Automated segmentation in medical image analysis is a challenging task that requires large amount of manually labeled data. However, most existing learning-based approaches usually suffer from limited annotated data, which poses major practical problem for accurate and robust segmentation. In addition, semi-supervised are not compared with the supervised counterparts, also lack explicit modeling geometric structure semantic information, both limit accuracy. this work, we present SimCVD,...

Transformers have made remarkable progress towards modeling long-range dependencies within the medical image analysis domain. However, current transformer-based models suffer from several disadvantages: (1) existing methods fail to capture important features of images due naive tokenization scheme; (2) information loss because they only consider single-scale feature representations; and (3) segmentation label maps generated by are not accurate enough without considering rich semantic...

The wind farm layout optimization problem (WFLOP) aims to maximize energy utilization efficiency and mitigate losses caused by wake effects optimizing the spatial of turbines. Although Genetic Algorithms (GAs) Particle Swarm Optimization (PSO) have been widely used in WFLOP due their discrete characteristics, they still limitations global exploration capability depth. Meanwhile, Differential Evolution algorithm (DE), known for its strong ability excellent performance handling complex...

Human-robot cooperative navigation is challenging in environments with incomplete information. We introduce CoNav-Maze, a simulated robotics environment where robot navigates using local perception while human operator provides guidance based on an inaccurate map. The can share its camera views to improve the operator's understanding of environment. To enable efficient human-robot cooperation, we propose Information Gain Monte Carlo Tree Search (IG-MCTS), online planning algorithm that...

In recent years, convolutional neural networks (CNNs) have dominated the field of computer vision. Compared to traditional methods, these network algorithms exhibit strong biomimetic performance advantages in complex visual tasks due their brain-like structure. However, because some necessary characteristics are ignored, differ greatly from computational mechanisms brain. This paper starts with extracting basic features such as motion direction information brain and abstracts, generalizes...

Complex single-objective bounded problems are often difficult to solve. In evolutionary computation methods, since the proposal of differential evolution algorithm in 1997, it has been widely studied and developed due its simplicity efficiency. These developments include various adaptive strategies, operator improvements, introduction other search methods. After 2014, research based on LSHADE also by researchers. However, although recently proposed improvement strategies have shown...

Automated segmentation in medical image analysis is a challenging task that requires large amount of manually labeled data. However, most existing learning-based approaches usually suffer from limited annotated data, which poses major practical problem for accurate and robust segmentation. In addition, semi-supervised are not compared with the supervised counterparts, also lack explicit modeling geometric structure semantic information, both limit accuracy. this work, we present SimCVD,...

Space-time video super-resolution (STVSR) aims to construct a high space-time resolution sequence from the corresponding low-frame-rate, low-resolution sequence. Inspired by recent success consider spatial-temporal information for super-resolution, our main goal in this work is take full considerations of spatial and temporal correlations within sequences fast dynamic events. To end, we propose novel one-stage memory enhanced graph attention network (MEGAN) super-resolution. Specifically,...

The wind farm layout optimization problem (WFLOP) aims to maximize energy utilization efficiency under different conditions by optimizing the spatial of turbines fully mitigate losses caused wake effects. Some high-performance continuous methods, such as differential evolution (DE) variants, exhibit limited performance when directly applied due WFLOP’s discrete nature. Therefore, metaheuristic algorithms with inherent characteristics like genetic (GAs) and particle swarm (PSO) have been...

Contrastive learning (CL) aims to learn useful representation without relying on expert annotations in the context of medical image segmentation. Existing approaches mainly contrast a single positive vector (i.e., an augmentation same image) against set negatives within entire remainder batch by simply mapping all input features into constant vector. Despite impressive empirical performance, those methods have following shortcomings: (1) it remains formidable challenge prevent collapsing...

Turn-Milling Combined NC machine tool is different from traditional tools in structure and process realization. As an important means the design stage, analysis method of geometric accuracy error also method. The actual errors compensation values are a pair "symmetry" data sets which connected by movement tools. authors try to make them more consistent through this work. terms were firstly determined topological analysis, then based on homogeneous coordinate transformation multibody system...

A desirable property of autonomous agents is the ability to both solve long-horizon problems and generalize unseen tasks. Recent advances in data-driven skill learning have shown that extracting behavioral priors from offline data can enable challenging tasks with reinforcement learning. However, generalization during prior training remains an outstanding challenge. To this end, we present Few-shot Imitation Skill Transition Models (FIST), algorithm extracts skills utilizes them given a few...

Mutual Information between agent Actions and environment States (MIAS) quantifies the influence of on its environment. Recently, it was found that maximization MIAS can be used as an intrinsic motivation for artificial agents. In literature, term empowerment is to represent maximum at a certain state. While has been shown solve broad range reinforcement learning problems, calculation in arbitrary dynamics challenging problem because relies estimation mutual information. Existing approaches,...

Recent advances in unsupervised representation learning significantly improved the sample efficiency of training Reinforcement Learning policies simulated environments. However, similar gains have not yet been seen for real-robot reinforcement learning. In this work, we focus on enabling data-efficient from pixels. We present Contrastive Pre-training and Data Augmentation Efficient Robotic (CoDER), a method that utilizes data augmentation to achieve sample-efficient arm sparse rewards. While...

In reinforcement learning, conducting task composition by forming cohesive, executable sequences from multiple tasks remains challenging. However, the ability to (de)compose is a linchpin in developing robotic systems capable of learning complex behaviors. Yet, compositional beset with difficulties, including high dimensionality problem space, scarcity rewards, and absence system robustness after composition. To surmount these challenges, we view through prism category theory -- mathematical...

Networked robotic systems balance compute, power, and latency constraints in applications such as self-driving vehicles, drone swarms, teleoperated surgery. A core problem this domain is deciding when to offload a computationally expensive task the cloud, remote server, at cost of communication latency. Task offloading algorithms often rely on precise knowledge system-specific performance metrics, sensor data rates, network bandwidth, machine learning model While these metrics can be modeled...