surface-construct 0.8__tar.gz → 0.8.2__tar.gz

{surface_construct-0.8/surface_construct.egg-info → surface_construct-0.8.2}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: surface_construct
-Version: 0.8
+Version: 0.8.2
 Summary: Surface termination construction especially for complex model, such as oxides or carbides.
 Home-page: https://gitee.com/pjren/surface_construct/
 Author: ren

{surface_construct-0.8 → surface_construct-0.8.2}/setup.py

@@ -18,7 +18,7 @@ install_requires = [
 
 setup(
     name='surface_construct',
-    version='0.8',
+    version='0.8.2',
     packages=['surface_construct'],
     url='https://gitee.com/pjren/surface_construct/',
     license='GPL',

{surface_construct-0.8 → surface_construct-0.8.2}/surface_construct/sampling.py

@@ -1,284 +1,318 @@
-"""
-TODO: 关键点采样：top 位置、hollow位，bridge 位等等。
-"""
-
-import numpy as np
-from ase.geometry import get_distances
-from scipy.spatial import ConvexHull
-from scipy.spatial.distance import cdist
-from sklearn.cluster import KMeans as Cluster
-import random
-
-
-def addition_samples(sg_obj, size=None, probability=None, **kwargs):
-    if 'seed' in kwargs:
-        seed = kwargs['seed']
-    else:
-        seed = None
-    if probability is None:
-        probability = {
-            "max_sigma": 0.2,  # 采样方法的概率
-            "max_diversity": 0.8,
-        }
-    if size is None:
-        size = 1
-
-    # 归一化
-    total = sum(probability.values())
-    if total != 1.0:
-        probability = {k: v / total for k, v in probability.items()}
-
-    rng = np.random.default_rng(seed)
-    method_list = rng.choice(list(probability.keys()), size=size, p=list(probability.values()))
-
-    point_idx = np.array([], dtype=int)
-    for method in method_list:
-        method_lower = method.lower()
-        if method_lower == 'max_sigma':
-            sampling_obj = MaxSigmaSampling(sg_obj)
-        elif method_lower == 'max_diversity':
-            sampling_obj = MaxDiversitySampling(sg_obj)
-        else:
-            raise NotImplementedError
-        point_idx = np.concatenate([point_idx, sampling_obj.samples(size=1, **kwargs)])  # 每种方法只采一个
-
-    return point_idx
-
-
-class SamplingBase:
-    def __init__(self, sg_obj, **kwargs):
-        self.sg_obj = sg_obj
-
-    @property
-    def _pop_size(self):
-        return len(self.sg_obj.points)
-
-    @property
-    def _population(self):
-        """
-        默认的全体是 sg_obj.points 的 index
-        :return:
-        """
-        return range(self._pop_size)
-
-    def _append_sample_to_sg(self, point_idx=None):
-        """
-        将采样点加入到 sg_obj.sample_points 和相应的 vector
-
-        :return:
-        """
-        if point_idx is not None:
-            if self.sg_obj.sample_idx is not None:
-                self.sg_obj.sample_idx = np.concatenate([self.sg_obj.sample_idx, point_idx])
-                self.sg_obj._sample_vector = np.concatenate([self.sg_obj._sample_vector, self.sg_obj.vector[point_idx]])
-                self.sg_obj.sample_points = np.concatenate([self.sg_obj.sample_points, self.sg_obj.points[point_idx]])
-            else:
-                self.sg_obj.sample_idx = np.array(point_idx)
-                self.sg_obj._sample_vector = self.sg_obj.vector[point_idx]
-                self.sg_obj.sample_points = self.sg_obj.points[point_idx]
-
-    def _samples(self, size, **kwargs):
-        raise NotImplementedError
-
-    def samples(self, size=1, **kwargs):
-        point_idx = self._samples(size=size, **kwargs)
-        self._append_sample_to_sg(point_idx=point_idx)
-
-        return point_idx
-
-
-class KeyPointSampling(SamplingBase):
-    """
-    基本思路：先定位 hollow位，再根据图论分析定位 bridge 位，最后定位top位。第二步，筛选等价位点。
-    需要一个基础func，从xy坐标，找到对应的格点。
-    """
-    def _samples(self, size, **kwargs):
-        pass
-
-
-class RandomSampling(SamplingBase):
-    """
-    完全随机的选择点，仅用于测试，效率太低。
-    """
-
-    def __init__(self, sg_obj, **kwargs):
-        super().__init__(sg_obj, **kwargs)
-        if 'seed' in kwargs:
-            self.seed = kwargs['seed']
-        else:
-            self.seed = None
-
-    def _samples(self, size, **kwargs):
-        rng = np.random.default_rng(self.seed)
-        pop_idx = rng.choice(self._population, size=size)
