Parallel support vector method and apparatus
View Patent ↗Disclosed is an improved technique for training a support vector machine using a distributed architecture. A training data set is divided into subsets, and the subsets are optimized in a first level of optimizations, with each optimization generating a support vector set. The support vector sets output from the first level optimizations are then combined and used as input to a second level of optimizations. This hierarchical processing continues for multiple levels, with the output of each prior level being fed into the next level of optimizations. In order to guarantee a global optimal solution, a final set of support vectors from a final level of optimization processing may be fed back into the first level of the optimization cascade so that the results may be processed along with each of the training data subsets. This feedback may continue in multiple iterations until the same final support vector set is generated during two sequential iterations through the cascade, thereby guaranteeing that the solution has converged to the global optimal solution. In various embodiments, various combinations of inputs may be used by the various optimizations. The individual optimizations may be processed in parallel.
1. A method for training a Cascade Support Vector Machine for use as a classifier, the method comprising the steps of:
a. encoding training data into a set of data vectors;
b. dividing the data set into a plurality of data subsets;
c. training a Cascade Support Vector Machine as a classifier for classifying data into different classes by:
i. performing a plurality of optimizations on a plurality of Support Vector Machines (SVMs) using the plurality of data subsets as inputs for the plurality of SVMs to obtain for each of the SVMs a set of support vectors;
ii. performing one or more additional optimizations on the plurality of SVMs using as inputs the sets of support vectors obtained by the previous optimizations; and
iii. repeating step ii until a single subset of the set of support vectors is obtained.
2. The method of claim 1 further comprising the step of performing further optimizations using the single subset of the set of support vectors and the plurality of data subsets as inputs into the plurality of SVMs, to test for convergence.
3. The method of claim 1 wherein the inputs for steps ii and iii further include at least a portion of the plurality of data subsets.
4. The method of claim 1 wherein the optimizations in at least one of steps i, ii and iii are performed in parallel.
5. The method of claim 1 wherein the optimizations in at least one of steps i, ii and iii are performed serially.
6. A computer readable medium comprising computer program instructions for training a Cascade Support Vector Machine for use as a classifier, the computer program instructions, when executed by a processor, defining the steps of:
a. encoding training data into a set of data vectors;
b. dividing the data set into a plurality of data subsets;
c. training a Cascade Support Vector Machine as a classifier for classifying data into different classes by:
i. performing a plurality of optimizations on a plurality of Support Vector Machines (SVMs) using the plurality of data subsets as inputs for the plurality of SVMs to obtain for each of the SVMs a set of support vectors;
ii. performing one or more additional optimizations on the plurality of SVMs using as inputs the sets of support vectors obtained by the previous optimizations; and
iii. repeating step ii until a single subset of the set of support vectors is obtained.
7. The computer readable medium of claim 6 further comprising the step of performing further optimizations using the single subset of the set of support vectors and the plurality of data subsets as inputs into the plurality of SVMs, to test for convergence.
8. The computer readable medium of claim 6 wherein the inputs for steps ii and iii further include at least a portion of the plurality of data subsets.
9. The computer readable medium of claim 6 wherein the optimizations in at least one of steps i, ii and iii are performed in parallel.
10. The computer readable medium of claim 6 wherein the optimizations in at least one of steps i, ii and iii are performed serially.