Merge remote-tracking branch 'origin/update.stepsize' into update

doc/en/source/algorithm/exchange.rst

@@ -33,15 +33,6 @@ Otherwise, continuous space is used.
     Description: Initial value of parameters.
     If not defined, these will be initialize randomly.
 
-  - ``unit_list``
-
-    Format: List of float. Length should be equal to ``dimension``.
-
-    Description:
-    Unit length of each parameter.
-    ``Algorithm`` makes parameters dimensionless and normalized by dividing these by ``unit_list``.
-    If not defined, each component will be 1.0.
-
   - ``min_list``
 
     Format: List of float. Length should be equal to ``dimension``.
@@ -61,6 +52,13 @@ Otherwise, continuous space is used.
     When a parameter exceeds this value during the Monte Carlo search,
     the solver is not evaluated and the value is considered infinite.
 
+  - ``step_list``
+
+    Format: List of float. Length should be equal to ``dimension``.
+
+    Description:
+    The step length in one Monte Carlo update (deviation of Gaussian distribution).
+
 - Discrete space
 
   - ``mesh_path``

doc/en/source/algorithm/pamc.rst

@@ -34,15 +34,6 @@ Otherwise, continuous space is used.
     Initial value of parameters.
     If not defined, these will be initialize randomly.
 
-  - ``unit_list``
-
-    Format: List of float. Length should be equal to ``dimension``.
-
-    Description:
-    Unit length of each parameter.
-    ``Algorithm`` makes parameters dimensionless and normalized by dividing these by ``unit_list``.
-    If not defined, each component will be 1.0.
-
   - ``min_list``
 
     Format: List of float. Length should be equal to ``dimension``.
@@ -61,6 +52,13 @@ Otherwise, continuous space is used.
     When a parameter exceeds this value during the Monte Carlo search,
     the solver is not evaluated and the value is considered infinite.
 
+  - ``step_list``
+
+    Format: List of float. Length should be equal to ``dimension``.
+
+    Description:
+    The step length in one Monte Carlo update (deviation of Gaussian distribution).
+
 - Discrete space
 
   - ``mesh_path``

doc/en/source/tutorial/exchange.rst

@@ -41,8 +41,7 @@ For details, see the input file section of the manual.
   [algorithm.param]
   min_list = [3.0, 3.0]
   max_list = [6.0, 6.0]
-  initial_list = [0.0, 0.0]
-  unit_list = [0.3, 0.3]
+  step_list = [0.3, 0.3]
 
   [algorithm.exchange]
   Tmin = 0.1
@@ -67,9 +66,7 @@ The contents of ``[base]``, ``[solver]``, and ``[runner]`` sections are the same
 
 - ``min_list`` is a lower bound and ``max_list`` is an upper bound.
 
-- ``initial_list`` is an initial value.
-
-- ``unit_list`` is the unit of each parameter.
+- ``unit_list`` is step length in one Monte Carlo update (deviation of Gaussian distribution).
 
 ``[algorithm.exchange]`` section sets the hyper parameters for RXMC.
 

doc/en/source/tutorial/pamc.rst

@@ -49,7 +49,7 @@ For details, see the input file section of the manual.
     [algorithm.param]
     max_list = [6.0, 6.0]
     min_list = [-6.0, -6.0]
-    unit_list = [0.3, 0.3]
+    step_list = [0.3, 0.3]
 
     [algorithm.pamc]
     bmin = 0.0
@@ -74,7 +74,7 @@ The contents of ``[base]``, ``[solver]``, and ``[runner]`` sections are the same
 
 - ``min_list`` is a lower bound and ``max_list`` is an upper bound.
 
-- ``unit_list`` is step length in one MC update (deviation of Gaussian distribution).
+- ``unit_list`` is step length in one Monte Carlo update (deviation of Gaussian distribution).
 
