GuideBP: Guiding Backpropagation Through Weaker Pathways of Parallel Logits
This addresses a training efficiency problem for researchers and practitioners using multi-branch CNNs, though it appears to be an incremental improvement over existing gradient distribution methods.
The paper tackles the problem of gradient distribution in convolutional neural networks with parallel logits, where traditional methods distribute gradients equally across all pathways. The proposed GuideBP approach guides gradients through weakest pathways, achieving better performance than traditional column merging techniques across various multi-objective scenarios.
Convolutional neural networks often generate multiple logits and use simple techniques like addition or averaging for loss computation. But this allows gradients to be distributed equally among all paths. The proposed approach guides the gradients of backpropagation along weakest concept representations. A weakness scores defines the class specific performance of individual pathways which is then used to create a logit that would guide gradients along the weakest pathways. The proposed approach has been shown to perform better than traditional column merging techniques and can be used in several application scenarios. Not only can the proposed model be used as an efficient technique for training multiple instances of a model parallely, but also CNNs with multiple output branches have been shown to perform better with the proposed upgrade. Various experiments establish the flexibility of the learning technique which is simple yet effective in various multi-objective scenarios both empirically and statistically.