The problem of the Flexible Pickup and Delivery Services Problem (FPDSP) arises from the actual needs of multi-warehouse management strategies and is one of the key challenges in the current urban distribution logistics industry. The problem aims to quickly calculate the route planning in complex scenarios to ensure that the total traveling time of the vehicle is minimized while meeting the time window requirements. To address this problem, we propose a deep reinforcement learning method based on the Actor-Critic algorithm to quickly calculate the approximate optimal solution of FPDSP. Specifically, we propose a Transformer Model with Parallel Encoders (TMPE). The model efficiently extracts order features through parallel encoders and then uses serial decoders to optimize the fusion of feature information to optimize the order selection process. In addition, we designed a reward function to reduce the number of repeated pickups made by the vehicle at the same consignor’s location between different orders, thereby effectively reducing the vehicle’s total travel time. Experimental results show that our method can quickly find feasible solutions to the problem compared with heuristic methods on seven different datasets. At the same time, compared with all baseline methods, the number of optimal solutions of our method reaches 14, which significantly improves the problem-solving ability. This result provides a new solution for optimizing pickup and delivery logistics in multiple warehouses in cities in the future.

The visual prompts have provided an efficient manner in addressing visual cross-domain problems. Previous works introduce domain prompts to tackle the classification Test-Time Adaptation (TTA) problem by placing image-level prompts on the input and fine-tuning prompts for each target domain. However, since the image-level prompts mask out continuous spatial details in the prompt-allocated region, it will suffer from inaccurate contextual information and limited domain knowledge extraction, particularly when dealing with dense prediction TTA problems. To overcome these challenges, we propose a novel Sparse Visual Domain Prompts (SVDP) approach, which applies minimal trainable parameters (e.g., 0.1%) to pixels across the entire image and reserves more spatial information of the input. To better apply SVDP in extracting domain-specific knowledge, we introduce the Domain Prompt Placement (DPP) method to adaptively allocates trainable parameters of SVDP on the pixels with large distribution shifts. Furthermore, recognizing that each target domain sample exhibits a unique domain shift, we design Domain Prompt Updating (DPU) strategy to optimize prompt parameters differently for each sample, facilitating efficient adaptation to the target domain. Extensive experiments were conducted on widely-used TTA and continual TTA benchmarks, and our proposed method achieves state-of-the-art performance in both semantic segmentation and depth estimation tasks.

Cluster assignment is crucial for analyzing single-cell RNA sequencing (scRNA-seq) data, essential for studying cell diversity and biological functions at the single-cell level. Deep learning-based clustering methods have been extensively applied in scRNA-seq data analysis. However, current deep models frequently overlook the deep exploration of potential gene correlation, resulting in networks lacking guidance from gene correlation information when learning hidden layer features. This paper proposes an adaptive feedback fuzzy clustering method based on deep gene correlation mining(scDCMAC). The deep gene correlation mining network employs contrast learning to compute the contrast loss on two cell similarity matrices, constructed from denoised autoencoder hidden layer features and original input features, to uncover potential gene correlation. The adaptive feedback clustering network further denoises the hidden layer features using a self-attention mechanism, computes the affiliation matrix, and supervises the denoising autoencoder’s dimensionality reduction using the affiliation matrix. Experiments on 13 real scRNA-seq datasets demonstrate that scDCMAC excels in clustering accuracy and stability, showing great potential in scRNA-seq sample analysis.

Single-cell sequencing technologies allow for a high-resolution examination of various biological activities, such as the genome, transcriptome, and epigenetic modifications, across thousands of cells. In contrast, single-omics approaches fall short in fully capturing the intricate molecular mechanisms. In this regrad, integrating multi-omics data can potentially provide a more comprehensive understanding of fundamental biological processes. Although sequencing technologies have emerged that can simultaneously obtain different omics from single cells, they are limited by throughput and the types of omics. A more common scenario is that data from different omics are measured independently, resulting in unpaired cell compositions. Therefore, efficient algorithms are needed to integrate and analyze unpaired multi-omics data.In this study, we propose a novel framework for integrating unpaired single-cell multi-omics data based on the Kolmogorov-Arnold deep embedding model (scKADE). This framework introduces Kolmogorov–Arnold Networks (KAN) to map different omics data to a unified common feature space for adversarial alignment. Furthermore, to assess the contribution of different cells in data integration, we employ an unsupervised weight estimation algorithm to enhance the alignment quality.

Neural 3D reconstruction from multi-view images has recently attracted increasing attention from the community. Existing methods normally learn a neural field for the whole scene, while it is still under-explored how to reconstruct a target object indicated by users. Considering the Segment Anything Model (SAM) has shown effectiveness in segmenting any 2D images, in this paper, we propose NTO3D, a novel high-quality Neural Target Object 3D (NTO3D) reconstruction method, which leverages the benefits of both neural field and SAM. We first propose a novel strategy to lift the multi-view 2D segmentation masks of SAM into a unified 3D occupancy field. The 3D occupancy field is then projected into 2D space and generates the new prompts for SAM. This process is iterative until convergence to sepa-rate the target object from the scene. After this, we then lift the 2D features of the SAM encoder into a 3D feature field in order to improve the reconstruction quality of the target object. NTO3D lifts the 2D masks and features of SAM into the 3D neural field for high-quality neural target object 3D reconstruction. We conduct detailed experiments on several benchmark datasets to demonstrate the advantages of our method. The code will be available at: https://github.com/ucwxb/NTO3D.

Photo finishing tuning aims to automate the manual tuning process of the photo finishing pipeline, like Adobe Lightroom or Darktable. Previous works either use zeroth-order optimization, which is slow when the set of parameters increases, or rely on a differentiable proxy of the target finishing pipeline, which is hard to train. To overcome these challenges, we propose a novel goal-conditioned reinforcement learning framework for efficiently tuning parameters using a goal image as a condition. Unlike previous approaches, our tuning framework does not rely on any proxy and treats the photo finishing pipeline as a black box. Utilizing a trained reinforcement learning policy, it can efficiently find the desired set of parameters within just 10 queries, while optimization based approaches normally take 200 queries. Furthermore, our architecture utilizes a goal image to guide the iterative tuning of pipeline parameters, allowing for flexible conditioning on pixel-aligned target images, style images, or any other visually representable goals. We conduct detailed experiments on photo finishing tuning and photo stylization tuning tasks, demonstrating the advantages of our method.

BACKGROUND: Personalized therapy has been at the forefront of cancer care, making cancer treatment more effective. Since cancer patients respond individually to drug therapy, predicting the sensitivity of each patient to specific drugs is very helpful to apply therapeutic agents. Traditional methods focus on node (molecular) information but ignore relevant interactions among different nodes, which has very limited application in complex situations, such as cancer drug responses in real clinical practice. RESULTS: Treatment evaluation with Quantified Network (TreeQNet) is a webserver which could predict sensitivity to drugs for patients through the innovative use of proteomic and phosphoproteomic network from tumor tissues. CONCLUSION: TreeQNet service: http://bioinfo.ustc.edu.cn/ . TreeQNet source code: https://github.com/Really00/treeqnet-web-front/ .