-        return pop_idx
-
-
-class MaxSigmaSampling(SamplingBase):
-    """
-    对最大误差的点进行采样
-    """
-
-    def _samples(self, size, **kwargs):
-        if 'energy' in self.sg_obj.grid_property:
-            # 如果已经读入了一些能量，则返回误差最大的点
-            idx = self.sg_obj.grid_property_sigma['energy'].argmax()
-            return [idx]
-        else:
-            raise "No energy for all population, pls do initial sampling first!"
-
-
-class InitialSampling(SamplingBase):
-    """
-    使用聚类-分层采样进行初始采样
-    """
-
-    def _samples(self, size, **kwargs):
-        # 进行分类，然后采样。这里使用 Kmeans 方法进行聚类
-        clusters = Cluster(n_clusters=size, random_state=0).fit(self.sg_obj.vector)
-        self.sg_obj._clusters = clusters  # 保存用于作图
-        # 对于每一类取距离 cluster 中心最小的点的 idx
-        centers = clusters.cluster_centers_
-        center_dist = cdist(centers, self.sg_obj.vector)  # 计算每个点到中心的距离
-        point_idx = np.argmin(center_dist, axis=-1)
-
-        # 找到 Hull 点，输出 hull 的个数
-        # TODO: 有了 keypoint 采样，这个就是不需要的
-        hull = ConvexHull(self.sg_obj.vector)
-        vertices = hull.vertices
-        points_hull_sample = self.sg_obj.points[vertices]
-
-        # 计算 hull 点与 cluster_sample 的距离, 排除太近的 cluster 采样点
-        hull_cluster_dist_array, hull_cluster_dist = get_distances(points_hull_sample,
-                                                                   self.sg_obj.points[point_idx],
-                                                                   cell=self.sg_obj.atoms.cell,
-                                                                   pbc=self.sg_obj.atoms.pbc)
-        cluster_index = vertices[np.all(hull_cluster_dist > self.sg_obj.interval * 2, axis=-1)]
-        print("There are {:d} vertex according to ConvexHull analysis, which can be appended to the initial "
-              "sampling set for better diversity.".format(len(cluster_index)))
-        # 加入采样
-        if 'include_vertex' in kwargs and kwargs['include_vertex']:
-            print("The vertex are appended into initial sampling.")
-            point_idx = np.concatenate([point_idx, cluster_index])
-        else:
-            print("The vertex weren't appended into initial sampling.")
-
-        return point_idx
-
-    def _append_sample_to_sg(self, point_idx=None):
-        """
-        将采样点加入到 sg_obj.sample_points 和相应的 vector
-        :return:
-        """
-        if point_idx is not None:
-            self.sg_obj.sample_idx = np.asarray(point_idx)
-            self.sg_obj._sample_vector = self.sg_obj.vector[point_idx]
-            self.sg_obj.sample_points = self.sg_obj.points[point_idx]
-
-
-class MaxDiversitySampling(SamplingBase):
-    """
-    对当前采样结构差异最大的点进行采样
-    基本思路是这样的：
-        * 重新进行聚类，
-        * 判断已经采样点属于的类别，找出没有点的类别，空类
-        * 如果空类不止一个，比较这些空类中心与旧点的距离，选择距离最大的点。
-    """
-
-    def _samples(self, size, center=False, **kwargs):
-        """
-
-        :param size:
-        :param center: 是否取中心。如果不是，则取能量最小值的点。如果没有能量则报错。
-        :param kwargs:
-        :return:
-        """
-        # 判断是否有过往的采样点，如果没有，调用 InitialSampling
-        if self.sg_obj.sample_idx is None:
-            raise "Please add initial samples (e.g. InitialSampling) before invoke this method!"
-        cluster_size = len(self.sg_obj.sample_idx) + size
-        nvirgin = 0
-        larger_clusters = None
-        larger_virgin = None
-        virgin = None
-        clusters = None
-        # 如果等于则停止，并保存 cluster
-        while nvirgin != size:
-            # 以 len(sample_idx) + size 作为新的聚类的size
-            clusters = Cluster(n_clusters=cluster_size).fit(self.sg_obj.vector)
-            labels = clusters.labels_[self.sg_obj.sample_idx]
-            labels_set = set(labels)
-            virgin = set(range(cluster_size)) - labels_set
-            nvirgin = len(virgin)
-            # 判断分类以后空类数目与size的大小
-            # 如果大于size，则减小size，并记录空类的数目
-            if nvirgin > size:
-                cluster_size -= 1
-                larger_clusters = clusters
-                larger_virgin = virgin
-            # 如果小于 size 则增大size，检查上一个size是否有记录，如果有记录则使用上个size 的记录。从中随机选择size个点作为采样点。
-            elif nvirgin < size:
-                cluster_size += 1
-                if larger_clusters is not None:
