 |
eVision Python API
|
|
eVision Python API
|
The type of learning rate indicates how the learning rate changes during training. More...
Static Public Attributes | |
| int | Standard = 0 |
| The learning rate is reduced after each training iterations where the training error has not significantly changed. | |
| int | Fixed = 1 |
| The learning rate value remains fixed with the value given as parameter during the training. | |
| int | Custom = 2 |
| The learning rate is set for each iterations at the corresponding value in the vector of parameters. | |
| int | Decay = 3 |
| The learning rate is reduced by a factor after each iterations until a minimum. | |
| int | WarmRestart = 4 |
| The learning rate goes from maxLr to minLr during the n-th period of periodN iterations. | |
The type of learning rate indicates how the learning rate changes during training.
|
static |
The learning rate is set for each iterations at the corresponding value in the vector of parameters.
Parameters: [lr0, lr1, ..., lrN]. The last learning rate in the parameters will be used when the number of training iterations is bigger than the vector of parameters.
|
static |
The learning rate is reduced by a factor after each iterations until a minimum.
Parameters: [Start learning rate, Minimum learning rate, Decay factor]
|
static |
The learning rate value remains fixed with the value given as parameter during the training.
Parameters= [lr]
|
static |
The learning rate is reduced after each training iterations where the training error has not significantly changed.
Parameters: None
|
static |
The learning rate goes from maxLr to minLr during the n-th period of periodN iterations.
The length of the first period is given by the parameter period0 and is multiplied by the factor f after each period. Thus, periodN = f^n * period0.
Parameters: [minMr, maxLr, period0, f]
1.17.0