Conditional Random Sampling (CRS), a technique particularly suitable for sparse data. In large-scale applications, the data are often highly sparse. CRS combines sketching and sampling in that it converts sketches of the data into conditional random samples online in the estimation stage, with the sample size determined retrospectively.Conditional Random Sampling (CRS) is a sketch-based sampling technique that effectively exploits data sparsity. In modern applications in learning, data mining, and information retrieval, the datasets are often very large and also highly sparse. The following codes were written in MATLAB in order to simulate the CRS algorithm,as part of the honours maths course (SHARP) taken during my undergraduate study in SUTD.