First page Back Continue Last page Overview Graphics

A Machine Learning Approach to Data Fusion


Notes:

Here is an example table of retrieved actives number for each size range and for each coefficients. Each column represents the numbers of retrieved actives for a particular coefficient, and each row represents the size range of the retrieved actives for every individual coefficient. The best combination of coefficients for data fusion is the one that retrieves the highest number of actives. Thus, we can find the coefficients with the highest number of retrieved actives from each size range and group these coefficients to form the best combination of data fusion. So far, our experiments show that the size distributions of each individual coefficient are not exactly same but very similar when two queries have the same size. The machine learning approach will learn the best coefficient for each size range and for each query size, and a program will allow user to enter a query then output the best combination for data fusion.