2.1.0
User Documentation for Apache MADlib
linalg.sql_in File Reference

SQL functions for linear algebra. More...

Functions

float8 norm1 (float8[] x)
 1-norm of a vector More...
 
float8 norm2 (float8[] x)
 2-norm of a vector More...
 
float8 dist_inf_norm (float8[] x, float8[] y)
 Infinity-norm of the difference between two vectors. More...
 
float8 dist_pnorm (float8[] x, float8[] y, float8 p)
 p-norm of the difference between two vectors More...
 
float8 dist_norm1 (float8[] x, float8[] y)
 1-norm of the difference between two vectors More...
 
float8 dist_norm2 (float8[] x, float8[] y)
 2-norm of the difference between two vectors More...
 
float8 cosine_similarity (float8[] x, float8[] y)
 cosine similarity score between two vectors More...
 
float8 squared_dist_norm2 (float8[] x, float8[] y)
 Squared 2-norm of the difference between two vectors. More...
 
float8 dist_angle (float8[] x, float8[] y)
 Angle between two vectors. More...
 
float8 dist_tanimoto (float8[] x, float8[] y)
 Tanimoto distance between two vectors. More...
 
float8 dist_jaccard (text[] x, text[] y)
 Jaccard distance between two vectors (treated as sets) More...
 
closest_column_result _closest_column (float8[] m, float8[] x, regproc dist, text dist_dn)
 
closest_column_result closest_column (float8[] m, float8[] x, regproc dist="squared_dist_norm2")
 Given matrix \( M \) and vector \( \vec x \) compute the column of \( M \) that is closest to \( \vec x \). More...
 
closest_column_result closest_column (float8[] m, float8[] x)
 
closest_columns_result _closest_columns (float8[] m, float8[] x, integer num, regproc dist, text dist_dn)
 Given matrix \( M \) and vector \( \vec x \) compute the columns of \( M \) that are closest to \( \vec x \). More...
 
closest_columns_result closest_columns (float8[] m, float8[] x, integer num, regproc dist)
 
closest_columns_result closest_columns (float8[] m, float8[] x, integer num)
 
float8 [] avg_vector_transition (float8[] state, float8[] x)
 
float8 [] avg_vector_merge (float8[] state_left, float8[] state_right)
 
float8 [] avg_vector_final (float8[] state)
 
aggregate float8 [] avg (float8[] x)
 Compute the average of vectors. More...
 
float8 [] normalized_avg_vector_transition (float8[] state, float8[] x)
 
float8 [] normalized_avg_vector_final (float8[] state)
 
aggregate float8 [] normalized_avg (float8[] x)
 Compute the normalized average of vectors. More...
 
float8 [] matrix_agg_transition (float8[] state, float8[] x)
 
float8 [] matrix_agg_final (float8[] state)
 
aggregate float8 [] matrix_agg (float8[] x)
 Combine vectors to a matrix. More...
 
float8 [] matrix_column (float8[] matrix, integer col)
 Return the column of a matrix. More...
 
set< record > deconstruct_2d_array (float8[] in_array)
 Construct a M-column N-row table of 2-D array. More...
 
set< record > __deconstruct_lower_triangle (float8[] in_array)
 Construct a M-column M-row table of using lower triangle of 2-D array. More...
 
float8 [] array_to_1d (float8[] in_array)
 Return the input array as 1-D. More...
 
float8 [] array_to_2d (float8[] in_array)
 Return A 2-D matrix that the number of rows is encoded as the first element of the input array and the number of cols second. More...
 
float8 [] index_2d_array (float8[] in_2d_array, integer index)
 Return A 1-D form of the specified row of the given 2-D array. More...
 
float8 [] get_row (float8[] in_2d_array, integer index)
 Get an indexed row of the given matrix (2-D array) More...
 
float8 [] get_col (float8[] in_2d_array, integer index)
 Get an indexed col of the given matrix (2-D array) More...
 

Detailed Description

See also
For an overview of linear-algebra functions, see the module description Norms and Distance Functions.

