tf.reduce_sum () is a summation function
tf.train.GradientDescentOptimizer () using stochastic gradient descent algorithm
tf.global_variables_initializer () initialization
tf.cast () execution tensorflow in tensor data type conversion, such as read into the picture if it is int8 type, usually in the format you want to convert the image data is float32 before training.
tf.enable_eager_execution()
function takes a negative value tf.negative