Fix cholesky computation.

This commit is contained in:
sebcrozet 2018-10-30 17:29:32 +01:00
parent 7ecbacacda
commit 9bf1d0280d
7 changed files with 425 additions and 187 deletions

View File

@ -52,6 +52,15 @@ where
pub(crate) vals: Vec<N>, pub(crate) vals: Vec<N>,
} }
impl<N: Scalar, R: Dim, C: Dim> CsVecStorage<N, R, C>
where
DefaultAllocator: Allocator<usize, C>,
{
pub fn values(&self) -> &[N] {
&self.vals
}
}
impl<N: Scalar, R: Dim, C: Dim> CsVecStorage<N, R, C> where DefaultAllocator: Allocator<usize, C> {} impl<N: Scalar, R: Dim, C: Dim> CsVecStorage<N, R, C> where DefaultAllocator: Allocator<usize, C> {}
impl<'a, N: Scalar, R: Dim, C: Dim> CsStorageIter<'a, N, R, C> for CsVecStorage<N, R, C> impl<'a, N: Scalar, R: Dim, C: Dim> CsStorageIter<'a, N, R, C> for CsVecStorage<N, R, C>
@ -187,10 +196,35 @@ where
} }
impl<N: Scalar, R: Dim, C: Dim, S: CsStorage<N, R, C>> CsMatrix<N, R, C, S> { impl<N: Scalar, R: Dim, C: Dim, S: CsStorage<N, R, C>> CsMatrix<N, R, C, S> {
pub fn from_data(data: S) -> Self {
CsMatrix {
data,
_phantoms: PhantomData,
}
}
pub fn len(&self) -> usize { pub fn len(&self) -> usize {
self.data.len() self.data.len()
} }
pub fn nrows(&self) -> usize {
self.data.shape().0.value()
}
pub fn ncols(&self) -> usize {
self.data.shape().1.value()
}
pub fn shape(&self) -> (usize, usize) {
let (nrows, ncols) = self.data.shape();
(nrows.value(), ncols.value())
}
pub fn is_square(&self) -> bool {
let (nrows, ncols) = self.data.shape();
nrows.value() == ncols.value()
}
pub fn transpose(&self) -> CsMatrix<N, C, R> pub fn transpose(&self) -> CsMatrix<N, C, R>
where where
DefaultAllocator: Allocator<usize, R>, DefaultAllocator: Allocator<usize, R>,