-                    clusters = larger_clusters
-                    virgin = larger_virgin
-                    break
-        # 从 virgin 里面选取 size 个点
-        rng = np.random.default_rng()
-        cluster_idx = rng.choice(list(virgin), size=size)
-        if center:
-            # 取中心位置的格点
-            centers = clusters.cluster_centers_[cluster_idx]
-            center_dist = cdist(centers, self.sg_obj.vector)  # 计算每个点到中心的距离
-            point_idx = np.argmin(center_dist, axis=-1)
-        else:
-            # 取这些 clusters 中能量最小值点
-            point_idx = []
-            for c_id in cluster_idx:
-                p_idx = np.arange(len(self.sg_obj.points))[clusters.labels_ == c_id]
-                # 求这些点的能量最小值
-                if 'energy' not in self.sg_obj.grid_property:
-                    raise NotImplementedError
-                p_energy = self.sg_obj.grid_energy[p_idx]
-                point_idx.append(p_idx[p_energy.argmin()])
-
-        return point_idx
-
-
-class NewtonSampling(SamplingBase):
-    """
-    沿着受力方向进行采样
-    """
-
-    def _samples(self, size, **kwargs):
-        raise NotImplementedError
-
-
-class RandomWalk(SamplingBase):
-    """
-    从给定点出发随机行走进行采样
-    """
-
-    def __init__(self, sg_obj=None, probability=1.0, **kwargs):
-        super().__init__(sg_obj, probability, **kwargs)
-
-    def _samples(self, size, **kwargs):
-        raise NotImplementedError
-
-
-class SystematicSampling(SamplingBase):
-    """
-    等距采样。主要用于测试。
-    """
-
-    def _samples(self, size, **kwargs):
-        if 'start' in kwargs:
-            start = kwargs['start']
-        else:
-            start = random.randint(0, self._pop_size)
-        stop = self._pop_size
-        indices = list(range(start, stop)) + list(range(0, start))
-        step = int(self._pop_size / size)
-        point_idx = [indices[i + n * step] for n, i in enumerate(range(size))]
-
-        return point_idx
+"""
+TODO: 关键点采样：top 位置、hollow位，bridge 位等等。
+"""
+import itertools
+import numpy as np
+from ase.geometry import get_distances
+from scipy.spatial import ConvexHull
+from scipy.spatial.distance import cdist
+from scipy.special import comb
+from sklearn.cluster import KMeans as Cluster
+import random
+
+from surface_construct.utils import furthest_sites
+
+MIN_HULL_ANGLE_COS = np.cos(np.pi * 30 / 180)
+
+def hull_vertices(hull):
+    hsimplices = hull.simplices
+    hvertices = hull.vertices
+    hnorms = hull.equations[:,0:-1]
+    ndim = hsimplices.shape[1]
+    vertices = []
+    # 去掉 hull 的 simplices 的角度较大的点
+    for i in hvertices:
+        p0_facets_idx = np.argwhere(hsimplices == i)[:,0]
+        p0_norms = hnorms[p0_facets_idx]
+        cosangle = lambda a,b: a.dot(b) / (np.linalg.norm(a) * np.linalg.norm(b))
+        # 求 i 凸点相邻的超平面的法向向量之间的夹角。如果存在夹角小于30度，即平面之间的夹角大于150度，则排除该点。反之，保留该点。
+        norm_angle_cos = np.absolute([cosangle(a,b) for a,b in itertools.combinations(p0_norms, 2)])
+        if np.sum(norm_angle_cos < MIN_HULL_ANGLE_COS) >= comb(ndim,2):
+            vertices.append(i)
+
+    return vertices
+
+
+def addition_samples(sg_obj, size=None, probability=None, **kwargs):
+    if 'seed' in kwargs:
+        seed = kwargs['seed']
+    else:
+        seed = None
+    if probability is None:
+        probability = {
+            "max_sigma": 0.2,  # 采样方法的概率
+            "max_diversity": 0.8,
+        }
+    if size is None:
+        size = 1
+
+    # 归一化
+    total = sum(probability.values())
+    if total != 1.0:
+        probability = {k: v / total for k, v in probability.items()}
+
+    rng = np.random.default_rng(seed)
+    method_list = rng.choice(list(probability.keys()), size=size, p=list(probability.values()))
+
+    point_idx = np.array([], dtype=int)
+    for method in method_list:
+        method_lower = method.lower()
+        if method_lower == 'max_sigma':
+            sampling_obj = MaxSigmaSampling(sg_obj)
+        elif method_lower == 'max_diversity':
+            sampling_obj = MaxDiversitySampling(sg_obj)
+        else:
+            raise NotImplementedError
+        point_idx = np.concatenate([point_idx, sampling_obj.samples(size=1, **kwargs)])  # 每种方法只采一个
+
+    return point_idx
+
+
+class SamplingBase:
