Merge pull request #2100 from nouiz/omlw

.gitignore

@@ -10,7 +10,6 @@
 *.log
 *.nav
 *.out
-*.pdf
 *.snm
 *.toc
 *.vrb

doc/index.txt

@@ -21,6 +21,8 @@ Montreal).
 News
 ====
 
+* Open Machine Learning Workshop 2014 `presentation <omlw2014/omlw_presentation.pdf>`_.
+
 * Colin Raffel `tutorial on Theano <http://nbviewer.ipython.org/github/craffel/theano-tutorial/blob/master/Theano%20Tutorial.ipynb>`_.
 
 * Ian Goodfellow did a `12h class with exercises on Theano <https://github.com/goodfeli/theano_exercises>`_.

doc/omlw2014/Makefile

+all: presentation.pdf sharing.pdf
+
+clean:
+	rm -f pygpu_ndarray.so core.* *.o *~
+
+cleantmp:
+	rm -f core.* *.o *~
+
+presentation.pdf: presentation.tex
+	pdflatex presentation
+	pdflatex presentation
+
+sharing.pdf: sharing.tex
+	pdflatex sharing
+	pdflatex sharing

doc/omlw2014/logreg.py

+import numpy
+import theano
+import theano.tensor as tt
+rng = numpy.random
+
+N = 400
+feats = 784
+D = (rng.randn(N, feats), rng.randint(size=N, low=0, high=2))
+training_steps = 10000
+
+# Declare Theano symbolic variables
+x = tt.matrix("x")
+y = tt.vector("y")
+w = theano.shared(rng.randn(feats), name="w")
+b = theano.shared(0., name="b")
+print "Initial model:"
+print w.get_value(), b.get_value()
+
+# Construct Theano expression graph
+p_1 = 1 / (1 + tt.exp(-tt.dot(x, w) - b))   # Probability that target = 1
+prediction = p_1 > 0.5                      # The prediction thresholded
+xent = -y * tt.log(p_1) - (1 - y) * tt.log(1 - p_1)  # Cross-entropy loss
+cost = xent.mean() + 0.01 * (w ** 2).sum()  # The cost to minimize
+gw, gb = tt.grad(cost, [w, b])
+
+# Compile
+train = theano.function(
+    inputs=[x, y],
+    outputs=[prediction, xent],
+    updates=[(w, w - 0.1 * gw),
+             (b, b - 0.1 * gb)],
+    name='train')
+
+predict = theano.function(inputs=[x], outputs=prediction,
+                          name='predict')
+
+# Train
+for i in range(training_steps):
+    pred, err = train(D[0], D[1])
+
+print "Final model:"
+print w.get_value(), b.get_value()
+print "target values for D:", D[1]
+print "prediction on D:", predict(D[0])

doc/omlw2014/sharing.tex

+\documentclass[utf8x,xcolor=pdftex,dvipsnames,table]{beamer}
+\usetheme{Malmoe}  % Now it's a beamer presentation with the lisa theme!
+\setbeamertemplate{footline}[page number]
+\usecolortheme{beaver}
+\usepackage[T1]{fontenc}
+\usepackage{amsmath}
+\usepackage[utf8x]{inputenc}
+%\logo{\includegraphics[width=.8in]{UdeM_NoirBleu_logo_Marie_crop}}
+\usepackage{listings}
+
+\newcommand{\superscript}[1]{\ensuremath{^{\textrm{#1}}}}
+
+\mode<presentation>
+
+\title{Theano, Pylearn2, libgpuarray: Sharing and Future}
+
+\author{%
+\footnotesize
+Frédéric Bastien, Bart van Merriënboer \newline
+Département d'Informatique et de Recherche Opérationnelle \newline
+Université de Montréal \newline
+Montréal, Canada \newline
+\texttt{\{bastienf, vanmerb\}@iro.umontreal.ca} \newline \newline
+}
+
+\date{OML Workshop 2014}
+
+\setbeamertemplate{navigation symbols}{}
+
+\begin{document}
+
+\begin{frame}[plain]
+ \titlepage
+ \vspace{-5em}
+ \includegraphics[width=1in]{../hpcs2011_tutorial/pics/lisabook_logo_text_3.png}
+ \hfill
+ \includegraphics[width=.8in]{../hpcs2011_tutorial/pics/UdeM_NoirBleu_logo_Marie_crop}
+\end{frame}
+
+\section{Future}
+\begin{frame}
+  \tableofcontents[currentsection]
+\end{frame}
+
+\begin{frame}{Theano}\setcounter{page}{1}
+\begin{itemize}
+\item Easier C code development and better documentation of that
+\item Faster compilation
+\item Multi-GPU
+\item Better looping (update to scan)
+\item Allow checkpoint with GPU to reload without GPU
+\item Less memory allocation(lower Theano overhead)
+\item Faster convolution
+\end{itemize}
+\end{frame}
+
+\begin{frame}{libgpuarray}
+\begin{itemize}
+\item Find other projects to use it?
+\item More functionality as NumPy
+\item Move some of the functionality from Python/Theano to the C level
+\item Optimize the kernel selection and parametrization based on the GPU
+\end{itemize}
+\end{frame}
+
+\begin{frame}{Pylearn2}
+\begin{itemize}
+\item RNN
+\item Better hyperparameter search support, using e.g. Hyperopt
+\item Documentation
+\item Checkpoint
+\item Better support for sparse dataset
+\item Machine translation examples
+\item Gated activations for conditional computation
+\item Variational Auto-Encoders
+\end{itemize}
+\end{frame}
+
+\begin{frame}
+\end{frame}
+
+\begin{frame}{Simplifying code sharing between}
+\begin{enumerate}
+  \item<1-> License: \begin{bf}Suggest BSD\end{bf} as it is used by many software in our field.
+    \begin{itemize}
+    \item Common license help share code.
+    \item When reusing code, don't forget to keep the license and the copyright notice
+    \end{itemize}
+  \item<2-> Common base object! \begin{bf}libgpuarray\end{bf}
+  \item<3-> Otherwise: put important implementation(e.g. convolution) in \begin{bf}separate file\end{bf} and \begin{bf}use raw ptr/shape/strides\end{bf} as inputs. Document that interface.
+  \item<4-> Acknowledge reuse \begin{bf}in section on web site\end{bf} AND \begin{bf}in papers\end{bf} about the software we reuse! (and use too)
+\end{enumerate}
+\end{frame}
+
+
+\end{document}