SGD Framework is basically two parts: gradient calculation and weight update. Gradient calculation load data from DDR and calculate gradient of preset weight and is a dataflow function. Weight update calculate next weight based on gradient and chosen method and is executed after gradient calculation. Block diagram is shown as below.
The correctness of Linear Regression/LASSO Regression/Ridge Regression Training using SGD framework is verified by comparing results with Spark mllib. The results are identical.