Rework optimization routines (!81) · Merge requests · CHS / FORCES

Merged Sebastian Müller requested to merge reduced_eval_interface into main 5 months ago

Mar 13, 2025
- opt_eval_utils: update docs and comments · 81078282
  Sebastian Müller authored 1 month ago
  
  81078282
Mar 12, 2025
- move eval and objective interfaces to mo_opt_eval_utils · cb0b9de1
  Sebastian Müller authored 1 month ago
  
  cb0b9de1
- tests: filter new test for conditional compilation · d0d39b81
  Sebastian Müller authored 1 month ago
  
  d0d39b81
- test: rename opt test · 0b951fb0
  Sebastian Müller authored 1 month ago
  
  0b951fb0
- opt: rename optimization modules · dbd31b15
  Sebastian Müller authored 1 month ago
  
  dbd31b15
- tests: bugfix · eaf37019
  Sebastian Müller authored 1 month ago
  
  eaf37019
- tests: add tests for sim_data_t · 5291a109
  Sebastian Müller authored 1 month ago
  
  5291a109
- sim_data_t: add set_data to directly set variable data without using pointers · e471e134
  Sebastian Müller authored 1 month ago
  
  e471e134
- rename opti_sim to sim_data · 3d2786fc
  Sebastian Müller authored 1 month ago
  
  3d2786fc
- more docs · 4f778064
  Sebastian Müller authored 1 month ago
  
  4f778064
- opt_types: add more function docs · 6771e227
  Sebastian Müller authored 1 month ago
  
  6771e227
- sim_var_t: remove time_avg_selector · 01625078
  Sebastian Müller authored 1 month ago
  
  01625078
- Removed old uncommented code, removed ``optidata`` and ``optidata_sim`` · 7df7af88
  Maren Kaluza authored 1 month ago
  
  7df7af88
Mar 06, 2025

Removed desctructor again: Not needed in this case. · ebbd9f1c

Maren Kaluza authored 1 month ago

If only allocated arrays must be deallocated Fortran does it on its own
and no destructor/final is needed.

ebbd9f1c

Feb 26, 2025
- Implemented final procedures for sim_data_t and sim_var_t. · d9244a47
  Maren Kaluza authored 1 month ago
  
  d9244a47
Jan 29, 2025
- opti_types: added sanity checks to pointer routines; added data_shape optional... · 5fe220bd
  Sebastian Müller authored 2 months ago
```
opti_types: added sanity checks to pointer routines; added data_shape optional input to sim_data_add; added is_allocated to sim_var; added optional raise to sim_data_get_id
```
  5fe220bd
- changed 'stop' to error_messages · 4d924975
  Maren Kaluza authored 2 months ago and Sebastian Müller committed 2 months ago
  
  4d924975
- renamed ndim to data_shape and dim to ndim · b3a7d2df
  Maren Kaluza authored 2 months ago and Sebastian Müller committed 2 months ago
  
  b3a7d2df
- Set procedures in sim_data_t to private/public, +minor changes · 9bec1f8c
  Maren Kaluza authored 2 months ago and Sebastian Müller committed 2 months ago
  
  9bec1f8c
- replaced name(256) with name allocatable · 20fc5c1c
  Maren Kaluza authored 3 months ago and Sebastian Müller committed 2 months ago
  
  20fc5c1c
- model_dp doesn't loop over domains anymore · 0ea2d885
  Maren Kaluza authored 3 months ago and Sebastian Müller committed 2 months ago
```
model_dp is designed for one dimensional data_sim.
```
  0ea2d885
- optidata_sim_init now gets shape instead of dataObs as argument · 0764950f
  Maren Kaluza authored 3 months ago and Sebastian Müller committed 2 months ago
```
dataObs was only used to derive the shape of the opti data.
```
  0764950f
- Switched two arguments in a subroutine to follow conventions. · 2daf85c8
  Maren Kaluza authored 3 months ago and Sebastian Müller committed 2 months ago
  
