Traditional channel-wise pruning methods by reducing network channels struggle to effectively prune efficient CNN models with depth-wise convolutional layers and certain efficient modules, such as popular inverted residual blocks. Prior depth pruning methods by reducing network depths are not suitable for pruning some efficient models due to the existence of some normalization layers. Moreover, finetuning subnet with directly removing activation layers would corrupt the original model weights, hindering the pruned model from achieving high performance. To address these issues, we propose a novel depth pruning method for efficient models. Our approach proposes a novel block pruning strategy and progressive training method for the subnet. Additionally, we extend our pruning method to vision transformer models. Experimental results demonstrate that our method consistently outperforms existing depth pruning methods across various pruning configurations. We obtained three pruned ConvNeXtV1 models with our method applying on ConvNeXtV1, which surpass most SOTA efficient models with comparable inference performance. Our method also achieves state-of-the-art pruning performance on the vision transformer model.

Diffusion models have achieved remarkable progress in the field of image generation due to their outstanding capabilities. However, these models require substantial computing resources because of the multi-step denoising process during inference. While traditional pruning methods have been employed to optimize these models, the retraining process necessitates large-scale training datasets and extensive computational costs to maintain generalization ability, making it neither convenient nor efficient. Recent studies attempt to utilize the similarity of features across adjacent denoising stages to reduce computational costs through simple and static strategies. However, these strategies cannot fully harness the potential of the similar feature patterns across adjacent timesteps. In this work, we propose a novel pruning method that derives an efficient diffusion model via a more intelligent and differentiable pruner. At the core of our approach is casting the model pruning process into a SubNet search process. Specifically, we first introduce a SuperNet based on standard diffusion via adding some backup connections built upon the similar features. We then construct a plugin pruner network and design optimization losses to identify redundant computation. Finally, our method can identify an optimal SubNet through few-step gradient optimization and a simple post-processing procedure. We conduct extensive experiments on various diffusion models including Stable Diffusion series and DiTs. Our DiP-GO approach achieves 4.4 x speedup for SD-1.5 without any loss of accuracy, significantly outperforming the previous state-of-the-art methods.

Accurate travel-time prediction of public transport is essential for reliable journey planning in urban transportation systems. However, existing studies on bus travel-/arrival-time prediction often focus only on improving the prediction accuracy of a single bus trip. This is inadequate in modern public transportation systems, where a bus journey usually consists of multiple bus trips. In this paper, we investigate the problem of travel-time prediction for bus journeys that takes into account a passenger's riding time on multiple bus trips, and also his/her waiting time at transfer points (interchange stations or bus stops). A novel framework is proposed to separately predict the riding and waiting time of a given journey from multiple datasets (i.e., historical bus trajectories, bus route, and road network), and combining the results to form the final travel-time prediction. We empirically determine the impact factors of bus riding times and develop a long short-term memory model that can accurately predict the riding time of each segment of the bus lines/routes. We also demonstrate that the waiting time at transfer points significantly impacts the total journey travel time, and estimating the waiting time is non-trivial as we cannot assume a fixed distribution waiting time. In order to accurately predict the waiting time, we introduce a novel interval-based historical average method that can efficiently address the correlation and sensitivity issues in waiting time prediction. Experiments on real-world data show that the proposed method notably outperforms six baseline approaches for all the scenarios considered.