View File

@ -1,184 +0,0 @@
use alga::general::{ClosedAdd, ClosedMul};
use num::{One, Zero};
use std::iter;
use std::marker::PhantomData;
use std::ops::{Add, Mul, Range};
use std::slice;
use allocator::Allocator;
use constraint::{AreMultipliable, DimEq, SameNumberOfRows, ShapeConstraint};
use sparse::{CsMatrix, CsStorage, CsVector};
use storage::{Storage, StorageMut};
use {DefaultAllocator, Dim, Matrix, MatrixMN, Real, Scalar, Vector, VectorN, U1};
pub struct SymbolicAnalysis {
pinv: Vec<usize>,
q: Vec<usize>,
elimination_tree: Vec<usize>,
cp: Vec<usize>,
leftmost: Vec<usize>,
m2: usize,
lnz: usize,
unz: usize,
}
#[derive(Copy, Clone, Debug)]
pub struct EliminationTreeNode {
parent: usize,
}
impl EliminationTreeNode {
pub fn root() -> Self {
EliminationTreeNode {
parent: usize::max_value(),
}
}
pub fn with_parent(parent: usize) -> Self {
EliminationTreeNode { parent }
}
pub fn is_root(&self) -> bool {
self.parent == usize::max_value()
}
pub fn parent(&self) -> usize {
self.parent
}
}
impl<N: Real, D: Dim, S: CsStorage<N, D, D>> CsMatrix<N, D, D, S> {
fn elimination_tree(&self) -> Vec<EliminationTreeNode> {
let (nrows, ncols) = self.data.shape();
assert_eq!(
nrows.value(),
ncols.value(),
"The matrix `self` must be square to compute its elimination tree."
);
let mut forest: Vec<_> = iter::repeat(EliminationTreeNode::root())
.take(nrows.value())
.collect();
let mut ancestor: Vec<_> = iter::repeat(usize::max_value())
.take(nrows.value())
.collect();
for k in 0..nrows.value() {
for irow in self.data.column_row_indices(k) {
let mut i = irow;
while i < k {
let i_ancestor = ancestor[i];
ancestor[i] = k;
if i_ancestor == usize::max_value() {
forest[i] = EliminationTreeNode::with_parent(k);
break;
}
i = i_ancestor;
}
}
}
forest
}
fn reach(
&self,
j: usize,
max_j: usize,
tree: &[EliminationTreeNode],
marks: &mut Vec<bool>,
out: &mut Vec<usize>,
) {
marks.clear();
marks.resize(tree.len(), false);
for irow in self.data.column_row_indices(j) {
let mut curr = irow;
while curr != usize::max_value() && curr <= max_j && !marks[curr] {
marks[curr] = true;
out.push(curr);
curr = tree[curr].parent;
}
}
}
fn column_counts(&self, tree: &[EliminationTreeNode]) -> Vec<usize> {
let len = self.data.shape().0.value();
let mut counts: Vec<_> = iter::repeat(0).take(len).collect();
let mut reach = Vec::new();
let mut marks = Vec::new();
for i in 0..len {
self.reach(i, i, tree, &mut marks, &mut reach);
for j in reach.drain(..) {
counts[j] += 1;
}
}
counts
}
fn tree_postorder(tree: &[EliminationTreeNode]) -> Vec<usize> {
// FIXME: avoid all those allocations?
let mut first_child: Vec<_> = iter::repeat(usize::max_value()).take(tree.len()).collect();
let mut other_children: Vec<_> =
iter::repeat(usize::max_value()).take(tree.len()).collect();
// Build the children list from the parent list.
// The set of children of the node `i` is given by the linked list
// starting at `first_child[i]`. The nodes of this list are then:
// { first_child[i], other_children[first_child[i]], other_children[other_children[first_child[i]], ... }
for (i, node) in tree.iter().enumerate() {
if !node.is_root() {
let brother = first_child[node.parent];
first_child[node.parent] = i;
other_children[i] = brother;
}
}
let mut stack = Vec::with_capacity(tree.len());
let mut postorder = Vec::with_capacity(tree.len());
for (i, node) in tree.iter().enumerate() {
if node.is_root() {
Self::dfs(
i,
&mut first_child,
&other_children,
&mut stack,
&mut postorder,
)
}
}
postorder
}
fn dfs(
i: usize,
first_child: &mut [usize],
other_children: &[usize],
stack: &mut Vec<usize>,
result: &mut Vec<usize>,
) {
stack.clear();
stack.push(i);
while let Some(n) = stack.pop() {
let child = first_child[n];
if child == usize::max_value() {
// No children left.
result.push(n);
} else {
stack.push(n);
stack.push(child);
first_child[n] = other_children[child];
}
}
}
}

