Changes

Summary

1. example on Gumbel 1d (details)
2. started with the Helmholtz equation (details)
3. integrated SING into TransportMaps (details)
4. cleaning SING files (details)
5. l1 tests and plotting scripts (details)
6. working on Helmholtz example from Ghattas' paper (details)
7. working on Helmholtz (details)
8. starting to work on actual probabilistic model (details)
9. renaming some sun_sing files (details)
10. updating run_sing files to run with SI.SING (details)
11. actually updating to run with SI.SING (ignore last commit) (details)
12. testing l1 (details)
13. l1 regularization (details)
14. Updated regularized map - SING now runs with 2 types of reg and no reg. (details)
15. relaxed rademacher, butterfly distribution, and more properties of (details)
16. relaxed rademacher uses uniform over (-1,1), not (0,1) (details)
17. node ordering fix (details)
18. changing parameters in Helmholtz (details)
19. added stub solver for EIT problem (details)
20. working on surface gradient (details)
21. working on CEM (details)
22. fixed EIT working on CEM (details)
23. plotting scripts for nips (details)
24. turned hessian off in sing (details)
25. mc studies 1step algs (details)
26. More MC Studies for SING (details)
27. working on models (details)
28. No electrodes model works (details)
29. working on discretized model (details)
30. working on discretized model (details)
31. gmsh mesh generation for discretized test (details)
32. working on complete model (details)
33. starting to work on adjoint equations (details)
34. messed up with no-electrodes model.. (details)
35. working on fast sensor function (details)
36. fixed sensor expression (details)
37. working on Dirichlet to Neumann model (details)
38. working on adjoint for no electrode model (details)
39. adjoint done (details)
40. gradient and Hessian working (details)
41. setting up the posterior distribution (details)
42. working on likelihood functions (details)
43. working on posterior distribution (details)
44. storage/loading of distribution works (details)
45. working on action of Hessian (details)
46. working on action_hess_a for transport map distributions (details)
47. working on minimization of kl divergence using action of the Hessian (details)
48. working on kl minimization tests (details)
49. unit tests are passing. todo: caching in Likelihood to avoid extra (details)
50. the distribution object is now pre-broadcasted to the children in (details)
51. fixed test_kl_divergence (details)
52. working on caching of Likelihood evaluations (details)
53. fixed init function (details)
54. The action of the Hessian is ready. Fixes #31. Caching issues reserved (details)
55. introducing cached and counted decorators (details)
56. adding counting output (details)
57. caching to be finished in TransportMaps and Maps in general. Then I need (details)
58. still working on maps... (details)
59. working on distributions (details)
60. testing caching. todo: caching in optimization problems (details)
61. added caching and counting wherever needed (maybe missing some spots). (details)
62. working on caching in Hessian. Todo: update and run unittests (details)
63. adding reset counters functionalities (details)
64. counting and resetting works. Caching also seems to work. TODO: (details)
65. caching does what is supposed to do! (details)
66. added manual test for caching and counting (details)
67. working on caching minimization routine (details)
68. working on unit tests for kl-divergence (details)
69. working on test_kl_divergence.py (details)
70. fixing  an error at a time... (details)
71. kl_min tests work! (details)
72. fixing all unit tests (details)
73. porting MPI implementation to mpi_map v2.0 (details)
74. porting to mpi_map v2.0 completed. ToBeTested!! (details)
75. mpi_map v2 integrated and passing unit tests (details)
76. gathering of ncalls and nevals count at the end of optimization done (details)
77. passes unit tests (details)
78. now pointing to MPI_Pool_v2 in mpi_map v2.0 (details)
79. returning latest cache within log dictionary (details)
80. fixed py2/py3 issue with cached decorator (details)
81. fixed py2/py3 issue with cached decorator (details)
82. fixed py2/py3 issue with cached decorator (details)
83. fixing py2/py3 handling of *args and **kwargs (PEP3102) (details)
84. working on target distribution (details)
85. working on mpi evaluation of target density (details)
86. working on constant field data generation and parallel implementation (details)
87. working on speeding up function and gradient evaluatiions (details)
88. implemented fungrad (details)
89. working of action of Hessian (details)
90. mpi implementation not working... (details)
91. staring solving simple problems.. (details)
92. working on constant field problem and parallelization (details)
93. working on the problem set up (details)
94. data generation for constant field works (details)
95. working on adjoint for field (details)
96. fixed gradient! (details)
97. working on Hessian of EIT problem (details)
98. working on hessian (details)
99. fixed counter reset in some classes (details)
100. fixing regularization (details)
101. fixed caching for tuple_grad_a (details)
102. fixed some issues with tests and new counting decorators (details)
103. action of the Hessian for EIT works. To be integrated in distribution (details)
104. eit hessian works with caching of solution and adjoints (details)
105. EIT field distribution works with MPI (details)
106. added possibility to add real fields by expression (details)
107. disabled plotting if not available (details)
108. distributions work on cluster (details)
109. made EIT solver into a class (details)
110. working on validators (details)
111. working on saa (details)
