首页|Patent Issued for Systems and methods for providing a nearest neighbors classifi cation pipeline with automated dimensionality reduction (USPTO 11934384)
Patent Issued for Systems and methods for providing a nearest neighbors classifi cation pipeline with automated dimensionality reduction (USPTO 11934384)
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
News editors obtained the following quote from the background information suppli ed by the inventors:“Nearest neighbors algorithms, such as the K-nearest neighb ors classification algorithm, classify newobservations by finding K-many closes t training observations to the new observation and use the majorityclass vote f rom this subset of training observations as the class for the new observation. T hese typesof algorithms are commonly used in machine learning solutions. Since the nearest neighbors algorithmsseek to find the closest neighbors for classifi cation, distance metrics are used to determine the nearestneighbors. However, w hen the dimensionality of the training dataset is large, such as hundreds of inp utfeatures, calculating distance metrics is not only computationally very expen sive, but also gives rise towhat is referred to as the “curse of dimensionality ,” which is a phenomenon in which the nearest neighborclassifications can resul t in lower accuracy classifications. Accordingly, to attempt to avoid or reducethese problems, it is often useful to reduce the dimensionality of high-dimensio nal datasets before applyingcertain distance-based machine learning algorithms to fit the data. However, it is very difficult for a data scientist to know how many dimensions to reduce the data by and still achieve meaningful classificatio nresults.
AlgorithmsBusinessCapital One Servic es LLCCyborgsDimensionality ReductionEmerging TechnologiesK-nearest Neig hborMachine Learning
NETL
NSTL
