Some optimizer-related flags need more accurate descriptions

[nobutoba]

Prerequisites

• [x] I checked to make sure that this issue has not been filed already.

1. The entire URL of the documentation with the issue

https://github.com/tensorflow/models/blob/master/research/slim/train_image_classifier.py

2. Describe the issue

The help message of train_image_classifier.py, as printed e.g. by python train_image_classifier.py --help, can be improved so that the following descriptions of optimizer-related flags become more accurate and easier to understand:

flag name	current description	dependency	new description candidate
momentum	"The momentum for the MomentumOptimizer and RMSPropOptimizer."	Has no effect when using RMSPropOptimizer.	"The momentum for MomentumOptimizer."
rmsprop_momentum	"Momentum."	Has effect only when using RMSPropOptimizer.	"The momentum for RMSPropOptimizer."
opt_epsilon	"Epsilon term for the optimizer."	Has effect only when using Adadelta, Adam, RMSProp.	"Epsilon term for AdadeltaOptimizer, AdamOptimizer, and RMSPropOptimizer."
adam_beta1	"The exponential decay rate for the 1st moment estimates."	Has effect only when using AdamOptimizer.	"The exponential decay rate for the 1st moment estimates. Relevant only for AdamOptimizer."
adam_beta2	"The exponential decay rate for the 2nd moment estimates."	Has effect only when using AdamOptimizer.	"The exponential decay rate for the 2nd moment estimates. Relevant only for AdamOptimizer."
ftrl_learning_rate_power	"The learning rate power."	Has effect only when using FtrlOptimizer.	"The FTRL learning rate power."
learning_rate_decay_factor	"Learning rate decay factor."	Has effect only when using exponential decay.	"Learning rate decay factor used by a exponential decay learning rate."

[marksandler2]

Thanks, can you send a pull request?