Abstract: Recommendation algorithms are known to suffer from popularity bias; a few popular items are recommended frequently while the majority of other items are ignored. These recommendations are then consumed by the users, their reaction will be logged and added to the system: what is generally known as a feedback loop. In this paper, we propose a method for simulating the users interaction with the recommenders in an offline setting and study the impact of feedback loop on the popularity bias amplification of several recommendation algorithms. We then show how this bias amplification leads to several other problems such as declining the aggregate diversity, shifting the representation of users' taste over time and also homogenization of the users experience. In particular, we show that the impact of feedback loop is generally stronger for the users who belong to the minority group.

Feedback Loop and Bias Amplification in Recommender Systems

Abstract: Recommendation algorithms are known to suffer from popularity bias; a few popular items are recommended frequently while the majority of other items are ignored. These recommendations are then consumed by the users, their reaction will be logged and added to the system: what is generally known as a feedback loop. In this paper, we propose a method for simulating the users interaction with the recommenders in an offline setting and study the impact of feedback loop on the popularity bias amplification of several recommendation algorithms. We then show how this bias amplification leads to several other problems such as declining the aggregate diversity, shifting the representation of users' taste over time and also homogenization of the users. In particular, we show that the impact of feedback loop is generally stronger for the users who belong to the minority group.

SOM-Hunter: Video Browsing with Relevance-to-SOM Feedback Loop

Abstract: This paper presents a prototype video retrieval engine focusing on a simple known-item search workflow, where users initialize the search with a query and then use an iterative approach to explore a larger candidate set. Specifically, users gradually observe a sequence of displays and provide feedback to the system. The displays are dynamically created by a self organizing map that employs the scores based on the collected feedback, in order to provide a display matching the user preferences. In addition, users can inspect various other types of specialized displays for exploitation purposes, once promising candidates are found.

Distributed Supervisory Secondary Control for a DC Microgrid

Abstract: This article investigates the current sharing and voltage regulation problem of a DC microgrid by a distributed supervisory control method. In the existing secondary control methods, current sharing feedback control loop and voltage regulation feedback control loop are linearly integrated to shift the droop equation. However, the linear combination of two loops may easily offset the control effect of each other, as the control dimension is less than the objective dimension. This fact motivates us to design a distributed supervisory secondary control (DSSC) scheme to achieve a more flexible configuring of each converter. The DSSC consists of two candidate controllers and a supervisor. The two controllers refer to current sharing configuring and DC bus voltage regulation respectively, and the supervisor is to orchestrate the switching between candidate controllers based on the system environment of DC microgrid. Under the proposed DSSC, each converter selects to configure the current sharing or regulates the DC bus voltage according to its need. Theoretical stability analysis is rigorously conducted. Finally, both simulation and experiment results are presented to demonstrate the effectiveness of DSSC.

Reservoir Computing Using a Spin-Wave Delay-Line Active-Ring Resonator Based on Yttrium-Iron-Garnet Film

Abstract: We demonstrate the use of propagating spin waves for implementing a reservoir-computing architecture. Our concept utilizes an active-ring resonator comprising a magnetic thin-film delay line with an integrated feedback loop. These systems exhibit strong nonlinearity and delayed response, two important properties required for an effective reservoir-computing implementation. In a simple design, we exploit the electric control of feedback gain to inject input data into the active-ring resonator and use a microwave diode to read out the amplitude of the spin waves circulating in the ring. We employ two baseline tasks, namely the short-term memory and parity-check tasks, to evaluate the suitability of this architecture for processing time-series data.