Function Documentation

◆ __deconstruct_lower_triangle()

set<record> __deconstruct_lower_triangle ( float8 []  in_array)
Parameters
in_array2-D array

◆ _closest_column()

closest_column_result _closest_column ( float8 []  m,
float8 []  x,
regproc  dist,
text  dist_dn 
)

◆ _closest_columns()

closest_columns_result _closest_columns ( float8 []  m,
float8 []  x,
integer  num,
regproc  dist,
text  dist_dn 
)

This function does essentially the same as closest_column(), except that it allows to specify the number of closest columns to return. The return value is a composite value:

  • columns_ids INTEGER[] - The 0-based indices of the num columns of \( M \) that are closest to \( x \). In case of ties, the first such indices are returned.
  • distances DOUBLE PRECISION[] - The distances between the columns of \( M \) with indices in columns_ids and \( x \). That is, distances[i] contains \( \operatorname{dist}(\vec{m_j}, \vec x) \), where \( j = \) columns_ids[i].

◆ array_to_1d()

float8 [] array_to_1d ( float8 []  in_array)
Parameters
in_array1-D or 2-D array

◆ array_to_2d()

float8 [] array_to_2d ( float8 []  in_array)
Parameters
in_arrayInput array with first 2 elements describing dimensions

◆ avg()

aggregate float8 [] avg ( float8 []  x)

Given vectors \( x_1, \dots, x_n \), compute the average \( \frac 1n \sum_{i=1}^n x_i \).

Parameters
xPoint \( x_i \)
Returns
Average \( \frac 1n \sum_{i=1}^n x_i \)

◆ avg_vector_final()

float8 [] avg_vector_final ( float8 []  state)

◆ avg_vector_merge()

float8 [] avg_vector_merge ( float8 []  state_left,
float8 []  state_right 
)

◆ avg_vector_transition()

float8 [] avg_vector_transition ( float8 []  state,
float8 []  x 
)

◆ closest_column() [1/2]

closest_column_result closest_column ( float8 []  m,
float8 []  x,
regproc  dist = "squared_dist_norm2" 
)
Parameters
MMatrix \( M = (\vec{m_0} \dots \vec{m_{l-1}}) \in \mathbb{R}^{k \times l} \)
xVector \( \vec x \in \mathbb R^k \)
distThe metric \( \operatorname{dist} \). This needs to be a function with signature DOUBLE PRECISION[] x DOUBLE PRECISION[] -> DOUBLE PRECISION.
Returns
A composite value:
  • columns_id INTEGER - The 0-based index of the column of \( M \) that is closest to \( x \). In case of ties, the first such index is returned. That is, columns_id is the minimum element in the set \( \arg\min_{i=0,\dots,l-1} \operatorname{dist}(\vec{m_i}, \vec x) \).
  • distance DOUBLE PRECISION - The minimum distance between any column of \( M \) and \( x \). That is, \( \min_{i=0,\dots,l-1} \operatorname{dist}(\vec{m_i}, \vec x) \).

◆ closest_column() [2/2]

closest_column_result closest_column ( float8 []  m,
float8 []  x 
)

◆ closest_columns() [1/2]

closest_columns_result closest_columns ( float8 []  m,
float8 []  x,
integer  num,
regproc  dist 
)

◆ closest_columns() [2/2]

closest_columns_result closest_columns ( float8 []  m,
float8 []  x,
integer  num 
)

◆ cosine_similarity()

float8 cosine_similarity ( float8 []  x,
float8 []  y 
)
Parameters
xVector \( \vec x = (x_1, \dots, x_n) \)
yVector \( \vec y = (y_1, \dots, y_n) \)
Returns
\( \frac{\langle \vec x, \vec y \rangle} {\| \vec x \| \cdot \| \vec y \|} \)

◆ deconstruct_2d_array()

set<record> deconstruct_2d_array ( float8 []  in_array)
Parameters
in_array2-D array

◆ dist_angle()

float8 dist_angle ( float8 []  x,
float8 []  y 
)
Parameters
xVector \( \vec x = (x_1, \dots, x_n) \)
yVector \( \vec y = (y_1, \dots, y_n) \)
Returns
\( \arccos\left(\frac{\langle \vec x, \vec y \rangle} {\| \vec x \| \cdot \| \vec y \|}\right) \)

◆ dist_inf_norm()

float8 dist_inf_norm ( float8 []  x,
float8 []  y 
)
Parameters
xVector \( \vec x = (x_1, \dots, x_n) \)
yVector \( \vec y = (y_1, \dots, y_n) \)
Returns
\( \| x - y \|_\infty = \max_{i=1}^n \|x_i - y_i\| \)

◆ dist_jaccard()

float8 dist_jaccard ( text []  x,
text []  y 
)
Parameters
xVector \( \vec x = (x_1, \dots, x_m) \)
yVector \( \vec y = (y_1, \dots, y_n) \)
Returns
\( 1 - \frac{|x \cap y|}{|x \cup y|} \)

