初步结果

app.py

@@ -1,22 +1,34 @@
 import streamlit as st
-from pages import page1, page2, page3
+# from subapps import page1, page2, page3
 
 # 设置页面配置
 st.set_page_config(layout="wide")
 
 # 主应用
 def main():
-    st.sidebar.title("导航")
-    pages = {
-        "NIPS 论文数据集高斯混合聚类分析": page1,
-        "第二个子应用": page2,
-        "第三个子应用": page3
-    }
+    # st.sidebar.title("导航")
+    # pages = {
+    #     "NIPS 论文数据集高斯混合聚类分析": page1,
+    #     "第二个子应用": page2,
+    #     "第三个子应用": page3
+    # }
     
-    page = st.sidebar.radio("选择子应用", tuple(pages.keys()))
+    # page = st.sidebar.radio("选择子应用", tuple(pages.keys()))
     
-    # 根据选择的子应用加载相应的页面
-    pages[page].main()
+    # # 根据选择的子应用加载相应的页面
+    # pages[page].main()
+    # app.py
+
+    # https://www.cnblogs.com/wang_yb/p/18502232
+
+    page1 = st.Page("my_pages/page1.py", title="无监督聚类动态变化", icon="📊", default=True)
+    page2 = st.Page("my_pages/page2.py", title="第二个子应用", icon="📈")
+    page3 = st.Page("my_pages/page3.py", title="第三个子应用", icon="📉")
+    # page3 = st.Page("dev/test.py", title="第三个子应用", icon="📉")
+
+    pg = st.navigation([page1, page2, page3])
+    pg.run()
+    # pass
 
 if __name__ == "__main__":
     main()

{pages → my_pages}/page1.py

@@ -10,10 +10,10 @@ from io import BytesIO
 import time
 from utils.data_processor import load_data, process_data, build_hyperedges
 from utils.visualizer import visualize_gmm, visualize_ratings
-from utils.hypergraph_drawer import draw_hypergraph
+from utils.hypergraph_drawer import draw_hypergraph, pyplot_fig_to_buffer
 
 def main():
-    st.title("NIPS 论文数据集高斯混合聚类分析")
+    st.title("NeurIPS 论文数据集高斯混合聚类分析")
 
     # 自动播放
     slider_max = 10
@@ -31,10 +31,11 @@ def main():
     if st.session_state.play_state:
         button_label = "暂停"
     else:
-        button_label = "播放"
+        button_label = "开始拟合"
     st.button(button_label, on_click=toggle_play, key="play_button")
     # 播放速度
-    speed = st.slider("播放速度", min_value=0.1, max_value=2.0, value=1.0, step=0.1, key="speed_slider")
+    speed = st.slider("拟合速度", min_value=0.1, max_value=2.0, value=1.0, step=0.1, key="speed_slider")
+    # speed = st.slider("播放速度", min_value=0.1, max_value=2.0, value=1.0, step=0.1, key="speed_slider")
     # 主页面布局
 
     # 显示迭代次数滑条
@@ -52,16 +53,18 @@ def main():
     # 使用 sidebar 控制参数
     with st.sidebar:
         st.header("控制面板")
+        draw_width = st.slider("绘图宽度", min_value=3, max_value=20, value=6, step=1, key="draw_width")
+        draw_height = st.slider("绘图高度",min_value=3, max_value=20, value=6, step=1, key="draw_height")
+
         max_samples = len(df)
         num_samples = st.slider("选择采样论文数量", min_value=1, 
                                 max_value=min(100, max_samples), value=min(10, max_samples), step=1)
-
-
+                                
         
         # 添加复选框选择显示 paper 的属性
         display_attribute = st.selectbox(
             "选择显示 paper 的属性",
-            ["index", "id", "title", "keywords", "author"]
+            ["order", "index", "id", "title", "keywords", "author"]
         )
         # 选择是 top k 还是 top p
         display_option = st.selectbox(
@@ -81,31 +84,89 @@ def main():
         sampled_df, probabilities, paper_attributes = process_data(df, iteration, num_samples)
         # print(display_attribute) # 字符串
         hyperedges = build_hyperedges(probabilities, paper_attributes, display_attribute, top_k=top_k, top_p=top_p)
+        hypergraph = hnx.Hypergraph(hyperedges)
+
         # print(hyperedges)
 
 
+    st.header("超图可视化")
 
+    draw_dual = st.toggle("对偶翻转", value=False, key="draw_dual")
+    euler_first = st.toggle("先看Euler Diagram", value=True, key="euler_first")
     
