trueskill/PHPSkills/Numerics/Matrix.php
2010-08-31 22:24:05 -04:00

365 lines
10 KiB
PHP

<?php
namespace Moserware\Numerics;
class Matrix
{
const ERROR_TOLERANCE = 0.0000000001;
private $_matrixRowData;
private $_rowCount;
private $_columnCount;
public function __construct($rows = 0, $columns = 0, $matrixData = null)
{
$this->_rowCount = $rows;
$this->_columnCount = $columns;
$this->_matrixRowData = $matrixData;
}
public function getRowCount()
{
return $this->_rowCount;
}
public function getColumnCount()
{
return $this->_columnCount;
}
public function getValue($row, $col)
{
return $this->_matrixRowData[$row][$col];
}
public function setValue($row, $col, $value)
{
$this->_matrixRowData[$row][$col] = $value;
}
public function getTranspose()
{
// Just flip everything
$transposeMatrix = array();
for ($currentRowTransposeMatrix = 0;
$currentRowTransposeMatrix < $this->_columnCount;
$currentRowTransposeMatrix++)
{
$transposeMatrixCurrentRowColumnValues = array();
$transposeMatrix[] = $transposeMatrixCurrentRowColumnValues;
for ($currentColumnTransposeMatrix = 0;
$currentColumnTransposeMatrix < $this->_rowCount;
$currentColumnTransposeMatrix++)
{
$transposeMatrix[$currentRowTransposeMatrix][$currentColumnTransposeMatrix] =
$this->_matrixRowData[$currentColumnTransposeMatrix][$currentRowTransposeMatrix];
}
}
return new Matrix($transposeMatrix);
}
private function isSquare()
{
return ($this->_rowCount == $this->_columnCount) && ($this->_rowCount > 0);
}
public function getDeterminant()
{
// Basic argument checking
if (!$this->isSquare())
{
throw new Exception("Matrix must be square!");
}
if ($this->_rowCount == 1)
{
// Really happy path :)
return $this->_matrixRowValues[0][0];
}
if ($this->_rowCount == 2)
{
// Happy path!
// Given:
// | a b |
// | c d |
// The determinant is ad - bc
$a = $this->_matrixRowData[0][0];
$b = $this->_matrixRowData[0][1];
$c = $this->_matrixRowData[1][0];
$d = $this->_matrixRowData[1][1];
return $a*$d - $b*$c;
}
// I use the Laplace expansion here since it's straightforward to implement.
// It's O(n^2) and my implementation is especially poor performing, but the
// core idea is there. Perhaps I should replace it with a better algorithm
// later.
// See http://en.wikipedia.org/wiki/Laplace_expansion for details
$result = 0.0;
// I expand along the first row
for ($currentColumn = 0; $currentColumn < $this->_columnCount; $currentColumn++)
{
$firstRowColValue = $this->_matrixRowValues[0][$currentColumn];
$cofactor = $this->getCofactor(0, $currentColumn);
$itemToAdd = $firstRowColValue*$cofactor;
$result = $result + $itemToAdd;
}
return $result;
}
public function getAdjugate()
{
if (!$this->isSquare())
{
throw new Exception("Matrix must be square!");
}
// See http://en.wikipedia.org/wiki/Adjugate_matrix
if ($this->_rowCount == 2)
{
// Happy path!
// Adjugate of:
// | a b |
// | c d |
// is
// | d -b |
// | -c a |
$a = $this->_matrixRowData[0][0];
$b = $this->_matrixRowData[0][1];
$c = $this->_matrixRowData[1][0];
$d = $this->_matrixRowData[1][1];
return new SquareMatrix( $d, -$b,
-$c, $a);
}
// The idea is that it's the transpose of the cofactors
$result = array();
for ($currentColumn = 0; $currentColumn < $this->_columns; $currentColumn++)
{
for ($currentRow = 0; $currentRow < $this->_rowCount; $currentRow++)
{
$result[$currentColumn][$currentRow] = $this->getCofactor($currentRow, $currentColumn);
}
}
return new Matrix($result);
}
public function getInverse()
{
if (($this->_rowCount == 1) && ($this->_columnCount == 1))
{
return new SquareMatrix(1.0/$this->_matrixRowData[0][0]);
}
// Take the simple approach:
// http://en.wikipedia.org/wiki/Cramer%27s_rule#Finding_inverse_matrix
$determinantInverse = 1.0 / $this->getDeterminant();
$adjugate = $this->getAdjugate();
return self::scalarMultiply($determinantInverse, $adjugate);
}
public static function scalarMultiply($scalar, $matrix)
{
$rows = $matrix->getRowCount();
$columns = $matrix->getColumnCount();
$newValues = array();
for ($currentRow = 0; $currentRow < $rows; $currentRow++)
{
for ($currentColumn = 0; $currentColumn < $columns; $currentColumn++)
{
$newValues[$currentRow][$currentColumn] = $scalarValue*$matrix->getValue($currentRow, $currentColumn);
}
}
return new Matrix($rows, $columns, $newValues);
}
public static function add($left, $right)
{
if (
($left->getRowCount() != $right->getRowCount())
||
($left->getColumnCount() != $right->getColumnCount())
)
{
throw new Exception("Matrices must be of the same size");
}
// simple addition of each item
$resultMatrix = array();
for ($currentRow = 0; $currentRow < $left->getRowCount(); $currentRow++)
{
for ($currentColumn = 0; $currentColumn < $right->getColumnCount(); $currentColumn++)
{
$resultMatrix[$currentRow][$currentColumn] =
$left->getValue($currentRow, $currentColumn)
+
$right->getValue($currentRow, $currentColumn);
}
}
return new Matrix($left->getRowCount(), $right->getColumnCount(), $resultMatrix);
}
public static function multiply($left, $right)
{
// Just your standard matrix multiplication.
