Adequate alignment and interaction for cross-modal retrieval

Background Cross-modal retrieval has attracted widespread attention in many cross-media similarity search applications,particularly image-text retrieval in the fields of computer vision and natural language processing.Recently,visual and semantic embedding(VSE)learning has shown promising improvements in image text retrieval tasks.Most existing VSE models employ two unrelated encoders to extract features and then use complex methods to contextualize and aggregate these features into holistic embeddings.Despite recent advances,existing approaches still suffer from two limitations:(1)without considering intermediate interactions and adequate alignment between different modalities,these models cannot guarantee the discriminative ability of representations;and(2)existing feature aggregators are susceptible to certain noisy regions,which may lead to unreasonable pooling coefficients and affect the quality of the final aggregated features.Methods To address these challenges,we propose a novel cross-modal retrieval model containing a well-designed alignment module and a novel multimodal fusion encoder that aims to learn the adequate alignment and interaction of aggregated features to effectively bridge the modality gap.Results Experiments on the Microsoft COCO and Flickr30k datasets demonstrated the superiority of our model over state-of-the-art methods.

Digital fixed-frequency hysteresis current control of a BLDC motor applied for aerospace electrically powered actuators

In the conventional cascade control structure of aerospace electrically powered actuators, the current（or electromagnetic torque） loop plays a critical role in realizing a rapid response for a digitally controlled Brush Less Direct Current（BLDC） motor. Hysteresis Current Control（HCC） is an effective method in improving the performance of current control for a BLDC motor.Nevertheless, the varying modulating frequency in the traditional HCC causes severe problems on the safety of power devices and the electromagnetic compatibility design. A triangular carrier-based fixed-frequency HCC strategy is expanded by relaxing the constraints on the rising and descending rates of the winding current to advance the capability of HCC to realize fixed-frequency modulation in the steady state. Based on that, a new flexible-bound-size quasi-fixed-frequency HCC is proposed, and the range feasible to realize fixed-frequency modulation control can cover the entire running process in the steady state. Meanwhile, a corresponding digital control strategy is designed,and four digitalization rules are proposed to extend the capacity to achieve fixed-frequency modulation control to the unsteady working state, that is, a novel fixed-frequency modulation is realized.Simulation and experimental results prove the effectiveness of this improved fixed-frequency HCC strategy.

Overcoming thermo-optical dynamics in broadband nanophotonic sensing

Advances in integrated photonics open up exciting opportunities for batch-fabricated optical sensors using high-quality-factor nanophotonic cavities to achieve ultrahigh sensitivities and bandwidths.The sensitivity improves with increasing optical power;however,localized absorption and heating within a micrometer-scale mode volume prominently distorts the cavity resonances and strongly couples the sensor response to thermal dynamics,limiting the sensitivity and hindering the measurement of broadband time-dependent signals.Here,we derive a frequency-dependent photonic sensor transfer function that accounts for thermo-optical dynamics and quantitatively describes the measured broadband optomechanical signal from an integrated photonic atomic force microscopy nanomechanical probe.Using this transfer function,the probe can be operated in the high optical power,strongly thermo-optically nonlinear regime,accurately measuring low-and intermediate-frequency components of a dynamic signal while reaching a sensitivity of 0.7fm/Hz^(1/2) at high frequencies,an improvement of=10x relative to the best performance in the linear regime.Counterintuitively,we discover that a higher transduction gain and sensitivity are achieved with lower quality-factor optical modes for low signal frequencies.Not limited to optomechanical transducers,the derived transfer function is generally valid for describing the small-signal dynamic responses of a broad range of technologically important photonic sensors subject to the thermo-optical effect.