-    # 并排展示超图和高斯混合分布
-    col1, col2 = st.columns(2)
-    with col1:
-        st.header("超图可视化")
-        hypergraph_image = draw_hypergraph(hyperedges)
-        st.image(hypergraph_image, caption="超图可视化", use_container_width=True)
-
-    with col2:
-        st.header("高斯混合分布聚类结果")
-        fig_gmm = visualize_gmm(sampled_df, iteration)
-        st.plotly_chart(fig_gmm, use_container_width=True)
+    def euler_visualization(hypergraph):
+        st.subheader("Euler Diagram 可视化")
+        col1, col2, col3, col4 = st.columns(4)
+        with col1:
+            fill_edges  = st.toggle("填充超边", value=True, key="fill_edges")
+        with col2:
+            toplexes    = st.toggle("只保留未被包含的边（toplexes）", value=False, key="toplexes")
+        with col3:
+            with_edge_labels = st.toggle("不看边名", value=False, key="with_edge_labels")
+            
+
+        col1, col2 = st.columns(2)
+        captions = ["论文聚类超图", "论文聚类对偶超图"]
+
+        if 'poses' not in st.session_state:
+            st.session_state.poses = [None, None]
+        if draw_dual:
+            hypergraph = hypergraph.dual()
+            captions = ["论文聚类对偶超图", "论文聚类超图"]
+            # st.session_state.poses = [st.session_state.poses[1], st.session_state.poses[0]]
+
+        with col1:
+            hypergraph_image, st.session_state.poses[0] = draw_hypergraph(hypergraph, draw_dual=False, 
+                                            fill_edges=fill_edges, toplexes=toplexes,
+                                            with_edge_labels=with_edge_labels
+                                            # pos=st.session_state.poses[0]
+                                            ,draw_width=draw_width, draw_height=draw_height
+                                            )
+            st.image(hypergraph_image, caption = captions[0], use_container_width=True)
+
+        with col2:
+            hypergraph_image, st.session_state.poses[1] = draw_hypergraph(hypergraph, draw_dual=True, 
+                                            fill_edges=fill_edges, toplexes=toplexes,
+                                            with_edge_labels=with_edge_labels
+                                            # pos=st.session_state.poses[1]
+                                            ,draw_width=draw_width, draw_height=draw_height
+                                            )
+            st.image(hypergraph_image, caption =captions[1], use_container_width=True)
+
+    def new_visualization(hypergraph):
+        st.subheader("新型可视化")
+        col1, col2 = st.columns(2)
+        with col1:
+            fig, ax = plt.subplots(figsize=(draw_width, draw_height))
+            hnx.draw_bipartite_using_euler(hypergraph)
+            st.image(pyplot_fig_to_buffer(plt.gcf()), caption="双列二分图可视化", use_container_width=True)
+        with col2:
+            fig, ax = plt.subplots(figsize=(draw_width, draw_height))
+            hnx.draw_incidence_upset(hypergraph)
+            st.image(pyplot_fig_to_buffer(plt.gcf()), caption="Incidence/UpSet 可视化", use_container_width=True)
+
+    if euler_first:
+        euler_visualization(hypergraph)
+        new_visualization(hypergraph)
+    else:
+        new_visualization(hypergraph)
+        euler_visualization(hypergraph)
+
+    st.header("高斯混合分布聚类结果")
+    fig_gmm = visualize_gmm(sampled_df, iteration)
+    st.plotly_chart(fig_gmm, use_container_width=True)
 
     # 显示采样论文的详细信息
     st.header("采样论文详细信息")
-    st.dataframe(sampled_df[["title", "keywords", "rating_avg", "confidence_avg", "site"]])
+    st.dataframe(sampled_df[["index", "title", "keywords", "rating_avg", "confidence_avg", "site"]
+                            ]
+                            # .style.highlight_max(axis=0)
+                            )
 
     # 增加第二种可视化方式
-    st.header("论文评分分布")
-    fig_bar = visualize_ratings(sampled_df)
-    st.plotly_chart(fig_bar, use_container_width=True)
+    # st.header("论文评分分布")
+    # fig_bar = visualize_ratings(sampled_df)
+    # st.plotly_chart(fig_bar, use_container_width=True)
 
 
     
@@ -131,8 +192,8 @@ def main():
 
 
     
-if __name__ == "__main__":
-    # 设置页面布局
-    st.set_page_config(layout="wide")
-    # 运行主函数
-    main()
+# if __name__ == "__main__":
+# # 设置页面布局
+#     st.set_page_config(layout="wide")
+#     # 运行主函数
+main()