 ``[algorithm.pamc]`` section sets the parameters for PAMC.
 

doc/ja/source/algorithm/exchange.rst

@@ -31,15 +31,6 @@ MPI並列を利用する場合は、あらかじめ `mpi4py <https://mpi4py.read
 
     説明: パラメータの初期値。 定義しなかった場合は一様ランダムに初期化されます。
 
-  - ``unit_list``
-
-    形式: 実数型のリスト。長さはdimensionの値と一致させます。
-
-    説明: 各パラメータの単位。
-          探索アルゴリズム中では、各パラメータをそれぞれこれらの値で割ることで、
-          簡易的な無次元化・正規化を行います。
-          定義しなかった場合にはすべての次元で 1.0 となります。
-
   - ``min_list``
 
     形式: 実数型のリスト。長さはdimensionの値と一致させます。
@@ -56,6 +47,12 @@ MPI並列を利用する場合は、あらかじめ `mpi4py <https://mpi4py.read
           モンテカルロ探索中にこの値を上回るパラメータが出現した場合、
           ソルバーは評価されずに、値が無限大だとみなされます。
 
+  - ``step_list``
+
+    形式: 実数型のリスト。長さはdimensionの値と一致させます。
+
+    説明: モンテカルロ更新の際の変化幅(ガウス分布の偏差)です。
+
 - 離散空間
 
   - ``mesh_path``

doc/ja/source/algorithm/pamc.rst

@@ -31,15 +31,6 @@ MPI 並列をする場合にはあらかじめ `mpi4py <https://mpi4py.readthedo
 
     説明: パラメータの初期値。 定義しなかった場合は一様ランダムに初期化されます。
 
-  - ``unit_list``
-
-    形式: 実数型のリスト。長さはdimensionの値と一致させます。
-
-    説明: 各パラメータの単位。
-          探索アルゴリズム中では、各パラメータをそれぞれこれらの値で割ることで、
-          簡易的な無次元化・正規化を行います。
-          定義しなかった場合にはすべての次元で 1.0 となります。
-
   - ``min_list``
 
     形式: 実数型のリスト。長さはdimensionの値と一致させます。
@@ -56,6 +47,12 @@ MPI 並列をする場合にはあらかじめ `mpi4py <https://mpi4py.readthedo
           モンテカルロ探索中にこの値を上回るパラメータが出現した場合、
           ソルバーは評価されずに、値が無限大だとみなされます。
 
+  - ``step_list``
+
+    形式: 実数型のリスト。長さはdimensionの値と一致させます。
+
+    説明: モンテカルロ更新の際の変化幅(ガウス分布の偏差)です。
+
 - 離散空間
 
   - ``mesh_path``

doc/ja/source/tutorial/exchange.rst

@@ -42,8 +42,7 @@
   [algorithm.param]
   min_list = [3.0, 3.0]
   max_list = [6.0, 6.0]
-  initial_list = [0.0, 0.0]
-  unit_list = [0.3, 0.3]
+  step_list = [0.3, 0.3]
 
   [algorithm.exchange]
   Tmin = 0.1
@@ -66,9 +65,7 @@
 
 - ``min_list`` と ``max_list`` はそれぞれ探索範囲の最小値と最大値を指定します。
 
-- ``initial_list`` は初期値を指定します。
-
-- ``unit_list`` は各パラメータの単位を指定します。
+- ``step_list`` はモンテカルロ更新の際の変化幅(ガウス分布の偏差)です。
 
 ``[algorithm.exchange]`` サブセクションは、レプリカ交換モンテカルロ法のハイパーパラメータを指定します。
 

doc/ja/source/tutorial/pamc.rst

@@ -45,7 +45,7 @@
     [algorithm.param]
     max_list = [6.0, 6.0]
     min_list = [-6.0, -6.0]
-    unit_list = [0.3, 0.3]
+    step_list = [0.3, 0.3]
 
     [algorithm.pamc]
     bmin = 0.0
@@ -69,7 +69,7 @@
 
 - ``min_list`` と ``max_list`` はそれぞれ探索範囲の最小値と最大値を指定します。
 
-- ``unit_list`` はモンテカルロ更新の際の変化幅(ガウス分布の偏差)です。
+- ``step_list`` はモンテカルロ更新の際の変化幅(ガウス分布の偏差)です。
 
