Angel是一个基于参数服务器(Parameter Server)理念开发的高性能分布式机器学习平台,它基于腾讯内部的海量数据进行了反复的调优,并具有广泛的适用性和稳定性,模型维度越高,优势越明显。 Angel由腾讯和北京大学联合开发,兼顾了工业界的高可用性和学术界的创新性。
Angel的核心设计理念围绕模型。它将高维度的大模型合理切分到多个参数服务器节点,并通过高效的模型更新接口和运算函数,以及灵活的同步协议,轻松实现各种高效的机器学习算法。
Angel基于Java和Scala开发,能在社区的Yarn上直接调度运行,并基于PS Service,支持Spark on Angel,集成了部分图计算和深度学习算法。
欢迎对机器学习有兴趣的同仁一起贡献代码,提交Issues或者Pull Requests。请先查阅: Angel Contribution Guide
- Angel
- Traditional Machine Learning Methods
- Deep Learning Methods
- Spark on Angel
- Online Learning
- Offline Learning
- QQ群:20171688
- Lele Yu, Bin Cui, Ce Zhang, Yingxia Shao. LDA*: A Robust and Large-scale Topic Modeling System. VLDB, 2017
- Jiawei Jiang, Bin Cui, Ce Zhang, Lele Yu. Heterogeneity-aware Distributed Parameter Servers. SIGMOD, 2017
- Jie Jiang, Lele Yu, Jiawei Jiang, Yuhong Liu and Bin Cui. Angel: a new large-scale machine learning system. National Science Review (NSR), 2017
- Jie Jiang, Jiawei Jiang, Bin Cui and Ce Zhang. TencentBoost: A Gradient Boosting Tree System with Parameter Server. ICDE, 2017
-
Angel: A Machine Learning Framework for High Dimensionality. Strata China, 2017
-
方圆并济:基于 Spark on Angel 的高性能机器学习. QCon ShangHai China, 2017
-
基于Angel和Spark Streaming的高维度Online Learning. GIAC China, 2017