Class: Matrix::LUPDecomposition

Inherits:
Object
  • Object
show all
Includes:
ConversionHelper
Defined in:
opal/stdlib/matrix/lup_decomposition.rb

Overview

For an m-by-n matrix A with m >= n, the LU decomposition is an m-by-n unit lower triangular matrix L, an n-by-n upper triangular matrix U, and a m-by-m permutation matrix P so that L*U = P*A. If m < n, then L is m-by-m and U is m-by-n.

The LUP decomposition with pivoting always exists, even if the matrix is singular, so the constructor will never fail. The primary use of the LU decomposition is in the solution of square systems of simultaneous linear equations. This will fail if singular? returns true.

Instance Attribute Summary collapse

Instance Method Summary collapse

Constructor Details

#initialize(a) ⇒ LUPDecomposition

Returns a new instance of LUPDecomposition

Raises:

  • (TypeError)


154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
# File 'opal/stdlib/matrix/lup_decomposition.rb', line 154

def initialize a
  raise TypeError, "Expected Matrix but got #{a.class}" unless a.is_a?(Matrix)
  # Use a "left-looking", dot-product, Crout/Doolittle algorithm.
  @lu = a.to_a
  @row_count = a.row_count
  @column_count = a.column_count
  @pivots = Array.new(@row_count)
  @row_count.times do |i|
     @pivots[i] = i
  end
  @pivot_sign = 1
  lu_col_j = Array.new(@row_count)

  # Outer loop.

  @column_count.times do |j|

    # Make a copy of the j-th column to localize references.

    @row_count.times do |i|
      lu_col_j[i] = @lu[i][j]
    end

    # Apply previous transformations.

    @row_count.times do |i|
      lu_row_i = @lu[i]

      # Most of the time is spent in the following dot product.

      kmax = [i, j].min
      s = 0
      kmax.times do |k|
        s += lu_row_i[k]*lu_col_j[k]
      end

      lu_row_i[j] = lu_col_j[i] -= s
    end

    # Find pivot and exchange if necessary.

    p = j
    (j+1).upto(@row_count-1) do |i|
      if (lu_col_j[i].abs > lu_col_j[p].abs)
        p = i
      end
    end
    if (p != j)
      @column_count.times do |k|
        t = @lu[p][k]; @lu[p][k] = @lu[j][k]; @lu[j][k] = t
      end
      k = @pivots[p]; @pivots[p] = @pivots[j]; @pivots[j] = k
      @pivot_sign = -@pivot_sign
    end

    # Compute multipliers.

    if (j < @row_count && @lu[j][j] != 0)
      (j+1).upto(@row_count-1) do |i|
        @lu[i][j] = @lu[i][j].quo(@lu[j][j])
      end
    end
  end
end

Instance Attribute Details

#pivotsObject (readonly)

Returns the pivoting indices



63
64
65
# File 'opal/stdlib/matrix/lup_decomposition.rb', line 63

def pivots
  @pivots
end

Instance Method Details

#detObject Also known as: determinant

Returns the determinant of +A+, calculated efficiently from the factorization.



79
80
81
82
83
84
85
86
87
88
# File 'opal/stdlib/matrix/lup_decomposition.rb', line 79

def det
  if (@row_count != @column_count)
    Matrix.Raise Matrix::ErrDimensionMismatch
  end
  d = @pivot_sign
  @column_count.times do |j|
    d *= @lu[j][j]
  end
  d
end

#lObject



22
23
24
25
26
27
28
29
30
31
32
# File 'opal/stdlib/matrix/lup_decomposition.rb', line 22

def l
  Matrix.build(@row_count, [@column_count, @row_count].min) do |i, j|
    if (i > j)
      @lu[i][j]
    elsif (i == j)
      1
    else
      0
    end
  end
end

#pObject

Returns the permutation matrix +P+



48
49
50
51
52
# File 'opal/stdlib/matrix/lup_decomposition.rb', line 48

def p
  rows = Array.new(@row_count){Array.new(@row_count, 0)}
  @pivots.each_with_index{|p, i| rows[i][p] = 1}
  Matrix.send :new, rows, @row_count
end

#singular?Boolean

Returns +true+ if +U+, and hence +A+, is singular.

Returns:



67
68
69
70
71
72
73
74
# File 'opal/stdlib/matrix/lup_decomposition.rb', line 67

def singular?
  @column_count.times do |j|
    if (@lu[j][j] == 0)
      return true
    end
  end
  false
end

#solve(b) ⇒ Object

Returns +m+ so that A*m = b, or equivalently so that L*U*m = P*b +b+ can be a Matrix or a Vector



95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
# File 'opal/stdlib/matrix/lup_decomposition.rb', line 95

def solve b
  if (singular?)
    Matrix.Raise Matrix::ErrNotRegular, "Matrix is singular."
  end
  if b.is_a? Matrix
    if (b.row_count != @row_count)
      Matrix.Raise Matrix::ErrDimensionMismatch
    end

    # Copy right hand side with pivoting
    nx = b.column_count
    m = @pivots.map{|row| b.row(row).to_a}

    # Solve L*Y = P*b
    @column_count.times do |k|
      (k+1).upto(@column_count-1) do |i|
        nx.times do |j|
          m[i][j] -= m[k][j]*@lu[i][k]
        end
      end
    end
    # Solve U*m = Y
    (@column_count-1).downto(0) do |k|
      nx.times do |j|
        m[k][j] = m[k][j].quo(@lu[k][k])
      end
      k.times do |i|
        nx.times do |j|
          m[i][j] -= m[k][j]*@lu[i][k]
        end
      end
    end
    Matrix.send :new, m, nx
  else # same algorithm, specialized for simpler case of a vector
    b = convert_to_array(b)
    if (b.size != @row_count)
      Matrix.Raise Matrix::ErrDimensionMismatch
    end

    # Copy right hand side with pivoting
    m = b.values_at(*@pivots)

    # Solve L*Y = P*b
    @column_count.times do |k|
      (k+1).upto(@column_count-1) do |i|
        m[i] -= m[k]*@lu[i][k]
      end
    end
    # Solve U*m = Y
    (@column_count-1).downto(0) do |k|
      m[k] = m[k].quo(@lu[k][k])
      k.times do |i|
        m[i] -= m[k]*@lu[i][k]
      end
    end
    Vector.elements(m, false)
  end
end

#to_aryObject Also known as: to_a

Returns +L+, +U+, +P+ in an array



56
57
58
# File 'opal/stdlib/matrix/lup_decomposition.rb', line 56

def to_ary
  [l, u, p]
end

#uObject

Returns the upper triangular factor +U+



36
37
38
39
40
41
42
43
44
# File 'opal/stdlib/matrix/lup_decomposition.rb', line 36

def u
  Matrix.build([@column_count, @row_count].min, @column_count) do |i, j|
    if (i <= j)
      @lu[i][j]
    else
      0
    end
  end
end