- Speech and dialogue systems
- Recommender Systems and Techniques
- Information Retrieval and Search Behavior
- Semantic Web and Ontologies
- Natural Language Processing Techniques
- Smart Agriculture and AI
- Identification and Quantification in Food
- Neural Networks and Applications
- Plant Virus Research Studies
- Topic Modeling
Gansu Agricultural University
2022
The target recognition algorithm is one of the core technologies Zanthoxylum pepper-picking robots. However, most existing detection algorithms cannot effectively detect fruit covered by branches, leaves and other fruits in natural scenes. To improve work efficiency adaptability Zanthoxylum-picking robot environments, to recognize complex environments under different lighting conditions, this paper presents a Zanthoxylum-picking-robot method based on improved YOLOv5s. Firstly, an CBF module...
Backpropagation is the foundational algorithm for training neural networks and a key driver of deep learning's success. However, its biological plausibility has been challenged due to three primary limitations: weight symmetry, reliance on global error signals, dual-phase nature training, as highlighted by existing literature. Although various alternative learning approaches have proposed address these issues, most either fail satisfy all criteria simultaneously or yield suboptimal results....
Retrieval-augmented generation (RAG) techniques have proven to be effective in integrating up-to-date information, mitigating hallucinations, and enhancing response quality, particularly specialized domains. While many RAG approaches been proposed enhance large language models through query-dependent retrievals, these still suffer from their complex implementation prolonged times. Typically, a workflow involves multiple processing steps, each of which can executed various ways. Here, we...
The capacity of large language models (LLMs) to generate honest, harmless, and helpful responses heavily relies on the quality user prompts. However, these prompts often tend be brief vague, thereby significantly limiting full potential LLMs. Moreover, harmful can meticulously crafted manipulated by adversaries jailbreak LLMs, inducing them produce potentially toxic content. To enhance capabilities LLMs while maintaining strong robustness against inputs, this study proposes a transferable...