Generative adversarial networks (GANs) have achieved remarkable success with realistic tasks such as creating images, texts, and audio. Combining GANs quantum computing, are thought to an exponential advantage over their classical counterparts due the stronger expressibility of circuits. In this research, a two-qubit silicon photonic chip is created, capable executing arbitrary controlled-unitary (CU) operations generating any 2-qubit pure state, thus making it excellent platform for GANs....

We demonstrated a hybrid quantum-classical generative adversarial network (GAN) with silicon photonic chip capable of generating arbitrary 2-qubit states. The was successfully applied for classical distribution loading and MNIST image generation.

Abstract Generative adversarial networks (GANs) have achieved remarkable success with realistic tasks such as creating images, texts, and audio. Combining GANs quantum computing, are thought to an exponential advantage over their classical counterparts due the stronger expressibility of circuits. In this research, a two‐qubit silicon photonic chip is created, capable executing arbitrary controlled‐unitary () operations generating any pure state, thus making it excellent platform for GANs. To...

Deep learning-based object detection algorithms play an increasingly important role in many computer vision tasks. Many are trying to increase the depth of networks improve their feature expression capacity. However, how better extract essential features from a fixed number training samples is often ignored. Thus, we propose CoReSh network, using prior knowledge intrinsic features, address this problem. First, relevant optical factors which may affect performance explored, and proven be able...

With the continuous development of silicon-based optoelectronic chips, their high-power applications in communication, ranging, and other fields are gradually increasing. However, nonlinear effects silicon (Si) cause significant output power loss when input optical surpasses a certain threshold. This relationship between chip's reduces its working performance. Therefore, importance regulating phenomena chip becomes prominent. In this regard, Sb₂Se₃ is an excellent...

Integrated silicon micro-ring resonator (MRR) has been widely used as on-chip single photon source. In this paper, we introduce an approach to generate frequency-degenerate photons by bidirectionally pumping add-drop within a Sagnac loop configuration. This scheme facilitates the concurrent generation of pairs from both clockwise (CW) and counterclockwise (CCW) directions, transforming MRR into dual-source system. Through CMOS fabrication techniques, realized proposed device conducted...

We propose a complex-valued matrix-vector multiplication method in this work, which make full use of the amplitude and phase information input signal light weight matrix. demonstrate our computing theoretically experimentally. In order to verify theory, photonic chip is designed used setup neural network system, corresponding accuracies reach more than 90%.

We propose and demonstrate a sub-gigahertz bandwidth photonic differentiator employing the self-induced optical modulation effect in silicon-on-insulator micro-ring resonator. The all-passive DIFF is controlled through an all-optical pump-probe scheme. Input Gaussian-like pulses with 7.5ns pulse width are differentiated at high processing accuracy. A semi-analytical model that agrees experimental results also derived. DIFF's energy efficiency higher than 45%, far surpassing all previously...

We propose a two-dimensional beam scanner based on hitless microring switch array. By changing the resonance state of interference between different wavelength selection switches is avoided. The operation complexity steering FPA chip array reduced.

Panoptic segmentation provides both holistic and detailed image parsing information at the pixel instance level. However, computational burdens restrict its applications in real-time scenarios. A potential approach to learn more efficient models is employ knowledge distillation. previous distillation schemes have focused mainly on classification with limited attention given rearession-related tasks which key for panoptic segmentation. In this paper, we establish a logits-based, hints-based,...

In the modern content-based image retrieval systems, there is an increasingly interest in constructing a computationally effective model to predict interestingness of images since measure could improve human-centered search satisfaction and user experience different applications. this paper, we propose unified framework binary based on discriminant correlation analysis (DCA) multiple kernel learning (MKL) techniques. More specially, one hand, reduce feature redundancy describing cues images,...

In the modern content-based image retrieval systems, there is an increasingly interest in constructing a computationally effective model to measure interestingness of images since measurement could, more or less, improve human-centered search satisfaction and user experience different applications. Typically, couple attributes (also called cues) are empirically selected corresponding features capture these cues elaborately leveraged order comprehensively represent content underlying be...