Abstract:Band selection plays a crucial role in hyperspectral image classification by removing redundant and noisy bands and retaining discriminative ones. However, most existing deep learning-based methods are aimed at dealing with a specific band selection dataset, and need to retrain parameters for new datasets, which significantly limits their <a class="link-external link-http" href="http://generalizability.To" rel="external noopener nofollow">this http URL</a> address this issue, a novel multi-teacher multi-objective meta-learning network (M$^3$BS) is proposed for zero-shot hyperspectral band selection. In M$^3$BS, a generalizable graph convolution network (GCN) is constructed to generate dataset-agnostic base, and extract compatible meta-knowledge from multiple band selection tasks. To enhance the ability of meta-knowledge extraction, multiple band selection teachers are introduced to provide diverse high-quality experiences.strategy Finally, subsequent classification tasks are attached and jointly optimized with multi-teacher band selection tasks through multi-objective meta-learning in an end-to-end trainable way. Multi-objective meta-learning guarantees to coordinate diverse optimization objectives automatically and adapt to various datasets simultaneously. Once the optimization is accomplished, the acquired meta-knowledge can be directly transferred to unseen datasets without any retraining or fine-tuning. Experimental results demonstrate the effectiveness and efficiency of our proposed method on par with state-of-the-art baselines for zero-shot hyperspectral band selection.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is the insufficient generalization ability of existing deep - learning methods in hyperspectral band selection. Specifically, most of the existing deep - learning - based methods can only handle specific band - selection datasets, and need to retrain model parameters when encountering new datasets, which greatly limits their versatility. To solve this problem, the paper proposes a new multi - teacher multi - objective meta - learning network (M3BS) for zero - shot hyperspectral band selection. By introducing the multi - teacher strategy and the multi - objective meta - learning framework, M3BS aims to extract dataset - independent meta - knowledge, so as to achieve direct transfer application to unseen datasets without any retraining or fine - tuning. ### Main Contributions 1. **General - purpose GCN**: A generalizable graph convolution network (GCN) is constructed, which can extract consistent meta - knowledge from different datasets, enabling M3BS to perform hyperspectral band selection in zero - shot situations. 2. **Multi - teacher diversity integration**: By introducing multiple representative and reliable teacher models, additional supervision signals are provided, enhancing the generalization ability and adaptability of M3BS to unseen samples in zero - shot learning. 3. **Multi - objective meta - learning optimization**: A multi - objective meta - learning framework is adopted to automatically coordinate multiple optimization objectives and improve multiple band - selection tasks from multiple datasets simultaneously, so that the obtained meta - knowledge can be directly transferred from seen samples to unseen samples. ### Method Overview 1. **Spatial - spectral general - purpose GCN**: A spatial - spectral graph structure is constructed, representing the bands of the hyperspectral image and their spatial and spectral relationships as vertices and edges of the graph. The dataset - independent basic features are extracted through GCN and decomposed into learnable dataset - independent bases and dataset - specific coefficients. 2. **CNN classifier**: A CNN - based classifier is attached to evaluate the effect of the selected band combinations and output the classification probability distribution. 3. **Multi - teacher diversity integration**: Multiple teacher models are introduced to provide diverse high - quality experiences to enhance the ability of meta - knowledge extraction. 4. **Multi - objective meta - learning optimization**: Through the uncertainty - based multi - objective meta - learning process, the band - selection and classification tasks are jointly optimized to ensure that the model has good generalization ability. In conclusion, this paper aims to improve the generalization ability of the hyperspectral band - selection model through the multi - teacher multi - objective meta - learning framework, enabling it to be directly applied to unseen datasets without retraining or fine - tuning.

Multi-Teacher Multi-Objective Meta-Learning for Zero-Shot Hyperspectral Band Selection

An Unsupervised Band Selection Method Via Contrastive Learning for Hyperspectral Images

An Unsupervised Band Selection Method Based on Triple Constraints and Attention Network for Hyperspectral Imagery

Cross-Domain Heterogeneous Hyperspectral Image Classification Based on Meta-Learning with Task-Adaptive Loss Function.

A Band Selection Method with Masked Convolutional Autoencoder for Hyperspectral Image

Graph Meta Transfer Network for Heterogeneous Few-Shot Hyperspectral Image Classification

Learning a Deep Ensemble Network with Band Importance for Hyperspectral Object Tracking.

High-Efficiency Hyperspectral Unmixing Based on Band Selection

Unsupervised Band Selection of Hyperspectral Images via Multi-dictionary Sparse Representation

Unsupervised Band Selection for Hyperspectral Imagery Classification Without Manual Band Removal.

Deep Reinforcement Learning for Band Selection in Hyperspectral Image Classification

BS-Nets: An End-to-End Framework For Band Selection of Hyperspectral Image

Multiobjective band selection approach via an adaptive particle swarm optimizer for remote sensing hyperspectral images

Band Selection Based Hyperspectral Unmixing

An Advanced Hyperspectral Band Selection Approach Based on Mutual Information.

Cross-Domain Meta-Learning Under Dual-Adjustment Mode for Few-Shot Hyperspectral Image Classification

Unsupervised Hyperspectral Band Selection With Multigraph Integrated Embedding and Robust Self-Contained Regression

Double Deep Q-Network for Hyperspectral Image Band Selection in Land Cover Classification Applications

A multi-strategy integrated multi-objective artificial bee colony for unsupervised band selection of hyperspectral images

BSFormer: Transformer-Based Reconstruction Network for Hyperspectral Band Selection

Multi-task multi-objective evolutionary network for hyperspectral image classification and pansharpening