The dynamics of big data reflects in the continuous update. Bounded incremental calculation thus needs to meet the new requirement. That is to say, when the data changes, there is no need to restart the calculation. The last calculation result and the data updates produce the new calculation result. Because the calculation result and data updates are usually much smaller than the original data, the improvement of computational efficiency should be significant. At present, the incremental programs specially designed for a specific problem face high barriers to entry. We are working on an effective and universal incremental method, using program language, compilers, and algorithm skills to build an incremental program.

Focusing on big data incremental computing models and algorithms:
1) Characterize the effectiveness of an incremental algorithm and whether a general method exists to allow such characterization;
2) Use the boundedness of the incremental algorithm, to characterize the cost of incremental calculation through the variation between input and output flow;
3) Develop a general incremental method based on the boundedness of different incremental algorithms.