+    def __init__(self, sg_obj, **kwargs):
+        self.sg_obj = sg_obj
+
+    @property
+    def _pop_size(self):
+        return len(self.sg_obj.points)
+
+    @property
+    def _population(self):
+        """
+        默认的全体是 sg_obj.points 的 index
+        :return:
+        """
+        return range(self._pop_size)
+
+    def _append_sample_to_sg(self, point_idx=None):
+        """
+        将采样点加入到 sg_obj.sample_points 和相应的 vector
+
+        :return:
+        """
+        if point_idx is not None:
+            if self.sg_obj.sample_idx is not None:
+                self.sg_obj.sample_idx = np.concatenate([self.sg_obj.sample_idx, point_idx])
+                self.sg_obj._sample_vector = np.concatenate([self.sg_obj._sample_vector, self.sg_obj.vector[point_idx]])
+                self.sg_obj.sample_points = np.concatenate([self.sg_obj.sample_points, self.sg_obj.points[point_idx]])
+            else:
+                self.sg_obj.sample_idx = np.array(point_idx)
+                self.sg_obj._sample_vector = self.sg_obj.vector[point_idx]
+                self.sg_obj.sample_points = self.sg_obj.points[point_idx]
+
+    def _samples(self, size, **kwargs):
+        raise NotImplementedError
+
+    def samples(self, size=1, **kwargs):
+        point_idx = self._samples(size=size, **kwargs)
+        self._append_sample_to_sg(point_idx=point_idx)
+
+        return point_idx
+
+
+class KeyPointSampling(SamplingBase):
+    """
+    基本思路：先定位 hollow位，再根据图论分析定位 bridge 位，最后定位top位。第二步，筛选等价位点。
+    需要一个基础func，从xy坐标，找到对应的格点。
+    """
+    def _samples(self, size, **kwargs):
+        pass
+
+
+class RandomSampling(SamplingBase):
+    """
+    完全随机的选择点，仅用于测试，效率太低。
+    """
+
+    def __init__(self, sg_obj, **kwargs):
+        super().__init__(sg_obj, **kwargs)
+        if 'seed' in kwargs:
+            self.seed = kwargs['seed']
+        else:
+            self.seed = None
+
+    def _samples(self, size, **kwargs):
+        rng = np.random.default_rng(self.seed)
+        pop_idx = rng.choice(self._population, size=size)
+        return pop_idx
+
+
+class MaxSigmaSampling(SamplingBase):
+    """
+    对最大误差的点进行采样
+    """
+
+    def _samples(self, size, **kwargs):
+        if 'energy' in self.sg_obj.grid_property:
+            # 如果已经读入了一些能量，则返回误差最大的点
+            idx = self.sg_obj.grid_property_sigma['energy'].argmax()
+            return [idx]
+        else:
+            raise "No energy for all population, pls do initial sampling first!"
+
+
+class InitialSampling(SamplingBase):
+    """
+    使用聚类-分层采样进行初始采样
+    """
+
+    def _samples(self, size, **kwargs):
+        hull = ConvexHull(self.sg_obj.vector)
+        #vertices = []
+        # 去掉 hull 的 simplices 的角度较大的点
+        #for i in hull.vertices:
+        #    p1_idx, p2_idx = np.argwhere(hull.simplices == i)
+        #    p0 = hull.points[i]
+        #    p1 = hull.points[hull.simplices[p1_idx[0],1-p1_idx[1]]]
+        #    p2 = hull.points[hull.simplices[p2_idx[0],1-p2_idx[1]]]
+        #    a = p1 - p0
+        #    b = p2 - p0
+        #    cosangle = a.dot(b)/(np.linalg.norm(a) * np.linalg.norm(b))
+        #    if cosangle > MIN_HULL_ANGLE_COS:
+        #        vertices.append(i)
+        # 聚类，vector_mesh
+        vertices = hull_vertices(hull)
+        n_vector_mesh = int(hull.volume / (self.sg_obj._vector_unit *
+                                           self.sg_obj.interval)**self.sg_obj.vector.shape[1]) + 1
+        cluster0 = Cluster(n_clusters=n_vector_mesh)
+        cluster0.fit(self.sg_obj.vector)
+        mesh_centers = cluster0.cluster_centers_
+        self.sg_obj._mesh_centers = mesh_centers
+        cluster = Cluster(n_clusters=size)
+        cluster.fit(mesh_centers)
+        self.sg_obj._clusters = cluster
+        nvert = len(vertices)
+        if nvert >= size:
+            print(f"Warning: Sample number better be larger than {nvert}!")
+            if size == 1:
+                sample_idx = np.random.choice(vertices,1)
+            elif size==nvert:
+                sample_idx = vertices
+            else:
+                sample_idx = [vertices[i] for i in
+                              furthest_sites(self.sg_obj.vector[vertices], size)]
+        else:
+            # 聚类
