2018-05-31 10:47:31 +02:00
|
|
|
(** Direct Inversion of the Iterative Subspace algorithm.
|
|
|
|
|
2018-05-31 10:56:01 +02:00
|
|
|
At iteration {% $m$ %}, one has:
|
2018-05-31 10:47:31 +02:00
|
|
|
|
2018-05-31 10:56:01 +02:00
|
|
|
- {% $\mathbf{p}_m$ %}, a vector of parameters
|
|
|
|
- {% $\mathbf{e}_m$ %}, an approximate error vector
|
|
|
|
|
2019-01-15 15:18:01 +01:00
|
|
|
The DIIS approximate solution for iteration {% $m+1$ %} is given by
|
2018-05-31 10:47:31 +02:00
|
|
|
|
|
|
|
{% \begin{align*}
|
2018-05-31 10:56:01 +02:00
|
|
|
\mathbf{p}_{m+1} & = \sum_{i=1}^m c_i (\mathbf{p}^f + \mathbf{e}_i) \\
|
2019-12-05 15:10:14 +01:00
|
|
|
& = \sum_{i=1}^m c_i \mathbf{p}^f + \sum_i c_i \mathbf{e}_i
|
2018-05-31 10:47:31 +02:00
|
|
|
\end{align*} %}
|
2018-05-31 10:56:01 +02:00
|
|
|
|
|
|
|
where {% $\mathbf{p}^f$ %} is the exact solution. One wants to minimize the
|
|
|
|
norm of the error vector imposing the constraint that {% $\sum_{i=1}^m c_i = 1$ %}
|
|
|
|
with a Langrange multiplier {% $\lambda$ %}.
|
2018-05-31 10:47:31 +02:00
|
|
|
|
|
|
|
{%
|
|
|
|
\begin{align*}
|
2018-05-31 10:56:01 +02:00
|
|
|
\mathcal{L} & = ||\sum_i c_i \mathbf{e}_i||^2 - \lambda \left(\sum_i c_i - 1\right) \\
|
2018-06-27 13:13:59 +02:00
|
|
|
& = \sum_{ij} c_i c_j B_{ij} - \lambda \left(\sum_{i=1}^m c_i - 1\right)
|
2018-05-31 10:47:31 +02:00
|
|
|
\end{align*}
|
2018-05-31 10:56:01 +02:00
|
|
|
%}
|
|
|
|
with
|
|
|
|
{% $B_{ij} = \langle \mathbf{e}_i | \mathbf{e}_j \rangle$ %}.
|
2018-05-31 10:47:31 +02:00
|
|
|
|
|
|
|
Equating zero to the derivatives of {% $\mathcal{L}$ %} with respect to {% $c_i$ %} and {% $\lambda$ %} leads to
|
|
|
|
|
|
|
|
{% \begin{equation*}
|
|
|
|
\begin{bmatrix}
|
2018-05-31 16:46:45 +02:00
|
|
|
B_{11} & B_{12} & B_{13} & ... & B_{1m} & 1 \\
|
|
|
|
B_{21} & B_{22} & B_{23} & ... & B_{2m} & 1 \\
|
|
|
|
B_{31} & B_{32} & B_{33} & ... & B_{3m} & 1 \\
|
2018-05-31 10:47:31 +02:00
|
|
|
\vdots & \vdots & \vdots & \vdots & \ddots & \vdots \\
|
2018-05-31 16:46:45 +02:00
|
|
|
B_{m1} & B_{m2} & B_{m3} & ... & B_{mm} & 1 \\
|
2018-05-31 10:47:31 +02:00
|
|
|
1 & 1 & 1 & ... & 1 & 0
|
2018-05-31 16:46:45 +02:00
|
|
|
\end{bmatrix} \begin{bmatrix} c_1 \\ c_2 \\ c_3 \\ \vdots \\ c_m \\ -\lambda \end{bmatrix}=
|
2018-05-31 10:47:31 +02:00
|
|
|
\begin{bmatrix} 0 \\ 0 \\ 0 \\ \vdots \\ 0 \\ 1 \end{bmatrix}
|
|
|
|
\end{equation*}
|
|
|
|
|
|
|
|
%}
|
|
|
|
|
|
|
|
The coefficients are then used to update {% $\mathbf{p}$ %} as
|
|
|
|
{% $$
|
|
|
|
\mathbf{p}_{m+1}=\sum_{i=1}^m c_i\mathbf{p}_i.
|
|
|
|
$$ %}
|
|
|
|
|
|
|
|
*)
|
|
|
|
|
|
|
|
type t
|
|
|
|
|
2018-05-31 16:46:45 +02:00
|
|
|
val make : ?mmax:int -> unit -> t
|
|
|
|
(** Initialize DIIS with a maximum size.*)
|
2018-05-31 10:47:31 +02:00
|
|
|
|
2018-05-31 16:46:45 +02:00
|
|
|
val append : p:Lacaml.D.Vec.t -> e:Lacaml.D.Vec.t -> t -> t
|
2018-05-31 10:47:31 +02:00
|
|
|
(** Append a parameter vector [p] and the corresponding error vector [e]. *)
|
|
|
|
|
|
|
|
val next : t -> Lacaml.D.Vec.t
|
|
|
|
(** Returns a new parameter vector. *)
|
|
|
|
|
|
|
|
|