ИСТИНА

In the field of object recognition, deep learning technology based on imitating the nervous system has made revolutionary progress. But due to the high complexity of deep learning models, the model often overfits some unrelated features, such as the background or the high-frequency characteristics of the picture. This will reduce the robustness and reliability of the model. Inspired by the "Lateral Inhibition" mechanism in the biological nervous system, we introduce the lateral inhibition algorithm in deep learning. This algorithm can significantly reduce the model’s dependence on irrelevant information, make the model focus on the target areas of images, and improve the model’s interpretability and robustness. Our lateral suppression algorithm is compatible with the current mainstream deep learning models.