+            cluster2 = Cluster(n_clusters=size-nvert)
+            cluster2.fit(mesh_centers)
+            center_dist = cdist(cluster2.cluster_centers_, self.sg_obj.vector)  # 计算每个点到中心的距离
+            sample_idx = vertices + np.argmin(center_dist, axis=-1).tolist()
+        return sample_idx
+
+    def _append_sample_to_sg(self, point_idx=None):
+        """
+        将采样点加入到 sg_obj.sample_points 和相应的 vector
+        :return:
+        """
+        if point_idx is not None:
+            self.sg_obj.sample_idx = np.asarray(point_idx)
+            self.sg_obj._sample_vector = self.sg_obj.vector[point_idx]
+            self.sg_obj.sample_points = self.sg_obj.points[point_idx]
+
+
+class MaxDiversitySampling(SamplingBase):
+    """
+    对当前采样结构差异最大的点进行采样
+    基本思路是这样的：
+        * 重新进行聚类，
+        * 判断已经采样点属于的类别，找出没有点的类别，空类
+        * 如果空类不止一个，比较这些空类中心与旧点的距离，选择距离最大的点。
+    """
+
+    def _samples(self, size, center=False, **kwargs):
+        """
+
+        :param size:
+        :param center: 是否取中心。如果不是，则取能量最小值的点。如果没有能量则报错。
+        :param kwargs:
+        :return:
+        """
+        # 判断是否有过往的采样点，如果没有，调用 InitialSampling
+        if self.sg_obj.sample_idx is None:
+            raise "Please add initial samples (e.g. InitialSampling) before invoke this method!"
+        cluster_size = len(self.sg_obj.sample_idx) + size
+        nvirgin = 0
+        larger_clusters = None
+        larger_virgin = None
+        virgin = None
+        clusters = None
+        # 如果等于则停止，并保存 cluster
+        while nvirgin != size:
+            # 以 len(sample_idx) + size 作为新的聚类的size
+            clusters = Cluster(n_clusters=cluster_size).fit(self.sg_obj.vector)
+            labels = clusters.labels_[self.sg_obj.sample_idx]
+            labels_set = set(labels)
+            virgin = set(range(cluster_size)) - labels_set
+            nvirgin = len(virgin)
+            # 判断分类以后空类数目与size的大小
+            # 如果大于size，则减小size，并记录空类的数目
+            if nvirgin > size:
+                cluster_size -= 1
+                larger_clusters = clusters
+                larger_virgin = virgin
+            # 如果小于 size 则增大size，检查上一个size是否有记录，如果有记录则使用上个size 的记录。从中随机选择size个点作为采样点。
+            elif nvirgin < size:
+                cluster_size += 1
+                if larger_clusters is not None:
+                    clusters = larger_clusters
+                    virgin = larger_virgin
+                    break
+        # 从 virgin 里面选取 size 个点
+        rng = np.random.default_rng()
+        cluster_idx = rng.choice(list(virgin), size=size)
+        if center:
+            # 取中心位置的格点
+            centers = clusters.cluster_centers_[cluster_idx]
+            center_dist = cdist(centers, self.sg_obj.vector)  # 计算每个点到中心的距离
+            point_idx = np.argmin(center_dist, axis=-1)
+        else:
+            # 取这些 clusters 中能量最小值点
+            point_idx = []
+            for c_id in cluster_idx:
+                p_idx = np.arange(len(self.sg_obj.points))[clusters.labels_ == c_id]
+                # 求这些点的能量最小值
+                if 'energy' not in self.sg_obj.grid_property:
+                    raise NotImplementedError
+                p_energy = self.sg_obj.grid_energy[p_idx]
+                point_idx.append(p_idx[p_energy.argmin()])
+
+        return point_idx
+
+
+class NewtonSampling(SamplingBase):
+    """
+    沿着受力方向进行采样
+    """
+
+    def _samples(self, size, **kwargs):
+        raise NotImplementedError
+
+
+class RandomWalk(SamplingBase):
+    """
+    从给定点出发随机行走进行采样
+    """
+
+    def __init__(self, sg_obj=None, probability=1.0, **kwargs):
+        super().__init__(sg_obj, probability, **kwargs)
+
+    def _samples(self, size, **kwargs):
+        raise NotImplementedError
+
+
+class SystematicSampling(SamplingBase):
+    """
+    等距采样。主要用于测试。
+    """
+
+    def _samples(self, size, **kwargs):
+        if 'start' in kwargs:
+            start = kwargs['start']
+        else:
+            start = random.randint(0, self._pop_size)
+        stop = self._pop_size
+        indices = list(range(start, stop)) + list(range(0, start))
+        step = int(self._pop_size / size)
+        point_idx = [indices[i + n * step] for n, i in enumerate(range(size))]
+
+        return point_idx