Hybrid Frequency-Tunable Parity-Time Symmetric Optoelectronic Oscillator

Abstract: An optoelectronic oscillator (OEO) is a microwave photonic system consisting of an amplified optoelectronic feedback loop to generate a microwave signal. To reduce the phase noise, the feedback loop must be long to ensure a high Q factor, which is usually implemented by incorporating a long optical fiber in the loop. The key limitation of using a long fiber is that a large number of closely spaced longitudinal modes exist in the feedback loop, which makes single-frequency oscillation difficult to achieve. In this article, we propose to incorporate a photonic integrated microdisk resonator (MDR) in an OEO to implement parity-time (PT) symmetry as well as frequency tuning. By employing the reciprocity of light propagation in the MDR, two mutually coupled optoelectronic loops having an identical geometry with one having a gain coefficient and the other a loss coefficient, identical in magnitude, to form a PT-symmetric OEO that support single-frequency oscillation, are implemented. The tuning of the frequency is realized by thermally tuning the MDR. Experimental results show a microwave signal with a frequency that is tunable from 2 to 12 GHz is realized. The phase noise of the generated microwave signal at a frequency of 11.5 GHz is −117.3 dBc/Hz at a 10-kHz frequency offset. The use of a photonic integrated microdisk resonator makes the entire PT symmetric system have high potential for full photonic integration.

Robust chaotic-shift-keying scheme based on electro-optical hybrid feedback system.

Abstract: A chaotic-shift-keying (CSK) scheme is designed based on a chaos system with electro-optical hybrid time delayed feedback structure. By switching the time delay parameter as a message feeding method, the generated chaotic signal is no longer suffered from return map attack, which is an innate vulnerability of traditional CSK. When the coupling of the seed electrical chaotic system and the nonlinear optical time delay feedback loop is carefully weighed, this CSK scheme shows a good robustness in terms of handling noise for transmitting digital signals. By demodulating the digital signals with the chaotic coherent detection method, a bit error rate of 6×10−4 is achieved at the signal-to-noise ratio of 10dB in the simulation. The proposed method has a promising application prospect in some harsh environments.

It's About Time: Minimizing Hardware and Software Latencies in Speech Research With Real-Time Auditory Feedback

Abstract: Purpose Various aspects of speech production related to auditory-motor integration and learning have been examined through auditory feedback perturbation paradigms in which participants' acoustic speech output is experimentally altered and played back via earphones/headphones "in real time." Scientific rigor requires high precision in determining and reporting the involved hardware and software latencies. Many reports in the literature, however, are not consistent with the minimum achievable latency for a given experimental setup. Here, we focus specifically on this methodological issue associated with implementing real-time auditory feedback perturbations, and we offer concrete suggestions for increased reproducibility in this particular line of work. Method Hardware and software latencies as well as total feedback loop latency were measured for formant perturbation studies with the Audapter software. Measurements were conducted for various audio interfaces, desktop and laptop computers, and audio drivers. An approach for lowering Audapter's software latency through nondefault parameter specification was also tested. Results Oft-overlooked hardware-specific latencies were not negligible for some of the tested audio interfaces (adding up to 15 ms). Total feedback loop latencies (including both hardware and software latency) were also generally larger than claimed in the literature. Nondefault parameter values can improve Audapter's own processing latency without negative impact on formant tracking. Conclusions Audio interface selection and software parameter optimization substantially affect total feedback loop latency. Thus, the actual total latency (hardware plus software) needs to be correctly measured and described in all published reports. Future speech research with "real-time" auditory feedback perturbations should increase scientific rigor by minimizing this latency.

A Robust Local Positive Feedback Based Performance Enhancement Strategy for Non-Recycling Folded Cascode OTA

Abstract: In this study, a local positive feedback technique is proposed to enhance the performance of the folded cascode (FC) operational transconductance amplifier (OTA) when working in its saturation region. The enhanced FC-OTA has a higher gain bandwidth (GBW), higher DC gain and higher slew rate (SR) over its conventional counterparts and popular recycling FC-OTAs. The proposed enhanced FC-OTA is implemented by reusing the current generated by the input transistors through the proposed local positive feedback loop to provide additional transconductance and negative resistance. It is then fabricated with SMIC 180nm process. To verify the effectiveness of the proposed FC-OTA performance enhancement strategy, extensive mathematical analysis, small-/large-signal stability analysis, simulation and experimentation are thoroughly investigated in this study. Simulation and experimental results agree well with the theoretical analysis and the expected performances, which corroborates the control design.

