Search

Large Language Models (LLMs), with their remarkable task-handling capabilities and innovative outputs, have catalyzed significant advancements across a spectrum of fields. However, their proficiency within specialized domains such as biomolecular studies remains limited. To address this challenge, we introduce Mol-Instructions, a meticulously curated, comprehensive instruction dataset expressly designed for the biomolecular realm. Mol-Instructions is composed of three pivotal components: molecule-oriented instructions, protein-oriented instructions, and biomolecular text instructions, each curated to enhance the understanding and prediction capabilities of LLMs concerning biomolecular features and behaviors. Through extensive instruction tuning experiments on the representative LLM, we underscore the potency of Mol-Instructions to enhance the adaptability and cognitive acuity of large models within the complex sphere of biomolecular studies, thereby promoting advancements in the biomolecular research community. Mol-Instructions is made publicly accessible for future research endeavors and will be subjected to continual updates for enhanced applicability., Comment: Project homepage: https://github.com/zjunlp/Mol-Instructions

In this paper, a new method using level set is proposed for abnormal event detection and localization in video surveillance. From an input video, five image descriptors, namely the color moments, the edge histogram descriptors, the color and edge directivity descriptors, the color layout descriptors, and the scalable color descriptors, are extracted for robust detection. We employ the local binary fitting model as the statistical learning model to update by the input videos in real time. We performed experiments on the publicly available UCSD anomaly detection dataset and showed that our method has good performance for detecting and localizing abnormality compared to the state-of-the-art methods.

Support vector machines (SVM) is an excellent method of machine learning. It can be reduced to solving large-scale quadratic programming problem. The Sequential Minimal Optimisation (SMO) algorithm is the best training algorithm for SVM, it optimizes a minimal subset of just two points at each iteration. This paper proposes an improved training algorithm on the premise of the optimization problem admits an analytical solution. The improved algorithm optimizes three points at each iteration. The modified algorithm performs significantly faster than the original SMO on the datasets tried.