utils.py 文件源码

def CVsplitter(taskcolumn, K):
    '''
    Divide tasks into roughly equal K sets, and do CV over such K sets.
    '''

    tasks = sorted(list(set(taskcolumn)))
    tasks_splitted = [[] for _ in range(K)]
    for (ind, task) in enumerate(tasks):
        tasks_splitted[ind % K].append(task)

    for fold in range(K):
        print 'fold:', fold, 'testtasks:', tasks_splitted[fold]
        test = np.logical_or.reduce([taskcolumn == taskid for taskid in
                                    tasks_splitted[fold]])

        yield (np.logical_not(test), test)