A Modified Closed-Loop Voltage Model Observer Based on Adaptive Direct Flux Magnitude Estimation in Sensorless Predictive Direct Voltage Control of an Induction Motor

Abstract: Voltage model observer is a simple and economical technique for flux estimation in induction motor sensorless drives. However, it shows poor performance in low-speed regions. Therefore, in most cases, the use of this observer is limited. On the other hand, using a simple but accurate estimator is important when the control method is sophisticated and requires heavy computation. This issue will be important in predictive control more than the other methods because the accuracy of the prediction is dependent on the flux estimation. In this paper, a modified closed-loop technique based on voltage model observer is proposed for flux estimation. The feedback loop is supported by the proposed model reference adaptive system direct flux magnitude estimation technique. The dependence of the feedback loop on the stator resistance is eliminated. Therefore, the drift error will be avoided. This will allow the method to withstand the high stator resistance error even at low speeds. Also, a new Lyapunov-based technique for the stator resistance estimation via reduced-order model is proposed. By using the proposed observer, the predictive direct voltage control technique is used as the control method in order to achieve a control method that requires low computation. The proposed method is validated through the experimental results.

Key space enhancement of a chaos secure communication based on VCSELs with a common phase-modulated electro-optic feedback.

Abstract: In this paper, a novel chaotic secure communication system based on vertical-cavity surface-emitting lasers (VCSEL) with a common phase-modulated electro-optic (CPMEO) feedback is proposed. The security of the CPMEO system is guaranteed by suppressing the time-delay signature (TDS) with a low-gain electro-optic (EO) feedback loop. Furthermore, the key space is enhanced through a unique secondary encryption method. The first-level encrypted keys are the TDS in the EO feedback loop, and the second-level keys are the physical parameters of the VCSEL under variable-polarization optical feedback. Numerical results show that, compared to the dual-optical feedback system, the TDS of the CPMEO system is suppressed 8 times to less than 0.05 such that they can be completely concealed when the EO gain is 3, and the bandwidth is doubled to over 22 GHz. The error-free 10 Gb/s secure optical transmission can be realized when the time-delay mismatch is controlled within 3 ps. It is shown that the proposed scheme can significantly improve the system performance in TDS concealment, as well as bandwidth and key space enhancement, which has great potential applications in secure dual-channel chaos communication.

Quantum trajectory theory of few-photon cavity-QED systems with a time-delayed coherent feedback

Abstract: We describe an efficient approach to modeling cavity quantum electrodynamics (QED) with a time-delayed coherent feedback using quantum trajectory simulations. An analytical set of equations is derived to exploit the advantages of trajectories in the presence of the non-Markovian dynamics, where adjustments to the standard stochastic dynamics are discussed. In the weak excitation regime, we first verify that our approach recovers known results obtained with other simulation methods and demonstrate how a coherent feedback loop can increase the photon lifetime in typical cavity-QED systems. We also highlight the underlying stochastic dynamics. We then explore the nonlinear few-photon regime of cavity QED, under the restriction of at most one photon at a time in the feedback loop. In particular, we show how feedback affects the cavity photoluminescence (populations versus laser detuning) and describe how one must account for conditioning in the presence of feedback; specifically, the system observables must be conditioned on no photon detections at the feedback output channel occurring.

Optimal Voltage Regulator for Inverter Interfaced Distributed Generation Units Part II: Application

Abstract: The inverter-interfaced distributed generation (IIDG) units are operated either in grid-forming or grid-feeding modes. To this end, the inner control loops are embedded into the inverters’ control system to achieve the control objectives. However, the dynamic performance of IIDG units are greatly affected by their control system and also by the grid's impedance characteristics. Optimal voltage regulator (OVR) previously has been proposed where the conventional inner loops have been replaced by the state feedback loop to compensate for the LC filter dynamics in order to achieve the desired dynamic performance. Utilizing the OVR, a universal model is proposed in this article which is useful for both grid-feeding and grid-forming modes. Each mode of operation is achieved through impedance shaping as a feedback gain adjustment. To this end, the optimal impedance shaping for the universal model is determined based on the desired dynamic performance, control objectives and grid's impedance characteristics. Eigenvalue-analysis and simulation results prove the effectiveness of the universal model in the grid-feeding and grid-forming modes, in unbalanced and harmonic conditions as well as being able to suppress circulating, transient and fault currents in autonomous networked MGs.