 ``[algorithm.pamc]`` セクションは、ポピュレーションアニーリングのハイパーパラメータを指定します。
 

extra/sim-trhepd-rheed/sample/single_beam/exchange/input.toml

@@ -10,6 +10,7 @@ seed = 12345
 [algorithm.param]
 min_list = [3.0, 3.0]
 max_list = [6.0, 6.0]
+step_list = [1.0, 1.0]
 
 [algorithm.exchange]
 numsteps = 1000

extra/sim-trhepd-rheed/sample/single_beam/pamc/input.toml

@@ -27,7 +27,7 @@ seed = 12345
 [algorithm.param]
 min_list = [3.0, 3.0]
 max_list = [6.0, 6.0]
-unit_list = [0.3, 0.3]
+step_list = [0.3, 0.3]
 
 [algorithm.pamc]
 numsteps_annealing = 5

extra/sim-trhepd-rheed/tests/single_beam/exchange/input.toml

@@ -10,6 +10,7 @@ seed = 12345
 [algorithm.param]
 min_list = [3.0, 3.0]
 max_list = [6.0, 6.0]
+step_list = [1.0, 1.0]
 
 [algorithm.exchange]
 numsteps = 1000

extra/sim-trhepd-rheed/tests/single_beam/pamc/input.toml

@@ -28,7 +28,7 @@ seed = 12345
 [algorithm.param]
 min_list = [3.0, 3.0]
 max_list = [6.0, 6.0]
-unit_list = [0.3, 0.3]
+step_list = [0.3, 0.3]
 
 [algorithm.pamc]
 numsteps_annealing = 5

sample/analytical/exchange/input.toml

@@ -13,8 +13,7 @@ seed = 12345
 [algorithm.param]
 max_list = [6.0, 6.0]
 min_list = [-6.0, -6.0]
-initial_list = [0.0, 0.0]
-unit_list = [0.3, 0.3]
+step_list = [0.3, 0.3]
 
 [algorithm.exchange]
 Tmin = 0.01

sample/analytical/pamc/input.toml

@@ -17,7 +17,7 @@ seed = 12345
 [algorithm.param]
 max_list = [6.0, 6.0]
 min_list = [-6.0, -6.0]
-unit_list = [0.3, 0.3]
+step_list = [0.3, 0.3]
 
 [algorithm.pamc]
 bmin = 0.0

src/py2dmat/algorithm/exchange.py

@@ -61,7 +61,8 @@ class Algorithm(py2dmat.algorithm.montecarlo.AlgorithmBase):
     x: np.ndarray
     xmin: np.ndarray
     xmax: np.ndarray
-    xunit: np.ndarray
+    #xunit: np.ndarray
+    xstep: np.ndarray
 
     numsteps: int
     numsteps_exchange: int

src/py2dmat/algorithm/montecarlo.py

@@ -68,7 +68,7 @@ class AlgorithmBase(py2dmat.algorithm.AlgorithmBase):
     x: np.ndarray
     xmin: np.ndarray
     xmax: np.ndarray
-    xunit: np.ndarray
+    xstep: np.ndarray
 
     # discrete problem
     inode: np.ndarray
@@ -115,16 +115,26 @@ def __init__(self, info: py2dmat.Info,
             if "mesh_path" in info_param:
                 self.iscontinuous = False
                 self.domain = py2dmat.domain.MeshGrid(info)
-
             else:
                 self.iscontinuous = True
                 self.domain = py2dmat.domain.Region(info)
 
         if self.iscontinuous:
             self.xmin = self.domain.min_list
             self.xmax = self.domain.max_list
-            self.xunit = self.domain.unit_list
 
