In a printing production process, paper often exhibits various surface defects due to objective factors. These defects manifest in different forms and are typically small and densely distributed, posing a challenge for existing object detection methods. To tackle this issue, this research reports a paper product appearance inspection system that combines a dual-layer template and an attention mechanism. The system effectively addresses the problem of surface detect detection by matching template images with the images to be inspected. This approach ensures robust detection performance and reduces false negatives through the utilization of the improved You Only Look Once v7 algorithm. The enhanced algorithm incorporates diverse attention mechanisms that effectively utilize features, thereby enhancing the model’s capability for defect detection. The system’s development includes the creation of a visual interface that provides a customized defect detection solution for the printing authority. Through comprehensive ablation experiments, the algorithm demonstrates significant improvements in both detection accuracy and speed. This system enables automatic inspection of the appearance of single-page paper products and automates the removal of products with surface defects. Consequently, it achieves a high level of automation in the inspection and handling processes for single-page paper products.

Fiber optic pressure sensors have been widely used in recent decades. Most of the fiber optic pressure sensors use fiber gratings and Fabry-Perot cavity. However, they have the drawbacks of in achieving distributed sensing. In this paper, a lossless fiber pressure sensor based on a structure composed of two ratios of the polydimethylsiloxane (PDMS) and single-mode fiber is proposed. The PDMS has often been used as a flexible structural material. When the pressure is acting radially on the PDMS, the internal fiber is driven to produce an axial strain of a larger range, and the optical frequency domain reflectometer (OFDR) is used to demodulate the fiber strain. The experimental results prove that the designed sensor has a repeatable response with a sensitivity of 17.519 με/N within the range of 20 N (2548 kPa). The minimum pressure resolution and spatial resolution of the sensor can reach values of 0.5 N and 0.5 cm, respectively. Compared to a single ratio PDMS sensor, which has a non-linear response to the pressure, the range is increased by about 1.4 times. Based on the obtained results, the proposed sensor has application prospects in pressure sensors.

In deep learning based stock trading strategy models, most of the research just use simple convolutional neural networks (CNN) to process stock data. But task-specific neural network structures have been proposed extensively, and their effectiveness has been demonstrated in computer vision (CV) and natural language processing (NLP) tasks. In this paper, we proposed a multi-scale convolutional neural feature extraction network (MS-CNN) for stock data, which can better extract stock trend features and thus make better decisions. The network structure inspired by the human stock trading model: in human behavior, we do not only look at a single set of stock data, but rather combine all the stock data, such as opening, closing, and trading volume, to make a comprehensive judgment. And humans will consider the current stock trend on different time scales, such as 3-Day Line and 5-Day Line. This is consistent with the two-dimensional convolution kernels commonly used in CV tasks, so we used convolution kernels of 3×3 and 5×5 in the network with two-dimensional convolution size and constructed a novel network structure for stock data. With double deep Q networks (DDQN) algorithm, we get the best performance for our network. The experimental results show that we can obtain high yield on the datasets of Dow Jones (DJI), AAPLE (AAPL), and General Electric (GE).

Algorithmic trading is playing an increasingly important role in the financial market, achieving more efficient trading strategies by replacing human decision-making. Among numerous trading algorithms, deep reinforcement learning is gradually replacing traditional high-frequency trading strategies and has become a mainstream research direction in the field of algorithmic trading. This paper introduces a novel approach that leverages reinforcement learning with human feedback (RLHF) within the double DQN algorithm. Traditional reward functions in algorithmic trading heavily rely on expert knowledge, posing challenges in their design and implementation. To tackle this, the reward-driven double DQN (R-DDQN) algorithm is proposed, integrating human feedback via a reward function network trained on expert demonstrations. Additionally, a classification-based training method is employed for optimizing the reward function network. The experiments, conducted on datasets including HSI, IXIC, SP500, GOOGL, MSFT, and INTC, show that the proposed method outperforms all baselines across six datasets and achieves a maximum cumulative return of 1502% within 24 months.