Experimental phase control of a 100 laser beam array with quasi-reinforcement learning of a neural network in an error reduction loop

Abstract: An innovative scheme is proposed for the dynamic control of phase in two-dimensional laser beam array. It is based on a simple neural network that predicts the complex field array from the intensity of the induced scattered pattern through a phase intensity transformer made of a diffuser. Iterated phase corrections are applied on the laser field array by phase modulators via a feedback loop to set the array to prescribed phase values. A crucial feature is the use of a kind of reinforcement learning approach for the neural network training which takes account of the iterated corrections. Experiments on a proof of concept system demonstrated the high performance and scalability of the scheme with an array of up to 100 laser beams and a phase setting at 1/30 of the wavelength.

Low-Feedback Sampling Rate Digital Predistortion Using Deep Neural Network for Wideband Wireless Transmitters

Abstract: In this paper, a low-feedback-sampling-rate digital predistortion (DPD) method is proposed for wideband wireless transmitters and radio-frequency power amplifiers (PAs). This DPD method inserts a non-ideal real band-pass filter into the feedback loop to limit the feedback bandwidth. Meanwhile, to recover the band-limited feedback signal, it introduces a signal-recovery module that is based on the deep neural network (DNN). A data-preprocessing technique is proposed to reduce the amount of input data for the DNN and thereby significantly reducing its structural complexity. Also, the use of DNN makes it feasible to implement off-line training. Both the simplicity of the proposed DNN and the availability of off-line training reduce the system complexity in DPD. Experimental validation was performed on a PA driven by wideband orthogonal-frequency-division-multiplexing (OFDM) signals. The results demonstrated the superiority of the proposed method in under-sampling applications. The proposed DPD method achieved the same linearization performance as the state-of-art DPD methods while requiring a sampling rate that was approximately 5% less. Meanwhile, the proposed DPD method has been validated to have stronger anti-noise and generalization capabilities.

An Adaptive Feed-Forward Phase Locked Loop for Grid Synchronization of Renewable Energy Systems under Wide Frequency Deviations

Abstract: Synchronization is a crucial problem in the grid-connected inverter’s control and operation. A phase-locked loop (PLL) is a typical grid synchronization strategy, which ought to have a high resistance to power system uncertainties since its sensitivity influences the generated reference signal. The traditional PLL catches the phase and frequency of the input signal via the feedback loop filter (LF). In general, to enhance the steady-state capability during distorted grid conditions generally, a filter tuned for nominal frequency is used. This PLL corrects large frequency deviations around the nominal frequency, which increases the PLL’s locking time. Therefore, this paper presents an adaptive feed-forward PLL, where the input signal frequency and phase under large frequency deviations are tracked precisely, which overcomes the above-mentioned limitations. The proposed adaptive PLL consists of a feedback loop that reduces the phase error. The feed-forward loop predicts the frequency and phase error, and the frequency adaptive FIR filter reduces the ripples in output, which is due to input distortions. The adaptive mechanism adjusts the gain of the filter in accordance with the supply frequency. This reduces the phase and frequency error and also decreases the locking time under wide frequency deviations. To verify the effectiveness of the proposed adaptive feed-forward PLL, the system was tested under different grid abnormal conditions. Further, the stability analysis has been carried out via a developed prototype test platform in the laboratory. To bring the proposed simulations into real-time implementations and for control strategies, an Altera Cyclone II field-programmable gate array (FPGA) board has been used. The obtained results of the proposed PLL via simulations and hardware are compared with conventional techniques, and it indicates the superiority of the proposed method. The proposed PLL effectively able to tackle the different grid uncertainties, which can be observed from the results presented in the result section.

