ENH we8there demos improved, bug fixed with display of topic top word…

…s in PrintTopics.py
bnpy · Dec 8, 2016 · 8754bc2 · 8754bc2
1 parent 2d2f6d4
commit 8754bc2
Show file tree

Hide file tree

Showing 6 changed files with 20 additions and 21 deletions.
diff --git a/bnpy/viz/PrintTopics.py b/bnpy/viz/PrintTopics.py
@@ -253,6 +253,11 @@ def plotCompsFromWordCounts(
         fig_h, ax_list = pylab.subplots(
             nrows=nrows, ncols=ncols,
             figsize=(ncols * W, nrows * H))
+    if isinstance(ax_list, np.ndarray):
+        ax_list = ax_list.flatten().tolist()
+    assert isinstance(ax_list, list)
+    n_images_viewable = len(ax_list)
+    n_images_to_plot = len(compListToPlot)
 
     for plotID, compID in enumerate(compListToPlot):
         cur_ax_h = ax_list[plotID] #pylab.subplot(nrows, ncols, plotID + 1)
@@ -278,11 +283,11 @@ def plotCompsFromWordCounts(
         if xlabels is not None:
             if len(xlabels) > 0:
                 cur_ax_h.set_xlabel(xlabels[plotID], fontsize=11)
-    # Disable empty plots!
-    #for kdel in xrange(plotID + 2, nrows * ncols + 1):
-    #    aH = pylab.subplot(nrows, ncols, kdel)
-    #    aH.axis('off')
-    # Fix margins between subplots
+
+    # Disable empty plots
+    for k, ax_h in enumerate(ax_list[n_images_to_plot:]):
+        ax_h.axis('off')
+
     return figH, ax_list
 
 def count2str(val, width=4, minVal=0.01, **kwargs):

diff --git a/...plot-demo=vb_single_run-model=mix+mult.py → ...-demo=mix_vb_single_run-model=mix+mult.py b/...plot-demo=vb_single_run-model=mix+mult.py → ...-demo=mix_vb_single_run-model=mix+mult.py
@@ -28,6 +28,7 @@
 doc_ids = np.flatnonzero(
     dataset.getDocTypeCountMatrix().sum(axis=1) >= 20)
 dataset = dataset.make_subset(docMask=doc_ids, doTrackFullSize=False)
+
 ###############################################################################
 #
 # Make a simple plot of the raw data

diff --git a/...ot-demo=vb+proposals-model=dp_mix+mult.py → ...emo=mix_vb+proposals-model=dp_mix+mult.py b/...ot-demo=vb+proposals-model=dp_mix+mult.py → ...emo=mix_vb+proposals-model=dp_mix+mult.py
@@ -39,38 +39,31 @@
 
 ###############################################################################
 #
-# Train with VB algorithm
-# -----------------------
+# Train with birth and merge proposals
+# ------------------------------------
 # 
 # Take the best of 1 initializations
+# 
+# Ideally, we'd run this longer, but this is convenient for rapid inspection.
 
 merge_kwargs = dict(
-    m_startLap=10,
+    m_startLap=5,
     m_pair_ranking_procedure='elbo',
     m_pair_ranking_direction='descending',
     m_pair_ranking_do_exclude_by_thr=1,
     m_pair_ranking_exclusion_thr=-0.0005,
     )
-'''
-trained_model, info_dict = bnpy.run(
-    dataset, 'DPMixtureModel', 'Mult', 'memoVB',
-    output_path='/tmp/we8there/helloworld-model=dp_mix+mult-K=30/',
-    nLap=1000, convergeThr=0.0001, nTask=1, nBatch=1,
-    K=30, initname='bregmankmeans+lam1+iter1',
-    gamma0=50.0, lam=0.1,
-    moves='birth,merge,shuffle',
-    b_startLap=2, b_Kfresh=5, b_stopLap=10,
-    **merge_kwargs)
-'''
+
 trained_model, info_dict = bnpy.run(
     dataset, 'DPMixtureModel', 'Mult', 'memoVB',
     output_path='/tmp/we8there/helloworld-model=dp_mix+mult-K=30/',
-    nLap=1000, convergeThr=0.0001, nTask=1, nBatch=1,
+    nLap=15, convergeThr=0.0001, nTask=1, nBatch=1,
     K=30, initname='bregmankmeans+lam1+iter1',
     gamma0=50.0, lam=0.1,
     moves='birth,merge,shuffle',
     b_startLap=2, b_Kfresh=5, b_stopLap=10,
     **merge_kwargs)
+
 bnpy.viz.PrintTopics.plotCompsFromHModel(
     trained_model,
     vocabList=dataset.vocabList,

diff --git a/...del_vb_single_run-model=hdp_topic+mult.py → ...pic_vb_single_run-model=hdp_topic+mult.py b/...del_vb_single_run-model=hdp_topic+mult.py → ...pic_vb_single_run-model=hdp_topic+mult.py
diff --git a/...odel_vb+proposals-model=hdp_topic+mult.py → ...opic_vb+proposals-model=hdp_topic+mult.py b/...odel_vb+proposals-model=hdp_topic+mult.py → ...opic_vb+proposals-model=hdp_topic+mult.py
diff --git a/setup.py b/setup.py
@@ -177,7 +177,7 @@ def make_list_of_datasets_specs():
 # Main function
 setup(
     name="bnpy",
-    version="0.1.4",
+    version="0.1.5",
     author="Michael C. Hughes",
     author_email="mike@michaelchughes.com",
     description=(