Extracting coupling matrix from given S-parameters can be viewed as an inverse problem for microwave filters, which is of importance for filter design and tuning. In this letter, a regularized deep belief network (R-DBN) is proposed to handle this inverse modeling problem. The training of an R-DBN consists of two steps. First, in unsupervised training, to accommodate the characteristics of input data, this model is constructed with a series of traditional restricted Boltzmann machines (RBMs), which are equipped with a continuous version of transfer function for continuous data processing. In addition, this training can provide suitable weights and bias for the following step. Second, in supervised training, Bayesian regularization is employed to increase modeling ability and prevent overfitting. Two experiments with different simulation environments are illustrated, and the calibration results show high accuracy and robustness in a more intelligent way using this method.

Air quality has always been one of the most important environmental concerns for the general public and society. Using machine learning algorithms for Air Quality Index (AQI) prediction is helpful for the analysis of future air quality trends from a macro perspective. When conventionally using a single machine learning model to predict air quality, it is challenging to achieve a good prediction outcome under various AQI fluctuation trends. In order to effectively address this problem, a genetic algorithm-based improved extreme learning machine (GA-KELM) prediction method is enhanced. First, a kernel method is introduced to produce the kernel matrix which replaces the output matrix of the hidden layer. To address the issue of the conventional limit learning machine where the number of hidden nodes and the random generation of thresholds and weights lead to the degradation of the network learning ability, a genetic algorithm is then used to optimize the number of hidden nodes and layers of the kernel limit learning machine. The thresholds, the weights, and the root mean square error are used to define the fitness function. Finally, the least squares method is applied to compute the output weights of the model. In order to verify the predictive ability of GA-KELM, based on the collected basic data of long-term air quality forecast at a monitoring point in a city in China, the optimized kernel extreme learning machine is applied to predict air quality ( SO ₂ , NO ₂ , _{PM 10} , CO , O ₃ , PM _2.5 concentration and AQI), with comparative experiments based on convolutional neural networks, recurrent neural networks and long short-term memory. The results show that the proposed model trains faster and makes more accurate predictions.

The purpose of filters' inverse modeling is to acquire the values of physical or geometrical parameters for the specified electrical response. In this article, a dimensionality reduction (DR) strategy is proposed through the artificial neural network (ANN) approach, for the first time, to simplify the filters' inverse modeling. First, all physical/geometrical variables are divided into two parts, partial variables and other ones, through a selection mechanism, and only partial variables are to be determined in the inverse modeling solution. Then, an ANN for DR (DR-ANN) is constructed to form the relationship between partial variables and other ones for given electrical response. Third, the predicted partial variables are set as known input and then run the DR-ANN to acquire the value of remaining variables. The proposed DR-ANN framework can be easily introduced to many existing inverse modeling problems. Specifically, a multivalued neural network approach and a well-known differential evolutional (DE) optimization method are combined with DR-ANN. The improvement of the modified methods is demonstrated with some examples, including coupling matrix extraction and waveguide filter optimization. The experiment results show that the proposed solution is more robust with faster convergence in comparison with existing methods.

This paper proposes a real-time winter road surface condition (RSC) monitoring solution that automatically generates descriptive RSC information in terms of snow and ice coverage by using images from fixed traffic and weather cameras. Several state-of-the-art pre-trained deep neural networks are customized and fine-tuned to address a specific domain, classifying the amount of snow coverage on a road surface. A thorough evaluation is conducted to identify and select the best model. This evaluation uses an extensive set of experiments to test the accuracy and generalization of each model and uses transfer-learning to fine-tune each of the pre-trained models on independent images from different traffic and weather cameras. The transferability of each model, relationship between model performance and data size, and the system settings of each model are then examined. Lastly, three online weight calibration methods are proposed to automatically update the model in new environments. The result shows that re-training the model using images from a mixed set of cameras has the most promising results.

Operational optimal control (OOC) is an essential component of wastewater treatment process (WWTP). The control variables usually are high-dimensional, nonlinear, and strongly coupled, which can easily fail traditional optimization control methods. Mathematically, these operational variables usually are in the unknown low-dimensional space embedded in the high-dimensional space. Therefore, the OOC problem of WWTP can be resolved as an optimization challenge involving low-dimensional space, and the unknown low-dimensional space is presented in the form of a set of controlled variables in a high-dimensional space, which is normal in real-world industries. Here, a dimension-reducible data-driven optimization control framework for WWTP is proposed. Considering the difficulty in elucidating the whole space of set points, a neural network is designed to approximate the constraint relationship between control variables. The search process is based on optimization methods in low-dimensional space embedded into Euclidean spaces. Furthermore, the convergence of the process is ensured via mathematical analysis. Finally, the experimental simulation of wastewater treatment revealed that this approach is effective for an optimal solution in control systems.

The rapid development of communication transmission, including 6G technology, is creating increasing challenges for real-world object recognition tasks in transportation, which now must operate within complex external environments and the requirement of time efficiency. Although machine learning-based hybrid intelligence has attracted significant attention and achieved much success in recent years, the current models are often ineffective and have poor generalization in extreme weather. This is because the training of a deep learning model is often uncourteous, meaning that the models can easily fail, even during the feature extraction step. An adaptive hybrid attention-based convolutional neural network (AHA-CNN) framework is proposed in this paper to address these shortcomings. First, fuzzy c-means and maximum entropy algorithms are utilized for image feature pre-extraction. A heuristic search-based adaptive attention mechanism is then presented, which adaptively combines the previously extracted features and generates fused images. By applying this mechanism, the key areas of an image are reinforced in a more intelligent and interpretable way, and less important areas are ignored. The processed images are then transferred into a modified region-CNN for further training. Finally, four real-world experiments on traffic sign detection, vehicle license plate recognition, road surface condition monitoring, and pavement disease detection are carried out. Results show that the proposed framework has high testing accuracy compared with other existing methods. The features fused with the cognition mechanism are also easier to interpret.