◆ dist_norm1()

float8 dist_norm1 ( float8 []  x,
float8 []  y 
)
Parameters
xVector \( \vec x = (x_1, \dots, x_n) \)
yVector \( \vec y = (y_1, \dots, y_n) \)
Returns
\( \| x - y \|_1 = \sum_{i=1}^n |x_i - y_i| \)

◆ dist_norm2()

float8 dist_norm2 ( float8 []  x,
float8 []  y 
)
Parameters
xVector \( \vec x = (x_1, \dots, x_n) \)
yVector \( \vec y = (y_1, \dots, y_n) \)
Returns
\( \| x - y \|_2 = \sqrt{\sum_{i=1}^n (x_i - y_i)^2} \)

◆ dist_pnorm()

float8 dist_pnorm ( float8 []  x,
float8 []  y,
float8  p 
)
Parameters
xVector \( \vec x = (x_1, \dots, x_n) \)
yVector \( \vec y = (y_1, \dots, y_n) \)
pScalar \( p > 0 \)
Returns
\( \| x - y \|_p = (\sum_{i=1}^n \|x_i - y_i\|^p)^{\frac{1}{p}} \)

◆ dist_tanimoto()

float8 dist_tanimoto ( float8 []  x,
float8 []  y 
)
Parameters
xVector \( \vec x = (x_1, \dots, x_n) \)
yVector \( \vec y = (y_1, \dots, y_n) \)
Returns
\( 1 - \frac{\langle \vec x, \vec y \rangle} {\| \vec x \|^2 \cdot \| \vec y \|^2 - \langle \vec x, \vec y \rangle} \)

◆ get_col()

float8 [] get_col ( float8 []  in_2d_array,
integer  index 
)
Parameters
in_2d_arrayInput 2-D array
index

◆ get_row()

float8 [] get_row ( float8 []  in_2d_array,
integer  index 
)
Parameters
in_2d_arrayInput 2-D array
index

◆ index_2d_array()

float8 [] index_2d_array ( float8 []  in_2d_array,
integer  index 
)
Parameters
in_2d_arrayInput 2-D array
index

◆ matrix_agg()

aggregate float8 [] matrix_agg ( float8 []  x)

Given vectors \( \vec x_1, \dots, \vec x_n \in \mathbb R^m \), return matrix \( ( \vec x_1 \dots \vec x_n ) \in \mathbb R^{m \times n}\).

Parameters
xVector \( x_i \)
Returns
Matrix with columns \( x_1, \dots, x_n \)

◆ matrix_agg_final()

float8 [] matrix_agg_final ( float8 []  state)

◆ matrix_agg_transition()

float8 [] matrix_agg_transition ( float8 []  state,
float8 []  x 
)

◆ matrix_column()

float8 [] matrix_column ( float8 []  matrix,
integer  col 
)
Parameters
matrixTwo-dimensional matrix
colColumn of the matrix to return (0-based index)

◆ norm1()

float8 norm1 ( float8 []  x)
Parameters
xVector \( \vec x = (x_1, \dots, x_n) \)
Returns
\( \| x \|_1 = \sum_{i=1}^n |x_i| \)

◆ norm2()

float8 norm2 ( float8 []  x)
Parameters
xVector \( \vec x = (x_1, \dots, x_n) \)
Returns
\( \| x \|_2 = \sqrt{\sum_{i=1}^n x_i^2} \)

◆ normalized_avg()

aggregate float8 [] normalized_avg ( float8 []  x)

Given vectors \( x_1, \dots, x_n \), define \( \widetilde{x} := \frac 1n \sum_{i=1}^n \frac{x_i}{\| x_i \|} \), and compute the normalized average \( \frac{\widetilde{x}}{\| \widetilde{x} \|} \).

Parameters
xPoint \( x_i \)
Returns
Normalized average \( \frac{\widetilde{x}}{\| \widetilde{x} \|} \)

◆ normalized_avg_vector_final()

float8 [] normalized_avg_vector_final ( float8 []  state)

◆ normalized_avg_vector_transition()

float8 [] normalized_avg_vector_transition ( float8 []  state,
float8 []  x 
)

◆ squared_dist_norm2()

float8 squared_dist_norm2 ( float8 []  x,
float8 []  y 
)
Parameters
xVector \( \vec x = (x_1, \dots, x_n) \)
yVector \( \vec y = (y_1, \dots, y_n) \)
Returns
\( \| x - y \|_2^2 = \sum_{i=1}^n (x_i - y_i)^2 \)