{pages → my_pages}/page2.py

@@ -2,4 +2,6 @@ import streamlit as st
 
 def main():
     st.title("第二个子应用")
-    st.write("这里是第二个子应用的内容。")
+    st.write("这里是第二个子应用的内容。")
+
+main()

{pages → my_pages}/page3.py

@@ -2,4 +2,6 @@ import streamlit as st
 
 def main():
     st.title("第三个子应用")
-    st.write("这里是第三个子应用的内容。")
+    st.write("这里是第三个子应用的内容。")
+
+main()

utils/data_processor.py

@@ -9,7 +9,7 @@ def load_data():
 
 def process_data(df, iteration, num_samples):
     # 随机采样论文
-    sampled_df = df.sample(n=num_samples, random_state=iteration)
+    sampled_df = df.sample(n=num_samples, random_state=iteration).reset_index()
 
     # 计算每个论文属于各个 cluster 的概率
     probabilities = []
@@ -20,6 +20,8 @@ def process_data(df, iteration, num_samples):
         probabilities.append(prob_list)
         paper_attributes.append(
             {
+                "order": idx,
+                "index": row['index'],
                 "id": row["id"],
                 "title": row["title"],
                 "keywords": row["keywords"],
@@ -40,8 +42,9 @@ def build_hyperedges(
     # 构建超图边
     hyperedges: Dict[str, List[str]] = {}
     for idx, (prob, paper_attr) in enumerate(zip(probabilities, paper_attributes)):
-        if display_attribute_name == "index":
-            display_attribute = f"Paper {idx}"
+        if display_attribute_name == "index" or display_attribute_name == "order":
+            # display_attribute = f"Paper {idx}"
+            display_attribute = f"Paper {paper_attr[display_attribute_name]}"
         else:
             display_attribute: str = paper_attr[display_attribute_name]
         if top_k is not None:
@@ -50,10 +53,10 @@ def build_hyperedges(
             # 累加起来，直到第一次大于等于 p
             selected_indices = []
             cumulative_prob = 0.0
-            for i, p in enumerate(prob):
+            for i, p in enumerate(np.sort(prob)[::-1]):
                 selected_indices.append(i)
                 cumulative_prob += p
-                if cumulative_prob >= top_p:
+                if cumulative_prob > top_p+1e-4:
                     break
 
             

utils/hypergraph_drawer.py

@@ -1,20 +1,53 @@
+from turtle import title
+from typing import Dict
 import hypernetx as hnx
 import matplotlib.pyplot as plt
 from matplotlib.backends.backend_agg import FigureCanvasAgg as FigureCanvas
 from io import BytesIO
 
-def draw_hypergraph(hyperedges):
-    # 构建超图
-    H = hnx.Hypergraph(hyperedges)
-
-    # 绘制超图
-    fig, ax = plt.subplots(figsize=(12, 8))
-    hnx.draw(H, ax=ax)
-
-    # 将超图保存为图像
+def pyplot_fig_to_buffer(fig):
+    """Draw the figure using Pyplot and return the image buffer."""
     canvas = FigureCanvas(fig)
     buffer = BytesIO()
     canvas.print_png(buffer)
     buffer.seek(0)
+    return buffer
+
 
-    return buffer
+def draw_hypergraph(
+        H:hnx.Hypergraph=None,
+        hyperedges:Dict = None, 
+        draw_dual:bool=False,
+        toplexes:bool=False,
+        fill_edges:bool=True,
+        with_edge_labels:bool=True, 
+        with_node_labels:bool=True, 
+        title:str=None,
+        pos = None
+    ,draw_width:int = 6, draw_height:int = 6
+                    ):
+    # 构建超图
+    if H is None and hyperedges is None:
+        raise ValueError("Either H or hyperedges must be provided.")
+    if H is None:
+        H = hnx.Hypergraph(hyperedges)
+
+    if draw_dual:
+        H = H.dual()
+    if toplexes:
+        H = H.toplexes(return_hyp=True)
+
+    # 绘制超图
+    fig, ax = plt.subplots(figsize=(draw_width, draw_height))
+    if title is not None:
+        ax.set_title(title)
+    pos = hnx.draw(H, ax=ax, fill_edges=fill_edges, 
+                with_edge_labels=with_edge_labels,
+            with_node_labels=with_node_labels,
+                node_label_alpha=0.0, 
+                edge_label_alpha=0.0, pos=pos,
+                return_pos=True)
+        # hnx.draw(H, ax=ax)
+
+    # 将超图保存为图像
+    return pyplot_fig_to_buffer(fig), pos