{surface_construct-0.8 → surface_construct-0.8.2}/surface_construct/surface_grid.py

@@ -24,12 +24,12 @@ from scipy.interpolate import griddata
 from scipy.spatial import ConvexHull
 from scipy.spatial.distance import euclidean, cdist
 from sklearn.cluster import KMeans as Cluster
-from sklearn.cluster import kmeans_plusplus
 from sklearn.decomposition import PCA
 from sklearn.gaussian_process import GaussianProcessRegressor
 from sklearn.gaussian_process.kernels import RBF, ConstantKernel, WhiteKernel
 from sklearn.preprocessing import StandardScaler
 
+from surface_construct.sampling import InitialSampling, addition_samples
 from surface_construct.utils import get_calc_info, GridGenerator, get_distances, furthest_sites
 
 
@@ -321,60 +321,18 @@ class SurfaceGrid:
         k = d_vector / self.interval
         return np.min(k)
 
-    def grid_sample(self, N=10):
+    def grid_sample(self, N=1, probability=None, **kwargs):
         """
         Warning: Obsoleted, replaced by Sampling class
+        :param probability:
         :param N:
         :return:
         """
-
         if 'energy' in self.grid_property:
-            # 如果已经读入了一些能量，则返回误差最大的点
-            idx = self.grid_property_sigma['energy'].argmax()
-            max_sigma_point = np.array([self.points[idx]])
-            self.sample_idx = np.concatenate([self.sample_idx, [idx]])
-            self._sample_vector = np.concatenate([self._sample_vector, [self.vector[idx]]])
-            self.sample_points = np.concatenate([self.sample_points, max_sigma_point])
-            return [idx]
-
-        assert N > 1
-        hull = ConvexHull(self.vector)
-        vertices = []
-        # 去掉 hull 的 simplices 的角度较大的点
-        for i in hull.vertices:
-            p1_idx, p2_idx = np.argwhere(hull.simplices == i)
-            p0 = hull.points[i]
-            p1 = hull.points[hull.simplices[p1_idx[0],1-p1_idx[1]]]
-            p2 = hull.points[hull.simplices[p2_idx[0],1-p2_idx[1]]]
-            a = p1 - p0
-            b = p2 - p0
-            cosangle = a.dot(b)/(np.linalg.norm(a) * np.linalg.norm(b))
-            if cosangle > np.cos(np.pi*150/180):
-                vertices.append(i)
-        # 聚类，vector_mesh
-        n_vector_mesh = int(hull.volume / (self._vector_unit * self.interval)**self.vector.shape[1]) + 1
-        cluster0 = Cluster(n_clusters=n_vector_mesh)
-        cluster0.fit(self.vector)
-        mesh_centers = cluster0.cluster_centers_
-        self._mesh_centers = mesh_centers
-        nvert = len(vertices)
-        cluster = Cluster(n_clusters=N)
-        cluster.fit(mesh_centers)
-        if nvert >= N:
-            warning("Sample number should be larger than {nvert}")
-            sample_idx = [vertices[i] for i in furthest_sites(self.vector[vertices], N)]
+            points_idx = addition_samples(self, size=N, probability=probability, **kwargs)
         else:
-            # 聚类
-            cluster2 = Cluster(n_clusters=N-nvert)
-            cluster2.fit(mesh_centers)
-            center_dist = cdist(cluster2.cluster_centers_, self.vector)  # 计算每个点到中心的距离
-            sample_idx = vertices + np.argmin(center_dist, axis=-1).tolist()
-
-        self._clusters = cluster
-        self.sample_idx = sample_idx
-        self.sample_points = self.points[sample_idx]
-        self._sample_vector = self.vector[sample_idx]  # 保存用于作图
-        return self.sample_idx
+            points_idx = InitialSampling(self).samples(size=N)
+        return points_idx
 