View File

@ -1 +1,331 @@
use alga::general::{ClosedAdd, ClosedMul};
use num::{One, Zero};
use std::iter;
use std::marker::PhantomData;
use std::mem;
use std::ops::{Add, Mul, Range};
use std::slice;
use allocator::Allocator;
use constraint::{AreMultipliable, DimEq, SameNumberOfRows, ShapeConstraint};
use sparse::{CsMatrix, CsStorage, CsStorageIter, CsVecStorage, CsVector};
use storage::{Storage, StorageMut};
use {DefaultAllocator, Dim, Matrix, MatrixMN, Real, Scalar, Vector, VectorN, U1};
pub struct CsCholesky<N: Real, D: Dim>
where
DefaultAllocator: Allocator<usize, D> + Allocator<N, D>,
{
// Non-zero pattern of the original matrix upper-triangular part.
// Unlike the original matrix, the `original_p` array does contain the last sentinel value
// equal to `original_i.len()` at the end.
original_p: Vec<usize>,
original_i: Vec<usize>,
original_len: usize, // Number of elements on the numerical value vector of the original matrix.
// Decomposition result.
l: CsMatrix<N, D, D>,
// Used only for the pattern.
// FIXME: store only the nonzero pattern instead.
u: CsMatrix<N, D, D>,
ok: bool,
// Workspaces.
work_x: VectorN<N, D>,
work_c: VectorN<usize, D>,
}
impl<N: Real, D: Dim> CsCholesky<N, D>
where
DefaultAllocator: Allocator<usize, D> + Allocator<N, D>,
{
/// Computes the cholesky decomposition of the sparse matrix `m`.
pub fn new(m: &CsMatrix<N, D, D>) -> Self {
let mut me = Self::new_symbolic(m);
let _ = me.decompose(&m.data.vals);
me
}
/// Perform symbolic analysis for the given matrix.
///
/// This does not access the numerical values of `m`.
pub fn new_symbolic(m: &CsMatrix<N, D, D>) -> Self {
assert!(
m.is_square(),
"The matrix `m` must be square to compute its elimination tree."
);
let (l, u) = Self::nonzero_pattern(m);
// Workspaces.
let work_x = unsafe { VectorN::new_uninitialized_generic(m.data.shape().0, U1) };
let work_c = unsafe { VectorN::new_uninitialized_generic(m.data.shape().1, U1) };
let mut original_p = m.data.p.as_slice().to_vec();
original_p.push(m.data.i.len());
CsCholesky {
original_p,
original_i: m.data.i.clone(),
original_len: m.data.i.len(),
l,
u,
ok: false,
work_x,
work_c,
}
}
pub fn l(&self) -> Option<&CsMatrix<N, D, D>> {
if self.ok {
Some(&self.l)
} else {
None
}
}
pub fn unwrap_l(self) -> Option<CsMatrix<N, D, D>> {
if self.ok {
Some(self.l)
} else {
None
}
}
// Performs the numerical Cholesky decomposition given the set of numerical values.
pub fn decompose(&mut self, values: &[N]) -> bool {
assert!(
values.len() >= self.original_len,
"The set of values is too small."
);
// Reset `work_c` to the column pointers of `l`.
self.work_c.copy_from(&self.l.data.p);
// Perform the decomposition.
for k in 0..self.l.nrows() {
// Scatter the k-th column of the original matrix with the values provided.
let column_range = self.original_p[k]..self.original_p[k + 1];
self.work_x[k] = N::zero();
for p in column_range.clone() {
let irow = self.original_i[p];
if irow <= k {
self.work_x[irow] = values[p];
}
}
let mut diag = self.work_x[k];
self.work_x[k] = N::zero();
// Triangular solve.
for irow in self.u.data.column_row_indices(k) {
if irow >= k {
continue;
}
let lki = self.work_x[irow] / self.l.data.vals[self.l.data.p[irow]];
self.work_x[irow] = N::zero();
for p in self.l.data.p[irow] + 1..self.work_c[irow] {
self.work_x[self.l.data.i[p]] -= self.l.data.vals[p] * lki;
}
diag -= lki * lki;
let p = self.work_c[irow];
self.work_c[irow] += 1;
self.l.data.i[p] = k;
self.l.data.vals[p] = lki;
}
if diag <= N::zero() {
self.ok = false;
return false;
}
// Deal with the diagonal element.
let p = self.work_c[k];
self.work_c[k] += 1;
self.l.data.i[p] = k;
self.l.data.vals[p] = diag.sqrt();
}
self.ok = true;
true
}
