Traditional Direction of Arrival (DOA) estimation algorithms for coherent signals in uniform circular array (UCA), such as mode space transformation, allow for the adaptation of advanced algorithms initially developed...
详细信息
Modeling stochastic multi-ship trajectories is vital for maritime safety and interaction efficiency. Recent researches show that diffusion models excel in trajectory prediction, surpassing GANs and VAEs in generation ...
详细信息
Continual learning algorithms aim to learn from a sequence of tasks, making the training distribution non-stationary. The majority of existing continual learning approaches in the literature rely on heuristics and do ...
详细信息
The intricacies and instability of introducing cryogenic propellants into the combustion system have piqued the curiosity of scientists studying the procedure. The latest innovation is utilizing data-driven machine le...
详细信息
Graph Convolutional Networks (GCNs) are powerful learning approaches for graph-structured data. GCNs are both computing- and memory-intensive. The emerging 3D-stacked computation-in-memory (CIM) architecture provides ...
详细信息
ISBN:
(纸本)9798350323481
Graph Convolutional Networks (GCNs) are powerful learning approaches for graph-structured data. GCNs are both computing- and memory-intensive. The emerging 3D-stacked computation-in-memory (CIM) architecture provides a promising solution to process GCNs efficiently. The CIM architecture can provide near-data computing, thereby reducing data movement between computing logic and memory. However, previous works do not fully exploit the CIM architecture in both dataflow and mapping, leading to significant energy *** paper presents Lift, an energy-efficient GCN accelerator based on 3D CIM architecture using software and hardware co-design. At the hardware level, Lift introduces a hybrid architecture to process vertices with different characteristics. Lift adopts near-bank processing units with a push-based dataflow to process vertices with strong re-usability. A dedicated unit is introduced to reduce massive data movement caused by high-degree vertices. At the software level, Lift adopts a hybrid mapping to further exploit data locality and fully utilize the hybrid computing resources. The experimental results show that the proposed scheme can significantly reduce data movement and energy consumption compared with representative schemes.
Federated Learning (FL) has emerged as a promising training framework that enables a server to effectively train a global model by coordinating multiple devices, i.e., clients, without sharing their raw data. Keeping ...
详细信息
Objective: This study aims to explore the application of Chain of Thought (CoT) reasoning in automating ICD coding, specifically focusing on lymphoma cases. By leveraging large language models (LLMs) and CoT...
详细信息
Dual incomplete multi-view multi-label learning (DIMVMLL) aims to make accurate predictions for multiple labels by exploiting incomplete multi-view information and incomplete multi-label information from input samples...
详细信息
ISBN:
(纸本)9798400712203
Dual incomplete multi-view multi-label learning (DIMVMLL) aims to make accurate predictions for multiple labels by exploiting incomplete multi-view information and incomplete multi-label information from input samples. Since the view information and label information of samples are incomplete in this task, existing DIMVMLL methods classify samples based on the semantic relationships between view features and labels. However, in the DIMVMLL task, exploring the consistent and complementary information between views is important for the classification performance of model. To learn high-quality view consistency information and complementary information, we propose random masked autoencoders framework (RMAF). First, we design a triple reconstruction loss for the dual-channel autoencoders to guide the model in extracting consistent and complementary information between views and introduce contrastive learning to guide the model in learning private information for each view. Second, we propose a feature-level random masking strategy to enhance high-level feature extract ability of the autoencoders. Furthermore, to preserve the feature semantic structure, we introduce structure loss. Extensive experiments have validated the effectiveness of our model.
Reinforcement Learning (RL) is a machine learning approach in which an agent learns to make decisions in an environment to maximize a cumulative reward. When combined with NeuroEvolution of Augmenting Topologies (NEAT...
详细信息
Reinforcement Learning (RL) is a machine learning approach in which an agent learns to make decisions in an environment to maximize a cumulative reward. When combined with NeuroEvolution of Augmenting Topologies (NEAT), RL offers several advantages. NEAT is a genetic algorithm that optimizes the development of artificial neural networks by modifying both their structure and weights. When integrated with RL, NEAT can improve the learning process by merging evolutionary optimization with RL techniques. NEAT has demonstrated significant potential in evolving neural networks for RL tasks. However, traditional centralized training methods encounter scalability and data privacy issues. This paper investigates the integration of NEAT with Federated Learning (FL) and its enhancement with Markov Chains and Gaussian Processes to address certain issues. We propose a new framework that combines NEAT for neural network evolution with TensorFlow Federated (TFF) for decentralized training across multiple clients. Our approach is assessed using the BipedalWalker-v3 environment from OpenAI Gym. The experimental results show that our federated NEAT framework, augmented with Markov Chains and Gaussian Processes, achieves competitive performance while maintaining data privacy and reducing computational overhead on central servers. Additionally, we implement parallelization techniques using concurrent futures to enhance the efficiency of NEAT generations.
The rapid advancements in time-series forecasting have led to the emergence of datasets with increasingly diverse characteristics. Researchers typically focus on designing robust algorithms to handle these datasets. H...
详细信息
暂无评论