     # TODO: 将中心重新映射回到Cartesian坐标 ：
         # 找到向量空间最紧邻的N个点，判断其实空间的距离是否小于 interval × 2，直到有三个点满足
@@ -555,15 +513,7 @@ class SurfaceGrid:
         fig, ax = plt.subplots()
         ax.set_aspect('equal')
 
-        print("Plot {} ...".format(key))
-        #X = self.points[:, 0].reshape((self.grid_ny, self.grid_nx))
-        #Y = self.points[:, 1].reshape((self.grid_ny, self.grid_nx))
-        #Z = self.grid_property[key].reshape((self.grid_ny, self.grid_nx))
-        #if vmax is None:
-        #    vmax = Z.max() + (Z.max() - Z.min()) * 0.2
-        #if vmin is None:
-        #    vmin = Z.min() - (Z.max() - Z.min()) * 0.2
-        #contourf0 = ax.contourf(X, Y, Z, levels=50, cmap="jet", vmin=vmin, vmax=vmax)
+        print(f"Plot {key} ...")
         x = self.points[:, 0]
         y = self.points[:, 1]
         z = self.grid_property[key]
@@ -579,8 +529,7 @@ class SurfaceGrid:
             ax.scatter(sample_points[nsampled:, 0], sample_points[nsampled:, 1], marker="o", s=100, linewidths=2,
                        color="w", zorder=10)
         fig.colorbar(contourf0, ax=ax)
-        title = "{} distribution, Max(sigma)= {:.3f}".format(
-            key.capitalize(), self.grid_property_sigma[key].max())
+        title = f"{key.capitalize()} distribution, Max(sigma)={self.grid_property_sigma[key].max():.3f}"
         ax.set_title(title)
         fig.set_dpi(300)
         fig.set_size_inches(10, 10)
@@ -651,11 +600,6 @@ class SurfaceGrid:
         fig, ax = plt.subplots()
         ax.set_aspect('equal')
 
-        #X = self.points[:, 0].reshape((self.grid_ny, self.grid_nx))
-        #Y = self.points[:, 1].reshape((self.grid_ny, self.grid_nx))
-        #Z = self.grid_property_sigma[key].reshape((self.grid_ny, self.grid_nx))
-        #contourf0 = ax.contourf(X, Y, Z, levels=50, cmap="RdPu", vmax=vmax, vmin=vmin)
-
         x = self.points[:, 0]
         y = self.points[:, 1]
         z = self.grid_property_sigma[key]
@@ -667,8 +611,7 @@ class SurfaceGrid:
         ax.scatter(max_sigma_point[0], max_sigma_point[1], marker="+", s=100, linewidths=2,
                    color="w", zorder=10)
         fig.colorbar(contourf0, ax=ax)
-        title = "Gaussian Process Error of {}, Max(sigma)= {:.3f} eV".format(
-            key.capitalize(), self.grid_property_sigma[key].max())
+        title = f"Gaussian Process Error of {key.capitalize()}, Max(sigma)={self.grid_property_sigma[key].max():.3f} eV"
         ax.set_title(title)
 
         fig.set_dpi(300)

