Search

With the advancements in materials and pattern recognition algorithms, electronic nose (E-nose) has been continuously developing and finding increasingly widespread applications. However, E-nose is prone to drift during practical usage. Currently, the primary methods for addressing the issue of sensor drift in sample classification are semi-supervised learning and unsupervised learning. The implementation of supervised learning has been difficult in the past due to the requirement for a large number of up-to-date samples to retrain the models. But now, this challenge can be overcome with the help of meta-learning. We applied the model-agnostic meta-learning (MAML) architecture from meta-learning and designed a residual classifier with learnable per-parameter learning rates. Leveraging the ability of 'learning to learn', our method do not require a large amount of target domain data and there is no need to repeat extensive computations when transferring to any other target domain compared to other algorithms, resulting it highly deployable. Through the evaluation of multiple experiments, we demonstrate the feasibility of our approach and achieve very high classification accuracy on a publicly available long-term drift dataset compared with other methods. • The MAML architecture in meta-learning is utilized to address the issue of gas sensor drift. • Learning a learning rate and direction for each parameter is conducted in our architecture. In the outer loop phase of our method, different learning rates are also applied to different parameters. • Compared to other algorithms, our model only requires two samples per class for knowledge transfer, without the need for extensive repeated training. [ABSTRACT FROM AUTHOR]

Accurate and real-time prediction of gas concentrations is a critical component of intelligent prediction systems and holds significant importance for production safety and quality of life. However, due to the constraints of the network topology in gas sensor networks and the temporal nature of gas concentration variations, concentration prediction has always been considered a significant challenge. To simultaneously capture the spatiotemporal correlations in gas concentration data with spatiotemporal characteristics, this study proposes a fully-connected temporal multilayer graph convolutional network (FTM-GCN). This method combines multiple graph convolutional layers (MGC), fully connected layers, and gated recurrent units (GRU). FTM-GCN exhibits three prominent features: (1) It leverages MGC to learn the topology of gas sensor networks, enhancing the model's ability to graph data and capture spatial characteristics. MGC is incorporated to capture spatial features in the data. (2) GRU is employed to capture the dynamic changes in sensor network data and the temporal characteristics of the data. (3) Continuous fully-connected layers are introduced to enhance the model's performance. Experimental results demonstrate that FTM-GCN effectively captures the spatiotemporal correlations in spatiotemporal data across various prediction perspectives and outperforms GCN, GRU, T-GCN and STGCN. • Spatio-temporal prediction of graph-structured data using an electronic nose. • Successful prediction of ethanol concentrations in space-time. • A new deep learning model called FTM-GCN is proposed which is based on GCN. • The proposed FTM-GCN achieves outstanding regression prediction performance. [ABSTRACT FROM AUTHOR]