首页|New Machine Learning Findings Reported from Tianjin University (Exploring the Learning Difficulty of Data: Theory and Measure)
New Machine Learning Findings Reported from Tianjin University (Exploring the Learning Difficulty of Data: Theory and Measure)
扫码查看
点击上方二维码区域，可以放大扫码查看
原文链接
NETL
NSTL
By a News Reporter-Staff News Editor at Robotics & Machine Learning DailyNews Daily News - Investigators publish new report on Ma chine Learning. According to news reportingoriginating in Tianjin, People’s Rep ublic of China, by NewsRx journalists, research stated, “Easy/hardsample’ is a popular parlance in machine learning. Learning difficulty of samples refers to h ow easy/harda sample is during a learning procedure.”Funders for this research include National Natural Science Foundation of China ( NSFC), TJF.The news reporters obtained a quote from the research from Tianjin University, “ An increasing needof measuring learning difficulty demonstrates its importance in machine learning (e.g., difficulty-basedweighting learning strategies). Prev ious literature has proposed a number of learning difficulty measures.However, no comprehensive investigation for learning difficulty is available to date, res ulting in that nearlyall existing measures are heuristically defined without a rigorous theoretical foundation. This study attemptsto conduct a pilot theoreti cal study for learning difficulty of samples. First, influential factors forlea rning difficulty are summarized. Under various situations conducted by summarize d influential factors,correlations between learning difficulty and two vital cr iteria of machine learning, namely, generalizationerror and model complexity, a re revealed. Second, a theoretical definition of learning difficulty is proposedon the basis of these two criteria. A practical measure of learning difficulty is proposed under thedirection of the theoretical definition by importing the b ias-variance trade-off theory. Subsequently, therationality of theoretical defi nition and the practical measure are demonstrated, respectively, by analysisof several classical weighting methods and abundant experiments realized under all situations conductedby summarized influential factors. The mentioned weighting methods can be reasonably explained underthe proposed theoretical definition an dconcerned propositions.”
TianjinPeople’s Republic of ChinaAsiaCyborgsEmerging TechnologiesMachine LearningTianjin University
NETL
NSTL
