Search

We have developed an algorithm for (re)configuration of Coupled Resonator Optical Waveguides filters. We analyze in detail the effect of the phase shift introduced by tunable Mach-Zehnder interferometers. An alternating pattern is proposed to drive every other tunable coupler. We verified this algorithm to realize a user-defined all-pole filter function in circuit simulation.

We propose a novel filter circuit that incorporates a balanced Mach-Zehnder interferometer (MZI) loaded with a double coupled ring resonator. A global optimization method is applied to optimize the proposed auto-regression/moving average (ARMA) filter to a specific target function.

Programmable photonic circuits are optical chips whose functionality can be (re)configured through electronics and software. This programmability enables new functionality, and opens the path to general-purpose photonic chips, similar to electronic FPGAs. This tutorial discusses the operational principles and technology stack for these chips.

In order to predict the future state of friction torque of satellite momentum wheel bearings, firstly, the 450 original data of bearing friction torque were obtained by experiment and the data were divided into 15 groups. Secondly, variation intensity and reliability of each group were calculated based on the grey confidence level and the closest 5 variation intensities were linearly fitted based on the bootstrap-least square method and the estimated truth values and intervals of variation intensity and reliability were obtained based on the maximum entropy method. Finally, the dynamic prediction was realized and prediction error is not more than 4.2%.

In order to solve the defect that the attribute has the same effect during the zero-shot classification decision, the objective weighting method is introduced to learn the attribute weights. Firstly, the convolutional neural network is used to encode the image information, and the two full connected layers are used to encode the semantic attributes, and a visual-semantic joint embedding model is constructed by combining the visual branch. Then, the weight parameters of each attribute are learned by the objective weighting method. The nearest neighbor search is estimated to predict the test category labels by the attributed-based weights. The experimental results show that it makes up for the different importance of attribute to classification decision and improves the recognition accuracy of classification model. At the same time, it is proved that the selecting of embedded spaces will also affect the recognition accuracy.

In the image classification task, traditional model can only recognize annotated image samples, but class labels can't involve all the object categories. In order to reduce the dependence on the labels and recognize unannotated object samples, this paper proposes zero-shot image classification based on attribute. The binary attribute is used as the intermediate knowledge to migrate learned knowledge from training samples domain to test samples domain. Using the classification model of the multi-loss function based on ResNet-50 to predict the object attributes. Then, using an attribute matrix to represent the correspondence between the object class and the attribute. Finally, the result of attribute prediction is combined with the prior knowledge of the attribute matrix to get the category. Compared with the traditional image classification method, the attribute learning model is applied to the zero-shot image classification. The experimental data show that the method improves the recognition accuracy of the image and improves the flexibility of the image classification task, which lays the foundation for the multi-source domain adaptation induction problem.

Photonic chips are becoming more programmable, with connectivity that is reconfigured using electronics and software. This evolution is propelled by artificial intelligence and quantum computing applications. We will discuss more general-purpose circuits that could be deployed in more diverse applications, similar to general-purpose programmable electronics.

We have developed a graph representation of programmable photonic mesh circuits that can be used by pathfinding algorithms. We modified the Dijkstra algorithm to observe only physically possible connections in the graph, demonstrating flexible rerouting in case of node malfunction.

Programmable photonics configure the flow of light on a photonic chip using software and electrical actuators. Such flexible chips are useful for prototyping new applications. We discuss the underlying technologies for hardware and software.