SGD is a family of algorithms that reuse concepts from Support Vector Machines and Logistic Regression.
SGD uses an optimized method to minimize the cost (or loss) function, making it particularly suitable for large datasets (or datasets with large number of features).