Reciprocal human-natural system feedback loops within the invasion process

Abstract: Biological invasions are inextricably linked to how people collect, move, interact with and perceive nonnative species. However, invasion frameworks generally do not consider reciprocal interactions between non-native species and people. Non-native species can shape human actions via beneficial or detrimental ecological and socioeconomic effects and people, in turn, shape invasions through their movements, behaviour and how they respond to the collection, transport, introduction and spread of non-natives. The feedbacks that stem from this ‘coupled human and natural system’ (CHANS) could therefore play a key role in mitigating (i.e. negative feedback loops) or exacerbating (i.e. positive feedback loops) ongoing and future invasions. We posit that the invasion process could be subdivided into three CHANS that span from the source region from which non-natives originate to the recipient region in which they establish and spread. We also provide specific examples of feedback loops that occur within each CHANS that have either reduced or facilitated new introductions and spread of established non-native species. In so doing, we add to exisiting invasion frameworks to generate new hypotheses about human-based drivers of biological invasions and further efforts to determine how ecological outcomes feed back into human actions.

A dual-feedback loop model of the mammalian circadian clock for multi-input control of circadian phase.

Abstract: The molecular circadian clock is driven by interlocked transcriptional-translational feedback loops, producing oscillations in the expressions of genes and proteins to coordinate the timing of biological processes throughout the body. Modeling this system gives insight into the underlying processes driving oscillations in an activator-repressor architecture and allows us to make predictions about how to manipulate these oscillations. The knockdown or upregulation of different cellular components using small molecules can disrupt these rhythms, causing a phase shift, and we aim to determine the dosing of such molecules with a model-based control strategy. Mathematical models allow us to predict the phase response of the circadian clock to these interventions and time them appropriately but only if the model has enough physiological detail to describe these responses while maintaining enough simplicity for online optimization. We build a control-relevant, physiologically-based model of the two main feedback loops of the mammalian molecular clock, which provides sufficient detail to consider multi-input control. Our model captures experimentally observed peak to trough ratios, relative abundances, and phase differences in the model species, and we independently validate this model by showing that the in silico model reproduces much of the behavior that is observed in vitro under genetic knockout conditions. Because our model produces valid phase responses, it can be used in a model predictive control algorithm to determine inputs to shift phase. Our model allows us to consider multi-input control through small molecules that act on both feedback loops, and we find that changes to the parameters of the negative feedback loop are much stronger inputs for shifting phase. The strongest inputs predicted by this model provide targets for new experimental small molecules and suggest that the function of the positive feedback loop is to stabilize the oscillations while linking the circadian system to other clock-controlled processes.

Implementation and Characterization of an Integrate-and-Fire Neuron Circuit Using a Silicon Nanowire Feedback Field-Effect Transistor

Abstract: In this article, we propose an integrate-and-fire (IF) neuron circuit using a single-gated silicon nanowire feedback field-effect transistor that utilizes a positive feedback loop. The IF operations are investigated through mixed-mode technology computer-aided design simulations. The neuron circuit composed of four component transistors (plus one capacitor) exhibits a high firing frequency of ~20 kHz and low power and energy consumption of $7~\mu \text{W}$ and $2.9\times 10^{-15}$ J. The firing frequency and spiking voltage can be controlled through external biasing voltages. Our novel neuron circuit demonstrates a promising potential for use in spiking neural network hardware for very large-scale integration.

