In Apache Spark, the sparse vector is a vector that has two parallel arrays, one for indices, and one for values. This is used for storing non-zero entities to save space.