uai: add a comparison table to forests in the appendix

wittawatj · Jun 6, 2015 · 83c3322 · 83c3322
1 parent 36abbe6
commit 83c3322
Show file tree

Hide file tree

Showing 7 changed files with 9,053 additions and 4,776 deletions.
diff --git a/code/KernelEP.NET/KernelEP.NET.userprefs b/code/KernelEP.NET/KernelEP.NET.userprefs
@@ -1,14 +1,15 @@
 <Properties>
   <MonoDevelop.Ide.Workspace ActiveConfiguration="Release|x86" />
-  <MonoDevelop.Ide.Workbench ActiveDocument="KernelEP.NET/Main.cs">
+  <MonoDevelop.Ide.Workbench ActiveDocument="KernelEP.NET/Tasks/CollectOnlineLogistic.cs">
     <Files>
-      <File FileName="KernelEP.NET/Tool/OnlineMap.cs" Line="178" Column="5" />
-      <File FileName="KernelEP.NET/Main.cs" Line="18" Column="8" />
-      <File FileName="KernelEP.NET/Tool/Mapper.cs" Line="343" Column="24" />
-      <File FileName="KernelEP.NET/Op/LogisticOp.cs" Line="794" Column="20" />
-      <File FileName="KernelEP.NET/Tool/Misc.cs" Line="315" Column="10" />
-      <File FileName="KernelEP.NET/Tool/LocalConfig.cs" Line="31" Column="9" />
-      <File FileName="KernelEP.NET/Tasks/CollectOnlineCG.cs" Line="23" Column="24" />
+      <File FileName="KernelEP.NET/Tool/OnlineMap.cs" Line="277" Column="47" />
+      <File FileName="KernelEP.NET/Main.cs" Line="17" Column="38" />
+      <File FileName="KernelEP.NET/Tool/RandFeature.cs" Line="178" Column="20" />
+      <File FileName="KernelEP.NET/Op/LogisticOp.cs" Line="199" Column="24" />
+      <File FileName="KernelEP.NET/Tasks/CollectOnlineLogistic.cs" Line="383" Column="4" />
+      <File FileName="KernelEP.NET/TestIdea/LogisticRegression.cs" Line="87" Column="47" />
+      <File FileName="KernelEP.NET/Op/CompoundGamma.cs" Line="562" Column="2" />
+      <File FileName="KernelEP.NET/Tasks/CollectOnlineCG.cs" Line="53" Column="4" />
     </Files>
   </MonoDevelop.Ide.Workbench>
   <MonoDevelop.Ide.DebuggingService.Breakpoints>

diff --git a/lskl_icml_workshop2015/kjit_lskl_workshop2015.tex b/lskl_icml_workshop2015/kjit_lskl_workshop2015.tex
@@ -451,7 +451,7 @@ \subsection{Random feature approximations}
 One approach to learning the mapping  $\approxMsg{\factor}{\outV}{\theta}$ from incoming to outgoing messages
 would be to employ Gaussian process regression, using the kernel \eqref{eq:gauss_joint_emb}.
 This approach is not suited to just-in-time (JIT) learning, however,
-as both prediction and storage cost grow with the size of the training set.
+as both prediction and storage costs grow with the size of the training set.
 Instead, we define 
 a finite-dimensional random feature map $\hat{\psi} \in \mathbb{R}^{D_\mathrm{out}}$ such that 
 $\kappa(\mathsf{r}, \mathsf{s}) \approx \hat{\psi}(\mathsf{r})^\top \hat{\psi}(\mathsf{s})$, 

diff --git a/uai2015/img/pred_var/fm_kgg_joint_-bw-out1-crop.pdf b/uai2015/img/pred_var/fm_kgg_joint_-bw-out1-crop.pdf