+            if "step_list" in info_param:
+                self.xstep = info_param.get("step_list")
+
+            elif "unit_list" in info_param:
+                # for compatibility, unit_list can also be accepted for step size.
+                if self.mpirank == 0:
+                    print("WARNING: unit_list is obsolete. use step_list instead")
+                self.xstep = info_param.get("unit_list")
+            else:
+                # neither step_list nor unit_list is specified, report error.
+                # default value not assumed.
+                raise ValueError("ERROR: algorithm.param.step_list not specified")
         else:
             self.node_coordinates = np.array(self.domain.grid)[:, 1:]
             self.nnodes = self.node_coordinates.shape[0]
@@ -213,7 +223,7 @@ def propose(self, current: np.ndarray) -> np.ndarray:
             proposal
         """
         if self.iscontinuous:
-            dx = self.rng.normal(size=(self.nwalkers, self.dimension)) * self.xunit
+            dx = self.rng.normal(size=(self.nwalkers, self.dimension)) * self.xstep
             proposed = current + dx
         else:
             proposed_list = [self.rng.choice(self.neighbor_list[i]) for i in current]

src/py2dmat/algorithm/pamc.py

@@ -65,7 +65,8 @@ class Algorithm(py2dmat.algorithm.montecarlo.AlgorithmBase):
     x: np.ndarray
     xmin: np.ndarray
     xmax: np.ndarray
-    xunit: np.ndarray
+    #xunit: np.ndarray
+    xstep: np.ndarray
 
     numsteps: int
     numsteps_annealing: int

tests/bayes/do.sh

@@ -1,3 +1,5 @@
+#!/bin/sh
+
 rm -rf output
 
 time python3 ../../src/py2dmat_main.py input.toml

tests/exchange/do.sh

@@ -1,3 +1,5 @@
+#!/bin/sh
+
 rm -rf output
 
 time mpiexec --oversubscribe -np 4 python3 ../../src/py2dmat_main.py input.toml

tests/exchange/input.toml

@@ -9,6 +9,7 @@ seed = 12345
 [algorithm.param]
 min_list = [-5.0, -5.0]
 max_list = [5.0, 5.0]
+step_list = [1.0, 1.0]
 
 [algorithm.exchange]
 numsteps = 10000

tests/exchange_continue/do.sh

@@ -1,6 +1,6 @@
 #!/bin/sh
 
-CMD="mpiexec -np 4 python3 -u ../../src/py2dmat_main.py"
+CMD="mpiexec --oversubscribe -np 4 python3 -u ../../src/py2dmat_main.py"
 
 rm -rf output1
 

tests/exchange_continue/input1a.toml

@@ -11,6 +11,7 @@ checkpoint_steps = 1000
 [algorithm.param]
 min_list = [-5.0, -5.0]
 max_list = [5.0, 5.0]
+step_list = [1.0, 1.0]
 
 [algorithm.exchange]
 numsteps = 10000

tests/exchange_continue/input1b.toml

@@ -11,6 +11,7 @@ checkpoint_steps = 5000
 [algorithm.param]
 min_list = [-5.0, -5.0]
 max_list = [5.0, 5.0]
+step_list = [1.0, 1.0]
 
 [algorithm.exchange]
 numsteps = 20000

tests/exchange_continue/input2.toml

@@ -11,6 +11,7 @@ checkpoint_steps = 1000
 [algorithm.param]
 min_list = [-5.0, -5.0]
 max_list = [5.0, 5.0]
+step_list = [1.0, 1.0]
 
 [algorithm.exchange]
 numsteps = 20000

tests/exchange_mesh/do.sh

@@ -1,3 +1,5 @@
+#!/bin/sh
+
 rm -rf output
 
 echo generate MeshData.txt

tests/mapper/do.sh

@@ -1,6 +1,8 @@
+#!/bin/sh
+
 rm -rf output
 
-time mpiexec -np 2 python3 ../../src/py2dmat_main.py input.toml
+time mpiexec --oversubscribe -np 2 python3 ../../src/py2dmat_main.py input.toml
 
 resfile=output/ColorMap.txt
 

tests/minsearch/do.sh

@@ -1,3 +1,5 @@
+#!/bin/sh
+
 rm -rf output
 
 time python3 ../../src/py2dmat_main.py input.toml

tests/pamc/do.sh

@@ -1,3 +1,5 @@
+#!/bin/sh
+
 rm -rf output
 
 time mpiexec --oversubscribe -np 2 python3 -m mpi4py ../../src/py2dmat_main.py input.toml