C++ API Reference for Intel® Data Analytics Acceleration Library 2020 Update 1
Parameter base class for the Adaptive gradient descent algorithm More...
Parameter | ( | const sum_of_functions::BatchPtr & | function_, |
size_t | nIterations_ = 100 , |
||
double | accuracyThreshold_ = 1.0e-05 , |
||
data_management::NumericTablePtr | batchIndices_ = data_management::NumericTablePtr() , |
||
const size_t | batchSize_ = 128 , |
||
data_management::NumericTablePtr | learningRate_ = data_management::HomogenNumericTable< double >::create(1, 1, data_management::NumericTableIface::doAllocate, 0.01) , |
||
double | degenerateCasesThreshold_ = 1.0e-08 , |
||
size_t | seed_ = 777 |
||
) |
Constructs the parameter base class of the Adaptive gradient descent algorithm
[in] | function_ | Objective function represented as sum of functions |
[in] | nIterations_ | Maximal number of iterations of the algorithm |
[in] | accuracyThreshold_ | Accuracy of the algorithm. The algorithm terminates when this accuracy is achieved |
[in] | batchIndices_ | Numeric table that represents 32 bit integer indices of terms in the objective function. If no indices are provided, the implementation will generate random indices. |
[in] | batchSize_ | Number of batch indices to compute the stochastic gradient. If batchSize is equal to the number of terms in objective function then no random sampling is performed, and all terms are used to calculate the gradient. This parameter is ignored if batchIndices is provided. |
[in] | learningRate_ | Numeric table that contains value of the learning rate |
[in] | degenerateCasesThreshold_ | Value needed to avoid degenerate cases in square root computing. |
[in] | seed_ | Seed for random generation of 32 bit integer indices of terms in the objective function. |
|
virtual |
Checks the correctness of the parameter
data_management::NumericTablePtr batchIndices |
Numeric table that represents 32 bit integer indices of terms in the objective function. If no indices are provided, the implementation will generate random indices.
double degenerateCasesThreshold |
Value needed to avoid degenerate cases in square root computing.
engines::EnginePtr engine |
Engine for random generation of 32 bit integer indices of terms in the objective function.
data_management::NumericTablePtr learningRate |
Numeric table that contains value of the learning rate
size_t seed |
Seed for random generation of 32 bit integer indices of terms in the objective function.
For more complete information about compiler optimizations, see our Optimization Notice.