112. implementing gradient based validators (details)
113. validators... (details)
114. working on adaptivity algorithm (details)
115. working on the adaptivity algorithm (details)
116. adding adaptivity options to tmap-tm (details)
117. working on launching script tmap-tm (details)
118. changing builder classes to imbed parameters and take distributions on (details)
119. going through banana example (details)
120. working on adaptivity on banana (details)
121. adaptivity on banana almost work (details)
122. adaptivity works on banana. a lot to do yet... (details)
123. improving parallelization (details)
124. splitted first variation computation for better parallelization (details)
125. adaptivity works on banana without hyper (details)
126. working on low-rank approximation of Hessians (details)
127. working on laplace approximation for Bayesian inverse problems, where (details)
128. working on low-rank approximation of covariance in bayesian problem (details)
129. working on low-rank approximation (details)
130. added logging information in postprocessing script (details)
131. low-rank Laplace approximation works (details)
132. applying Laplace to EIT (details)
133. working on composition of Likelihood with maps (details)
134. working on composition of Likelihood with maps (details)
135. composition of functions and map now is possible. working on laplace of (details)
136. working on laplace approximation of EIT (details)
137. debugging action of Hessian (details)
138. checking hessian action.. (details)
139. working on finite difference test for Hessian (details)
140. checking Hessian for distribution (details)
141. fixed withening script. Now action of Hessian works with MPI (details)
142. working on EIT adaptivity (details)
143. added missing functions to LinearTransportMap. Added unit tests for it (details)
144. modified caching of EIT solutions so that the cache is now pickable. (details)
145. added verbosity option in adaptivity algorithms and fixed bug in action (details)
146. added intermediate storage capability for adaptivity (details)
147. added output for lower and upper bounds in saa (details)
148. fixed bug in tmap-tm and in test_distribution.py for EIT (details)
149. added distributed sampling for MC sampling in minimize_kl_divergence (details)
150. fixed sampling for inverse maps from samples (disabled distributed (details)
151. fixed distributed sampling for linear span kl minimization (details)
152. fixed action Hessian with regularization (details)
153. simplified regularization of action hessian regularization (details)
154. fixed relative tolerance in validator (details)
155. fixed relative tolerance in validator (details)
156. remove unnecessary output form EIT (details)
157. changed active variable indicator to E[|x|] (details)
158. tryna generate non-gaussian data (details)
159. working on reload feature for tmap-tm (details)
160. reloading done for first variation adaptivity. to be checked in other (details)
161. added some output to First variation algorithm (details)
162. fixed restarting and grad_x for target in EIT (details)
163. added field full_basis_list to function approximations, to be used in (details)
164. fixed adaptivity accordingly to previous full_basis_list at the function (details)
165. running aerostruct and copula data (details)
166. committing data samples for copula data (details)
167. fixed first variation computation (details)
168. added plot sample for post process in EIT (details)
169. added a non-gaussian chain dataset (details)
170. added full_basis_list feature to xml map structure loading (details)
171. fixed full_basis_list to save in memory allocation (details)
172. minor testing datasets changes (details)
173. check for product distribution (details)
174. removed matplotlib from imports (details)
175. scikit-sparse reqs (details)
176. changed googleanalytics requirements (details)
177. starting jupyter notebook for sing (details)
178. looking at copula data inverse covariance (details)
179. working on coarser test case (details)
180. fixed several bugs in adaptivity and tests (details)
181. fixed get state in first variation builder (details)
182. added line search options in tmap-tm (details)
183. fixed kladaptivity serialization and regression regularization (details)
184. fixed gx_gt_kl batched evaluation (details)
185. added sing notebook example (details)
186. fixing adaptivity here and there.. (details)
187. integrated documentation for SING algorithm (details)
188. added postprocessing functionalities for adaptivity (details)
189. added double circle phantom (details)
190. fixed store function in tmap-tm (details)
191. added argument for line search delta (details)
192. added delta parameter for line search (details)
193. fixed documentation SING (details)
194. fixing unit tests; (details)
195. fixed laplace approximation starting point (details)
196. documenting fv adaptivity (details)
197. preparing release v2.0b0 (details)
198. v2.0b0: ready for unit-testing (details)
199. fixed matplotlib dependence (details)
200. fixed process_time PyV3 PyV2 (details)
201. fixed processtime name (details)
202. fixing bugs for release (details)
203. working to fix grax_gradt_kl (details)
204. fixed grad_x of first variation (details)
205. subtract identity_coeffs + l1_reg to component kl (details)
206. removed dependency on FileNotFoundError for compatibility with py2 (details)
207. fixed compatibility issue with py2 (details)
208. removed unnecessary folders in Examples and updated mirroring script (details)
209. added algorithms documentation section (details)
210. updated script for public mirroring (details)
211. throwing an error if ders!=[0,1] used in decoupled LinearSpan min_kl (details)

example on Gumbel 1d

started with the Helmholtz equation

integrated SING into TransportMaps

cleaning SING files

l1 tests and plotting scripts

working on Helmholtz example from Ghattas' paper

working on Helmholtz

starting to work on actual probabilistic model

renaming some sun_sing files

updating run_sing files to run with SI.SING

actually updating to run with SI.SING (ignore last commit)

testing l1

l1 regularization

Updated regularized map - SING now runs with 2 types of reg and no reg.

relaxed rademacher, butterfly distribution, and more properties of
gaussian graphs

relaxed rademacher uses uniform over (-1,1), not (0,1)

node ordering fix

changing parameters in Helmholtz

added stub solver for EIT problem

working on surface gradient

working on CEM

fixed EIT working on CEM

plotting scripts for nips

turned hessian off in sing

mc studies 1step algs

More MC Studies for SING

working on models

No electrodes model works

working on discretized model

working on discretized model

gmsh mesh generation for discretized test

working on complete model

starting to work on adjoint equations

messed up with no-electrodes model..

working on fast sensor function

fixed sensor expression

working on Dirichlet to Neumann model

working on adjoint for no electrode model

adjoint done

gradient and Hessian working

setting up the posterior distribution

working on likelihood functions

working on posterior distribution

storage/loading of distribution works

working on action of Hessian

working on action_hess_a for transport map distributions

working on minimization of kl divergence using action of the Hessian

working on kl minimization tests

unit tests are passing. todo: caching in Likelihood to avoid extra
function and gradient evaluations

the distribution object is now pre-broadcasted to the children in
kl-minimization. Fixes #82

fixed test_kl_divergence

working on caching of Likelihood evaluations

fixed init function

The action of the Hessian is ready. Fixes #31. Caching issues reserved
to #83.

introducing cached and counted decorators

adding counting output

caching to be finished in TransportMaps and Maps in general. Then I need
to move to caching for distributions

still working on maps...

working on distributions

testing caching. todo: caching in optimization problems

added caching and counting wherever needed (maybe missing some spots).
To be extensively tested...

working on caching in Hessian. Todo: update and run unittests

adding reset counters functionalities

counting and resetting works. Caching also seems to work. TODO:
Integrate into the optimization routines

caching does what is supposed to do!

added manual test for caching and counting

working on caching minimization routine

working on unit tests for kl-divergence

working on test_kl_divergence.py

fixing  an error at a time...

kl_min tests work!

fixing all unit tests

porting MPI implementation to mpi_map v2.0

porting to mpi_map v2.0 completed. ToBeTested!!

mpi_map v2 integrated and passing unit tests

gathering of ncalls and nevals count at the end of optimization done

passes unit tests

now pointing to MPI_Pool_v2 in mpi_map v2.0

returning latest cache within log dictionary

fixed py2/py3 issue with cached decorator

fixed py2/py3 issue with cached decorator

fixed py2/py3 issue with cached decorator

fixing py2/py3 handling of *args and **kwargs (PEP3102)

working on target distribution

working on mpi evaluation of target density

working on constant field data generation and parallel implementation

working on speeding up function and gradient evaluatiions

implemented fungrad

working of action of Hessian

mpi implementation not working...

staring solving simple problems..

working on constant field problem and parallelization

working on the problem set up

data generation for constant field works

working on adjoint for field

fixed gradient!

working on Hessian of EIT problem

working on hessian

fixed counter reset in some classes

fixing regularization

fixed caching for tuple_grad_a

fixed some issues with tests and new counting decorators

action of the Hessian for EIT works. To be integrated in distribution
definition

eit hessian works with caching of solution and adjoints

EIT field distribution works with MPI

added possibility to add real fields by expression

disabled plotting if not available

distributions work on cluster

made EIT solver into a class

working on validators

working on saa

implementing gradient based validators

validators...

working on adaptivity algorithm

working on the adaptivity algorithm

adding adaptivity options to tmap-tm

working on launching script tmap-tm

changing builder classes to imbed parameters and take distributions on
solve

going through banana example

working on adaptivity on banana

adaptivity on banana almost work

adaptivity works on banana. a lot to do yet...

improving parallelization

splitted first variation computation for better parallelization

adaptivity works on banana without hyper

working on low-rank approximation of Hessians

working on laplace approximation for Bayesian inverse problems, where
one has access only on the action of the Hessian

working on low-rank approximation of covariance in bayesian problem