The MSE (Mean squared error) gain, is a measure of set purity based on the variance of response values present in the node.
More...
#include <mse_gain.hpp>
|
| std::tuple< double, double > | BinaryGains () |
| | Calculates the mean squared error gain for the left and right children for the current index. More...
|
| |
| template<bool UseWeights, typename ResponsesType , typename WeightVecType > |
| void | BinaryScanInitialize (const ResponsesType &responses, const WeightVecType &weights, const size_t minimum) |
| | Caches the prefix sum of squares to efficiently compute gain value for each split. More...
|
| |
| template<bool UseWeights, typename ResponsesType , typename WeightVecType > |
| void | BinaryStep (const ResponsesType &responses, const WeightVecType &weights, const size_t index) |
| | Updates the statistics for the given index. More...
|
| |
|
| template<bool UseWeights, typename VecType , typename WeightVecType > |
| static double | Evaluate (const VecType &values, const WeightVecType &weights, const size_t begin, const size_t end) |
| | Evaluate the mean squared error gain of values from begin to end index. More...
|
| |
| template<bool UseWeights, typename VecType , typename WeightVecType > |
| static double | Evaluate (const VecType &values, const WeightVecType &weights) |
| | Evaluate the MSE gain on the complete vector. More...
|
| |
The MSE (Mean squared error) gain, is a measure of set purity based on the variance of response values present in the node.
This is same thing as negation of variance of dependent variable in the node as we will try to maximize this quantity to maximize gain (and thus reduce variance of a set).
◆ BinaryGains()
| std::tuple<double, double> mlpack::tree::MSEGain::BinaryGains |
( |
| ) |
|
|
inline |
Calculates the mean squared error gain for the left and right children for the current index.
X = array of values of size n.
\begin{eqnarray*} MSE = \sum\limits_{i=1}^n {X_i}^2 - {\dfrac{\sum\limits_{j=1}^n X_j}{n}}^2 \end{eqnarray*}
◆ BinaryScanInitialize()
template<bool UseWeights, typename ResponsesType , typename WeightVecType >
| void mlpack::tree::MSEGain::BinaryScanInitialize |
( |
const ResponsesType & |
responses, |
|
|
const WeightVecType & |
weights, |
|
|
const size_t |
minimum |
|
) |
| |
|
inline |
Caches the prefix sum of squares to efficiently compute gain value for each split.
It also computes the initial mean for left and right child.
- Parameters
-
| responses | The set of responses on which statistics are computed. |
| weights | The set of weights associated to each response. |
| minimum | The minimum number of elements in a leaf. |
◆ BinaryStep()
template<bool UseWeights, typename ResponsesType , typename WeightVecType >
| void mlpack::tree::MSEGain::BinaryStep |
( |
const ResponsesType & |
responses, |
|
|
const WeightVecType & |
weights, |
|
|
const size_t |
index |
|
) |
| |
|
inline |
Updates the statistics for the given index.
- Parameters
-
| responses | The set of responses on which statistics are computed. |
| weights | The set of weights associated to each response. |
| index | The current index. |
◆ Evaluate() [1/2]
template<bool UseWeights, typename VecType , typename WeightVecType >
| static double mlpack::tree::MSEGain::Evaluate |
( |
const VecType & |
values, |
|
|
const WeightVecType & |
weights, |
|
|
const size_t |
begin, |
|
|
const size_t |
end |
|
) |
| |
|
inlinestatic |
Evaluate the mean squared error gain of values from begin to end index.
Note that gain can be slightly greater than 0 due to floating-point representation issues. Thus if you are checking for perfect fit, be sure to use 'gain >= 0.0' and not 'gain == 0.0'. The values vector should always be of type arma::Row<double> or arma::rowvec.
- Parameters
-
| values | Set of values to evaluate MAD gain on. |
| weights | Weights associated to each value. |
| begin | Start index. |
| end | End index. |
◆ Evaluate() [2/2]
template<bool UseWeights, typename VecType , typename WeightVecType >
| static double mlpack::tree::MSEGain::Evaluate |
( |
const VecType & |
values, |
|
|
const WeightVecType & |
weights |
|
) |
| |
|
inlinestatic |
Evaluate the MSE gain on the complete vector.
- Parameters
-
| values | Set of values to evaluate MSE gain on. |
| weights | Weights associated to each value. |
The documentation for this class was generated from the following file:
