文件名称：MNIST_2019_09_28.py
文件大小：8KB
文件格式：PY
更新时间：2022-10-26 04:41:00
CNN
tremendous strides have been made in machine learning, one of the remaining open challenges is to achieve real-time speed as well as maintain high performance, Deep learning is a complex machine learning algorithm, which achieves much better results in speech and image recognition than previous related technologies. The Convolutional Network architecture is a remarkably versatile, yet conceptually simple paradigm that can be applied to a wide spectrum of perceptual tasks. In the designed CNN, the first convolutional layers has 6 feature image, the second convolution layer has 12 feature. Next is the full connection layer, the first full connection has 1024 neurons, and the second layer has 784 neurons. And the output layer is 10 neurons, which is the 10 digit labels. And Tensorflow is adopted to design setup network, When the training of network is completed, 500 epochs, the accuracy of the recognition is 95.6% for training set, and is 98.1% for test set.