theano summary

doc/omlw2014/presentation.tex

@@ -13,7 +13,7 @@
 
 \title{Theano, Pylearn2, libgpuarray Presentation}
 
-\author{
+\author{%
 \footnotesize
 Frédéric Bastien\superscript{a}, Bart van Merriënboer\superscript{a} \newline
 \superscript{a}Département d'Informatique et de Recherche Opérationnelle \newline
@@ -163,7 +163,7 @@ Montréal, Canada \newline
       \item Datasets (MNIST, CIFAR-10) and preprocessors (LCN, ZCA)
     \end{itemize}
     \item Experiments can be specified through a YAML config file, or by a Python script
-    \item Scripts for visualizing weights, plot monitored values during training
+    \item Scripts for visualizing weights, plot monitored values
   \end{itemize}
 
 \end{frame}
@@ -174,9 +174,74 @@ Montréal, Canada \newline
 
 
 \section{Theano}
-\begin{frame}{Theano details}
+\begin{frame}{Description}
+  \begin{itemize}
+    \item Mathematical symbolic expression compiler
+    \item Expressions mimic NumPy's syntax and semantics (easier to use)
+    \item Dynamic C/CUDA code generation
+    \begin{itemize}
+      \item Is used with other technologie to generate fast code: C/C++, CUDA, OpenCL, PyCUDA, Cython, Numba, \ldots
+    \end{itemize}
+    \item Efficient symbolic differeniation
+    \begin{itemize}
+      \item Theano computes derivatives of functions with one or many inputs.
+      \item Also support computation of the Jacobian, Hessian, R and L op.
+    \end{itemize}
+    \item Speed and stability optimizations
+    \begin{itemize}
+      \item Gives the right answer for ``log(1+x)'' even if x is really tiny.
+    \end{itemize}
+    \item Works on Linux, Mac and Windows
+    \item Transparent use of a GPU
+    \begin{itemize}
+      \item {\tt float32} only for now (working on other data types)
+      \item Still in experimental state on Windows
+    \end{itemize}
+
+    \item Extensive unit-testing and self-verification
+    \begin{itemize}
+      \item Detects and diagnoses many types of errors
+    \end{itemize}
+
+%    \item Statically typed and purely functional
+    \item Sparse operations (CPU only)
+  \end{itemize}
+\end{frame}
+
+\begin{frame}{Simple example}
+%--------------
+%
+%>>> import theano
+%>>> a = theano.tensor.vector("a")      # declare symbolic variable
+%>>> b = a + a ** 10                    # build symbolic expression
+%>>> f = theano.function([a], b)        # compile function
+%>>> print f([0, 1, 2])                 # prints `array([0, 2, 1026])`
+%
+%
+%======================================================  =====================================================
+%        Unoptimized graph                                    Optimized graph
+%======================================================  =====================================================
+%.. image:: ../hpcs2011_tutorial/pics/f_unoptimized.png   .. image:: ../hpcs2011_tutorial/pics/f_optimized.png
+%======================================================  =====================================================
+%
+%Symbolic programming = *Paradigm shift*: people need to use it to understand it.
 \end{frame}
+
+
 \begin{frame}{Project status?}
+  \begin{itemize}
+    \item Mature: Theano has been developed and used since January 2008 (6.5 yrs old)
+    \item Driven over 100 research papers
+    \item Good user documentation
+    \item Active mailing list with participants from outside our lab
+    \item Core technology for a few Silicon-Valley start-ups
+    \item Many contributors (some from outside our lab)
+    \item Used to teach many university classes
+    \item Has been used for research at Google and Yahoo.
+  \end{itemize}
+  Theano: \url{deeplearning.net/software/theano/}
+
+  Deep Learning Tutorials: \url{deeplearning.net/tutorial/}
 \end{frame}
 
 
@@ -205,7 +270,7 @@ Montréal, Canada \newline
 
 \begin{frame}{Why a new implementation?}
 \begin{block}{There are already a number of existing GPU computing codebases:}
-Theano, PyCUDA/PyOpenCL, CUDAmat, Gnumpy, Thrust, ...
+Theano, PyCUDA/PyOpenCL, CUDAmat, Gnumpy, Thrust, \ldots
 \end {block}
 \begin{enumerate}
 \item<2-> All are incompatible, which hinders code sharing.