In our experiments, the network was constructed by the one-shot model which consists of 100×10feature nodes and 1×9000 enhancement nodes. Compared with M the deep and complex structure of DBN and HELM, which is 4000-4000-4000 and 3000-3000-15000, respectively, the proposed BLS presents faster training time.The testing results present a pleasant performance, especially the training time of the proposed broad learning. Similarly to the MNIST cases, although the accuracy is not the best one, the performance matches with previous work with a testing time of 6.0299s in the server. Considering the superfast speed in computation, which is the best among the existing methods, the proposed broad learning and network is very attractive.