  2daf85c8
- added 'sim_data_get_id' and reduced code repetition. · 951b767d
  Maren Kaluza authored 3 months ago and Sebastian Müller committed 2 months ago
  
  951b767d
- allocate of sim_data uses an array instead of integers for dimension now · 9ac428d1
  Maren Kaluza authored 3 months ago and Sebastian Müller committed 2 months ago
  
  9ac428d1
- opti types: rename procedures of sim_data_t to follow naming schema · b7d5839d
  Sebastian Müller authored 3 months ago
  
  b7d5839d
- opti types: renaming: opti_sim => sim_data, opti_sim_single => sim_var,... · 6fdbc0a4
  Sebastian Müller authored 3 months ago
```
opti types: renaming: opti_sim => sim_data, opti_sim_single => sim_var, opti_vars => variables, dimen => ndim
```
  6fdbc0a4
- Merge branch 'add_optimizee_type' into 'reduced_eval_interface' · 40da53be
  Sebastian Müller authored 3 months ago and Sebastian Müller committed 2 months ago
```
Add optimizee type

See merge request !83
```
  40da53be
- Added ToDos as comments · 3379b596
  Maren Kaluza authored 4 months ago and Sebastian Müller committed 2 months ago
  
  3379b596
- mcmc: refactor to use optimizee; update tests · 252684e9
  Sebastian Müller authored 4 months ago
  
  252684e9
- Updated to type opti_sim_t with flexible opti data, minimal eval interface · 3b5b887a
  Maren Kaluza authored 6 months ago and Sebastian Müller committed 2 months ago
```
New features:

- The eval interface only has two variables: config, opti_sim(:)
    - config contains the parameters. Also for example the indices. The
      indices are helpful for parallelization: the eval would only use a
      subset of domains.
    - opti_sim is an array that will have the size nDomains in case of
      mHM. Each element of the array is then of type opti_sim and can
      have an array of different optimization variables 'opti_vars'.
- opti_vars replaces specific optimization variables like runoff, et,
  tws, etc. You can add an arbitrary variable with dimension 1, 2, 3, 4
  or 5 with an arbitrary name like "runoff", "et", "tws", etc.
    - optivars contains:
      - add: to add a variable. -- This might not work correctly if
	a second variable is added and so on.
      - has: by name, if a variable is added.
      - allocate: allocates the space for the variable. It cannot be
	done through a pointer.
      - set_pointer: sets a pointer to the allocated space of a
	variable.

Changes of plan:

- The opti_sim input variable in the eval interface was not supposed to
  be an array. But it was more convenient that way and would have added
  another complicated layer to the code otherwise.
- There was not supposed to be a procedure 'allocate' in opti_sim. But
  int turned out, associated arrays cannot be allocated through
  pointers.
```
  3b5b887a
- DDS: use optimizee input; update tests · 62e6c6dd
  Sebastian Müller authored 4 months ago
  
  62e6c6dd
- created new type variables_optidata_sim for eval_interface · c1eb67ee
  Maren Kaluza authored 1 year ago and Sebastian Müller committed 2 months ago
```
instead of many optization variables the eval_interfaces now gets a
collection of all optimization variables in one. By this, the
eval_interface is short and does not need to be changed each time we
would like to optimize something else.
```
  c1eb67ee
- sce: remove unneeded imports · 1dfb9fb3
  Sebastian Müller authored 4 months ago
  
  1dfb9fb3
- test: update test_mo_anneal to use optimizee input · 6da312a4
  Sebastian Müller authored 4 months ago
  
  6da312a4
- anneal: update optimizer to use optimizee · bec1d851
  Sebastian Müller authored 4 months ago
  
  bec1d851
- tests: update SCE unit test · 490999a1
  Sebastian Müller authored 4 months ago
  
  490999a1
- SCE: use optimizee as input for the objective function · de25b7ac
  Sebastian Müller authored 4 months ago
  
  de25b7ac
- opti: add optimizee abstract type and implementations for functinos and likelihoods · 3adfa19f
  Sebastian Müller authored 4 months ago
  
  3adfa19f

Rework optimization routines

Merge request reports

Activity