fn elimination_tree<S: CsStorage<N, D, D>>(m: &CsMatrix<N, D, D, S>) -> Vec<usize> {
let nrows = m.nrows();
let mut forest: Vec<_> = iter::repeat(usize::max_value()).take(nrows).collect();
let mut ancestor: Vec<_> = iter::repeat(usize::max_value()).take(nrows).collect();
for k in 0..nrows {
for irow in m.data.column_row_indices(k) {
let mut i = irow;
while i < k {
let i_ancestor = ancestor[i];
ancestor[i] = k;
if i_ancestor == usize::max_value() {
forest[i] = k;
break;
}
i = i_ancestor;
}
}
}
forest
}
fn reach<S: CsStorage<N, D, D>>(
m: &CsMatrix<N, D, D, S>,
j: usize,
max_j: usize,
tree: &[usize],
marks: &mut Vec<bool>,
out: &mut Vec<usize>,
) {
marks.clear();
marks.resize(tree.len(), false);
// FIXME: avoid all those allocations.
let mut tmp = Vec::new();
let mut res = Vec::new();
for irow in m.data.column_row_indices(j) {
let mut curr = irow;
while curr != usize::max_value() && curr <= max_j && !marks[curr] {
marks[curr] = true;
tmp.push(curr);
curr = tree[curr];
}
tmp.append(&mut res);
mem::swap(&mut tmp, &mut res);
}
out.append(&mut res);
}
fn nonzero_pattern<S: CsStorage<N, D, D>>(
m: &CsMatrix<N, D, D, S>,
) -> (CsMatrix<N, D, D>, CsMatrix<N, D, D>) {
let etree = Self::elimination_tree(m);
let (nrows, ncols) = m.data.shape();
let mut rows = Vec::with_capacity(m.len());
let mut cols = unsafe { VectorN::new_uninitialized_generic(m.data.shape().0, U1) };
let mut marks = Vec::new();
// NOTE: the following will actually compute the non-zero pattern of
// the transpose of l.
for i in 0..nrows.value() {
cols[i] = rows.len();
Self::reach(m, i, i, &etree, &mut marks, &mut rows);
}
let mut vals = Vec::with_capacity(rows.len());
unsafe {
vals.set_len(rows.len());
}
vals.shrink_to_fit();
let data = CsVecStorage {
shape: (nrows, ncols),
p: cols,
i: rows,
vals,
};
let u = CsMatrix::from_data(data);
// XXX: avoid this transpose.
let l = u.transpose();
(l, u)
}
/*
*
* NOTE: All the following methods are untested and currently unused.
*
*
fn column_counts<S: CsStorage<N, D, D>>(
m: &CsMatrix<N, D, D, S>,
tree: &[usize],
) -> Vec<usize> {
let len = m.data.shape().0.value();
let mut counts: Vec<_> = iter::repeat(0).take(len).collect();
let mut reach = Vec::new();
let mut marks = Vec::new();
for i in 0..len {
Self::reach(m, i, i, tree, &mut marks, &mut reach);
for j in reach.drain(..) {
counts[j] += 1;
}
}
counts
}
fn tree_postorder(tree: &[usize]) -> Vec<usize> {
// FIXME: avoid all those allocations?
let mut first_child: Vec<_> = iter::repeat(usize::max_value()).take(tree.len()).collect();
let mut other_children: Vec<_> =
iter::repeat(usize::max_value()).take(tree.len()).collect();
// Build the children list from the parent list.
// The set of children of the node `i` is given by the linked list
// starting at `first_child[i]`. The nodes of this list are then:
// { first_child[i], other_children[first_child[i]], other_children[other_children[first_child[i]], ... }
for (i, parent) in tree.iter().enumerate() {
if *parent != usize::max_value() {
let brother = first_child[*parent];
first_child[*parent] = i;
other_children[i] = brother;
}
}
let mut stack = Vec::with_capacity(tree.len());
let mut postorder = Vec::with_capacity(tree.len());
for (i, node) in tree.iter().enumerate() {
if *node == usize::max_value() {
Self::dfs(
i,
&mut first_child,
&other_children,
&mut stack,
&mut postorder,
)
}
}
postorder
}
fn dfs(
i: usize,
first_child: &mut [usize],
other_children: &[usize],
stack: &mut Vec<usize>,
result: &mut Vec<usize>,
) {
stack.clear();
stack.push(i);
while let Some(n) = stack.pop() {
let child = first_child[n];
if child == usize::max_value() {
// No children left.
result.push(n);
} else {
stack.push(n);
stack.push(child);
first_child[n] = other_children[child];
}
}
}
*/
}

