Broad Learning

Broad learning systems (BLS) aim to improve the efficiency and effectiveness of deep learning models, particularly in image classification, by employing novel architectures and training methods. Current research focuses on developing efficient convolutional BLS variants, often incorporating multi-scale feature fusion and incremental learning capabilities, with some work exploring the integration of BLS modules into existing deep learning backbones to enhance performance without extensive retraining. These advancements offer potential for significant improvements in computational efficiency and performance across various applications, including medical image analysis and neural architecture search.