{surface_construct-0.8 → surface_construct-0.8.2}/surface_construct/utils.py

@@ -14,11 +14,11 @@ from skimage.measure import marching_cubes
 def calc_hull_vertices(v):
     shape = v.shape
     if len(shape) != 2:
-        print("Warning: The vector should be 2D, however {}D vector was provided!)".format(len(shape)))
+        print(f"Warning: The vector should be 2D, however {len(shape)}D vector was provided!)")
         print("The Convex Hull Vertices won't be calculated.")
         return None
     if shape[1] > 5:
-        print("Warning: The vector.shape[1]={} is too large to be calculated!)".format(shape[1]))
+        print(f"Warning: The vector.shape[1]={shape[1]} is too large to be calculated!)")
         print("The Convex Hull Vertices won't be calculated.")
         return None
     try:
@@ -117,7 +117,7 @@ class GridGenerator:
         """
         self.atoms = atoms
         self._grid = None
-        self.atoms_num_type = set(atoms.numbers)
+        self.atoms_num_type = sorted(set(atoms.numbers))
         self.interval = interval
 
         if subtype is None:
@@ -261,7 +261,7 @@ class GridGenerator:
         xyz = rattle(atoms.cell.cartesian_positions(fxyz), stdev=self.interval / 3)
         grid_tree = cKDTree(xyz, copy_data=True)
 
-        # 对atoms 在 xy 方向超胞. Adapt from ase.geometry.geometry.general_find_mic
+        # 对atoms 在 xyz 方向超胞. Adapt from ase.geometry.geometry.general_find_mic
         ranges = [np.arange(-1 * p, p + 1) for p in atoms.pbc]
         hkls = np.array(list(itertools.product(*ranges)))
         vrvecs = hkls @ atoms.cell
@@ -299,6 +299,38 @@ class GridGenerator:
             print("Too much grid number, it will be very slow.")
         view(self.atoms + ase.Atoms(symbols=['X'] * len(self.grid), positions=self.grid))
 
+    def get_grid_site_type(self, site_dict=None):
+        """
+        根据第一近邻原子返回格点所对应的类型
+        :param site_dict: 格点的类型字典
+            example: {0:((atom_num, count),(atom_num, count),...), ..., 'next_idx':int}
+        :return: site label for each grid, site_dict
+        """
+
+        if site_dict is None:
+            site_dict = {'next_idx': 0}
+        _, Dga = get_distances(self.grid, self.atoms.positions, use_ase=True, cell=self.atoms.cell, pbc=self.atoms.pbc)
+        Lga = (Dga - (self.rsub+self.rads)*1.1 < 0)  # 格点与原子的连接性。如果距离小于半径之和，则为不连接
+        LTga = np.asarray([(Lga & (self.atoms.numbers==n1)).sum(-1) for n1 in self.atoms_num_type]).T  # 每个格点相连类矩阵
+        label_T_set = sorted(set(tuple(i) for i in LTga))  # 以原子类型区分，不同格点的类别集合
+        site_label = tuple(tuple((self.atoms_num_type[idx], l) for idx, l in enumerate(label)) for label in label_T_set)
+        if site_dict is None:
+            site_dict = {i:v for i,v in enumerate(site_label)}
+            site_dict['next_idx'] = len(site_label)
+        else:
+            # 比较新旧字典，如果有新的，就加入到 site_dict 中
+            old_site_label = tuple(v for k,v in site_dict.items() if type(k) == int)
+            new_site = [i for i in site_label if i not in old_site_label]
+            for ns in new_site:
+                idx = site_dict['next_idx']
+                site_dict[idx] = ns
+                site_dict['next_idx'] = idx + 1
+        site_dict_reverse = {tuple(iv[1] for iv in v):i for i,v in site_dict.items() if type(i)==int}
+        grid_T_label = [site_dict_reverse[tuple(i)] for i in LTga]
+        self.grid_site_type = grid_T_label
+        self.site_type_dict = site_dict
+        return grid_T_label, site_dict
+
 
 def rattle(positions, stdev=0.001, rng=None, seed=None):
     """Rattle the grid to make the vector distribution more smooth.

{surface_construct-0.8 → surface_construct-0.8.2/surface_construct.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: surface_construct
-Version: 0.8
+Version: 0.8.2
 Summary: Surface termination construction especially for complex model, such as oxides or carbides.
 Home-page: https://gitee.com/pjren/surface_construct/
 Author: ren