// See http://en.wikipedia.org/wiki/Matrix_multiplication for details
if ($left->getColumnCount() != $right->getRowCount())
{
throw new Exception("The width of the left matrix must match the height of the right matrix");
}
$resultRows = $left->getRowCount();
$resultColumns = $right->getColumnCount();
$resultMatrix = array();
for ($currentRow = 0; $currentRow < $resultRows; $currentRow++)
{
for ($currentColumn = 0; $currentColumn < $resultColumns; $currentColumn++)
{
$productValue = 0;
for ($vectorIndex = 0; $vectorIndex < $left->getColumnCount(); $vectorIndex++)
{
$leftValue = $left->getValue($currentRow, $vectorIndex);
$rightValue = $right->getValue($vectorIndex, $currentColumn);
$vectorIndexProduct = $leftValue*$rightValue;
$productValue = $productValue + $vectorIndexProduct;
}
$resultMatrix[$currentRow][$currentColumn] = $productValue;
}
}
return new Matrix($resultRows, $resultColumns, $resultMatrix);
}
private function getMinorMatrix($rowToRemove, $columnToRemove)
{
// See http://en.wikipedia.org/wiki/Minor_(linear_algebra)
// I'm going to use a horribly naïve algorithm... because I can :)
$result = array();
$actualRow = 0;
for ($currentRow = 0; $currentRow < $this->_rowCount; $currentRow++)
{
if ($currentRow == $rowToRemove)
{
continue;
}
$actualCol = 0;
for ($currentColumn = 0; $currentColumn < $this->_columnCount; $currentColumn++)
{
if ($currentColumn == $columnToRemove)
{
continue;
}
$result[$currentRow][$currentColumn] = $this->_matrixRowData[$currentRow][$currentColumn];
$actualCol++;
}
$actualRow++;
}
return new Matrix($this->_rowCount - 1, $this->_columnCount - 1, $result);
}
private function getCofactor($rowToRemove, $columnToRemove)
{
// See http://en.wikipedia.org/wiki/Cofactor_(linear_algebra) for details
// REVIEW: should things be reversed since I'm 0 indexed?
$sum = $rowToRemove + $columnToRemove;
$isEven = ($sum%2 == 0);
if ($isEven)
{
return $this->getMinorMatrix($rowToRemove, $columnToRemove)->getDeterminant();
}
else
{
return -1.0*$this->getMinorMatrix($rowToRemove, $columnToRemove)->getDeterminant();
}
}
}
class Vector extends Matrix
{
public function __construct(array $vectorValues)
{
parent::__construct(count($vectorValues), 1, array($vectorValues));
}
}
class SquareMatrix extends Matrix
{
public function __construct()
{
$allValues = \func_get_args();
$rows = (int) sqrt(count($allValues));
$cols = $rows;
$matrixData = array();
$allValuesIndex = 0;
for ($currentRow = 0; $currentRow < $rows; $currentRow++)
{
for ($currentColumn = 0; $currentColumn < $cols; $currentColumn++)
{
$matrixData[$currentRow][$currentColumn] = $allValues[$allValuesIndex++];
}
}
parent::__construct($rows, $cols, $matrixData);
}
}
class DiagonalMatrix extends Matrix
{
public function __construct(array $diagonalValues)
{
$diagonalCount = count($diagonalValues);
$rowCount = $diagonalCount;
$colCount = $rowCount;
parent::__construct($rowCount, $colCount);
for($i = 0; $i < $diagonalCount; $i++)
{
$this->setValue($i, $i, $diagonalValues[$i]);
}
}
}
class IdentityMatrix extends DiagonalMatrix
{
public function __construct($rows)
{
parent::__construct(\array_fill(0, $rows, 1));
}
}
?>