View File

@ -1,7 +1,9 @@
pub use self::cs_matrix::{CsMatrix, CsStorage, CsStorageMut, CsVector}; pub use self::cs_matrix::{
CsMatrix, CsStorage, CsStorageIter, CsStorageMut, CsVecStorage, CsVector,
};
pub use self::cs_matrix_cholesky::CsCholesky;
mod cs_matrix; mod cs_matrix;
mod cs_matrix_analysis;
mod cs_matrix_cholesky; mod cs_matrix_cholesky;
mod cs_matrix_conversion; mod cs_matrix_conversion;
mod cs_matrix_ops; mod cs_matrix_ops;

View File

@ -0,0 +1,55 @@
#![cfg_attr(rustfmt, rustfmt_skip)]
use na::{CsMatrix, CsVector, CsCholesky, Cholesky, Matrix5, Vector5};
#[test]
fn cs_cholesky() {
let mut a = Matrix5::new(
40.0, 0.0, 0.0, 0.0, 0.0,
2.0, 60.0, 0.0, 0.0, 0.0,
1.0, 0.0, 11.0, 0.0, 0.0,
0.0, 0.0, 0.0, 50.0, 0.0,
1.0, 0.0, 0.0, 4.0, 10.0
);
a.fill_upper_triangle_with_lower_triangle();
test_cholesky(a);
let a = Matrix5::from_diagonal(&Vector5::new(40.0, 60.0, 11.0, 50.0, 10.0));
test_cholesky(a);
let mut a = Matrix5::new(
40.0, 0.0, 0.0, 0.0, 0.0,
2.0, 60.0, 0.0, 0.0, 0.0,
1.0, 0.0, 11.0, 0.0, 0.0,
1.0, 0.0, 0.0, 50.0, 0.0,
0.0, 0.0, 0.0, 4.0, 10.0
);
a.fill_upper_triangle_with_lower_triangle();
test_cholesky(a);
let mut a = Matrix5::new(
2.0, 0.0, 0.0, 0.0, 0.0,
0.0, 2.0, 0.0, 0.0, 0.0,
1.0, 1.0, 2.0, 0.0, 0.0,
0.0, 0.0, 0.0, 2.0, 0.0,
1.0, 1.0, 0.0, 0.0, 2.0
);
a.fill_upper_triangle_with_lower_triangle();
test_cholesky(a);
}
fn test_cholesky(a: Matrix5<f32>) {
let cs_a: CsMatrix<_, _, _> = a.into();
let chol_a = Cholesky::new(a).unwrap();
let chol_cs_a = CsCholesky::new(&cs_a);
let l = chol_a.l();
println!("{:?}", chol_cs_a.l());
let cs_l: Matrix5<_> = chol_cs_a.unwrap_l().unwrap().into();
println!("{}", l);
println!("{}", cs_l);
assert_eq!(l, cs_l);
}

View File

@ -1,5 +1,6 @@
mod cs_cholesky;
mod cs_construction; mod cs_construction;
mod cs_conversion; mod cs_conversion;
mod cs_matrix; mod cs_matrix;
mod cs_ops; mod cs_ops;
mod cs_linalg; mod cs_solve;