Theoretical Modeling of the Iterative Properties of User Discovery in a Collaborative Filtering Recommender System

Abstract: The closed feedback loop in recommender systems is a common setting that can lead to different types of biases. Several studies have dealt with these biases by designing methods to mitigate their effect on the recommendations. However, most existing studies do not consider the iterative behavior of the system where the closed feedback loop plays a crucial role in incorporating different biases into several parts of the recommendation steps. We present a theoretical framework to model the asymptotic evolution of the different components of a recommender system operating within a feedback loop setting, and derive theoretical bounds and convergence properties on quantifiable measures of the user discovery and blind spots. We also validate our theoretical findings empirically using a real-life dataset and empirically test the efficiency of a basic exploration strategy within our theoretical framework. Our findings lay the theoretical basis for quantifying the effect of feedback loops and for designing Artificial Intelligence and machine learning algorithms that explicitly incorporate the iterative nature of feedback loops in the machine learning and recommendation process.

Input-to-State Stable Bilateral Teleoperation by Dynamic Interconnection and Damping Injection: Theory and Experiments

Abstract: In bilateral teleoperation, the human operating the master and the environment interacting with the slave are part of the force feedback loop. Yet, both have time-varying and unpredictable dynamics and are challenging to model. Conventional sidestepping of the demand for their models in the stability analysis assumes passive user and environment, and controls the master-communications-slave system to be passive too. This paper circumvents the need for user and environment models in a novel way: it regards their forces as external excitations for a semiautonomous feedback loop, which it outfits with a dynamic interconnection and damping injection controller that renders time-delay teleoperation exponentially input-to-state stable. The controller uses the position and velocity of the local robot and the delayed position transmitted from the other side to robustly synchronize the master and slave under the user and environment perturbations. Lyapunov–Krasovskii stability analysis shows that the strategy, first, can confine the position error between the master and the slave to an invariant set, and, second, can drive it exponentially to a globally attractive set. The approach has practical relevance for telemanipulation tasks with given precision requirements. Experiments with a pair of Geomagic Touch robots validate the strategy compared to state-of-the-art robust position tracking designs.

Using Exploration to Alleviate Closed Loop Effects in Recommender Systems

Abstract: Recommendation systems are often trained and evaluated based on users' interactions obtained through the use of an existing, already deployed, recommendation system. Hence the deployed recommendation systems will recommend some items and not others, and items will have varying levels of exposure to users. As a result, the collected feedback dataset (including most public datasets) can be skewed towards the particular items favored by the deployed model. In this manner, training new recommender systems from interaction data obtained from a previous model creates a feedback loop, i.e. a closed loop feedback. In this paper, we first introduce the closed loop feedback and then investigate the effect of closed loop feedback in both the training and offline evaluation of recommendation models, in contrast to a further exploration of the users' preferences (obtained from the randomly presented items). To achieve this, we make use of open loop datasets, where randomly selected items are presented to users for feedback. Our experiments using an open loop Yahoo! dataset reveal that there is a strong correlation between the deployed model and a new model that is trained based on the closed loop feedback. Moreover, with the aid of exploration we can decrease the effect of closed loop feedback and obtain new and better generalizable models.

Frequency-Response Analysis and Design Rules for Capacitive Feedback Transimpedance Amplifier

Abstract: Precision instrumentation systems, such as optical receivers and other current-output measurement systems, often contain a transimpedance amplifier (TIA). The commonly used resistive feedback TIA (RF-TIA) has a drawback in its practical implementation; the system bandwidth is limited by the feedback resistor and its parasitic capacitance in high-speed applications. To overcome this drawback, a capacitive feedback TIA (CF-TIA) can be used. The gain and bandwidth performance of the CF-TIA are theoretically equivalent to those of the RF-TIA. Although CF-TIA has been introduced in previous studies for reducing thermal noise and addressing the difficulty in integrating a high resistance with the CMOS technology, past analyses have not been in sufficient depth. The ultimate goal of this study is providing designers in this field with helpful design rules and analytical tools. In particular, a transfer function model, the critical design parameters, and a stability analysis are presented. Furthermore, a new CF-TIA configuration involving a simplified integrator dc-feedback loop is introduced, which can be deployed as a single stage rather than the conventional two-stage topology. While the applications considered in this study were high-speed sensor measurements and devices with a high input capacitance, such as a laser position detector, the analytical results obtained herein are also applicable to a variety of other applications such as emerging biosensors and optical communication system.