Multitask Learning Deep Neural Networks Enable Embedded Design of Active Metamaterials.

In this study, we propose and implement a deep neural network framework based on multitask learning aimed at simplifying the forward modeling and inverse design process of photonic devices integrating active metasurfaces. We demonstrate and validate our approach by constructing a continuously tunable bandpass filter that is effective in the midwave infrared region. The key to this filter is the combination of a metasurface and Fabry-Perot (F-P) cavity structure of the tunable phase-change material Ge2Sb2Se4Te (GSST) and the precise control of the crystallinity of the GSST by a silicon-based heater. With the help of a deep learning framework, we are able to independently model the crystallinity and geometric parameters of the filter to maximize the use of GSST tuning for bandpass filtering. Our model discusses the self-attention mechanism and the effect of noise and compares several existing popular algorithms, and the results show that a multitask deep learning strategy can better assist the on-demand reverse design of photonic structures with phase change materials. This opens up new possibilities for personalization and functional extension of optical devices.

A method to classify bone marrow cells with rejected option.

Bone marrow cell morphology has always been an important tool for the diagnosis of blood diseases. Still, it requires years of experience from a suitable person. Furthermore, the outcomes of their recognition are subjective and there is no objective quantitative standard. As a result, developing a deep learning automatic classification system for bone marrow cells is extremely important. However, typical classification machine learning systems only produce classification answers, and will not refuse to generate predictions when the prediction reliability is low. It will pose a big problem in some high-risk systems such as bone marrow cell recognition. This paper proposes a bone marrow cell classification method with rejected option (CMWRO) to classify 11 bone marrow cells. CMWRO is based on convolutional neural networks, ICP and SoftMax (CNN-ICP-SoftMax), containing a classifier with rejected option. When the rejected rate (RR) of tested samples is 0.3143, it can ensure that the precision, sensitivity, accuracy of the accepted samples reach 0.9921, 0.9917 and 0.9944 respectively. And the rejected samples will be handled by other ways, such as identified by doctors. Besides, the method has a good filtering effect on cell types that the classifier is not trained, such as abnormal cells and cells with less sample distribution. It can reach more than 82% in filtering efficiency. CMWRO improves the doctors' trust in the results of accepted samples to a certain extent. They only need to carefully identify the samples that CMWRO refuses to recognize, and finally combines the two results. It can greatly improve the efficiency and accuracy of bone marrow cell recognition.

Accuracy improvement in plastics classification by laser-induced breakdown spectroscopy based on a residual network.

The whole ecosystem is suffering from serious plastic pollution. Automatic and accurate classification is an essential process in plastic effective recycle. In this work, we proposed an accurate approach for plastics classification using a residual network based on laser-induced breakdown spectroscopy (LIBS). To increasing efficiency, the LIBS spectral data were compressed by peak searching algorithm based on continuous wavelet, then were transformed to characteristic images for training and validation of the residual network. Acrylonitrile butadiene styrene (ABS), polyamide (PA), polymethyl methacrylate (PMMA), and polyvinyl chloride (PVC) from 13 manufacturers were used. The accuracy of the proposed method in few-shot learning was evaluated. The results show that when the number of training image data was 1, the verification accuracy of classification by plastic type under residual network still kept 100%, which was much higher than conventional classification algorithms (BP, kNN and SVM). Furthermore, the training and testing data were separated from different manufacturers to evaluate the anti-interference properties of the proposed method from various additives in plastics, where 73.34% accuracy was obtained. To demonstrate the superiority of classification accuracy in the proposed method, all the evaluations were also implemented by using conventional classification algorithm (kNN, BP, SVM algorithm). The results confirmed that the residual network has a significantly higher accuracy in plastics classification and shows great potential in plastic recycle industries for pollution mitigation.

Experimental investigation of laser-induced breakdown spectroscopy assisted with laser-induced fluorescence for trace aluminum detection in steatite ceramics.

Laser-induced breakdown spectroscopy (LIBS) assisted with laser-induced fluorescence (LIF) was introduced to detect trace aluminum in steatite ceramics in this work. The mechanism and transition process of laser-induced aluminum atomic fluorescence in laser-induced plasma was described and discussed. Selective enhancement of LIF and temporal synchronicity between radiation laser and fluorescence were studied. The influences of ablation laser energy, power density of the radiation laser, and interpulse delay were experimentally investigated. The results showed that 60 mJ in ablation laser energy and 4 µs in interpulse delay were the optimal choice for fluorescent intensity. The fluorescence was increased to the saturation level over 4 MW/cm2. Spectral stability improvement of LIBS-LIF was also discovered in this work. The results proved that LIBS-LIF is a feasible and effective modification of LIBS for ceramics analysis.

Sensitive determination of silicon contents in low-alloy steels using micro laser-induced breakdown spectroscopy assisted with laser-induced fluorescence.

Silicon element plays an important role in strength and hardness improvement in steels, but is harmful for ductility, tenacity, and anti-corrosion. Therefore, silicon content should be fast determined in steel manufacture to keep silicon in moderation. In this work, micro laser-induced breakdown spectroscopy assisted with laser-induced fluorescence (µLIBS-LIF) was proposed to sensitively determine silicon in low-alloy steels. The mechanism and excitation selection of laser-induced silicon atomic fluorescence in laser-induced plasma were discussed. Under 10 µm laser-ablated scatters, the results showed that µLIBS-LIF had analytical performance with R2 of 0.9998, LoD of 2.8 µg g-1, and RMSECV of 63 µg g-1, significantly better than µLIBS under their respective optimal conditions. The analytical sensitivity in µLIBS-LIF was even better than macro LIBS in others' works. As our best knowledge, the silicon LoD in LIBS technique was improved to better than 10 µg g-1 in steel matrix for the first time. This work demonstrates µLIBS-LIF as a capable and potential approach for fast determining silicon element in steel industries.