隨機投影 (Random Projection) 方法,又被稱為 JL-embeddings,是近年發展成功的一項重要方法,尤其在處理大數據問題時,提供了一項很好的選擇,本文第一部份整理並敘述了這方面的成果,包含 Sparse Random Projection 之發展。隨機投影之用途之一是用以估計原資料矩陣之 singular values,本文對生成變數為N(0,1)時該估計式之分佈進行了探討,但沒有獲得可用之成果。模擬則顯示該估計式有高估現象,應非不偏估計式。 The Random Projection method, also known as JL-embeddings, is an important method for successful development in recent years, especially when dealing with big data problems. This is a good choice. Describes the results in this area, including the development of the Sparse Random Projection.One of the purposes of random projection is to estimate the singular values of the original data matrix. In this paper, the distribution of the estimator when the generator is N(0,1) is discussed, but no available results are obtained. The simulation shows that the estimate has an overestimation and should not be biased.