diff options
Diffstat (limited to 'aied2018/presentation/aied_poster.tex')
| -rw-r--r-- | aied2018/presentation/aied_poster.tex | 33 |
1 files changed, 14 insertions, 19 deletions
diff --git a/aied2018/presentation/aied_poster.tex b/aied2018/presentation/aied_poster.tex index 73de0e6..7f7db2f 100644 --- a/aied2018/presentation/aied_poster.tex +++ b/aied2018/presentation/aied_poster.tex @@ -13,10 +13,10 @@ \usepackage{color} \newcommand\red[1]{{\begingroup\color[rgb]{0.9,0.2,0.2}#1\endgroup}} \newcommand\blue[1]{{\begingroup\color[rgb]{0.15,0.15,0.8}#1\endgroup}} -\newcommand\green[1]{{\begingroup\color[rgb]{0.15,0.8,0.15}#1\endgroup}} +\newcommand\green[1]{{\begingroup\color[rgb]{0.10,0.7,0.10}#1\endgroup}} \usepackage{fancyvrb,courier} -\fvset{commandchars=\\\{\},baselinestretch=0.98,samepage=true,xleftmargin=2.5mm} +\fvset{commandchars=\\\{\},baselinestretch=0.98,samepage=true,xleftmargin=2.5mm,fontsize=\small} \usepackage{tikz} \usepackage{forest} @@ -53,7 +53,7 @@ \setbeamercolor*{block title}{fg=white,bg=FRIRed} \setbeamercolor*{block body}{fg=black, bg=white} - \begin{myblock}{Motivation and Research Questions} + \begin{myblock}{Motivation} \input{motivation.tex} \end{myblock} \setbeamercolor*{block title}{fg=white,bg=TitleBG} @@ -83,27 +83,22 @@ \setbeamercolor*{block title}{fg=white,bg=TitleBG} \setbeamercolor*{block body}{fg=black, bg=white} - \begin{myblock}{Learning Rules and Results} + \begin{myblock}{Rules and results} \input{rules.tex} \end{myblock}\vfill \setbeamercolor*{block title}{fg=white,bg=FRIRed} \begin{myblock}{Conclusions} \begin{itemize} - \item Abstract-syntax-tree (AST) patterns for representing program patterns. - \item Patterns are extracted automatically and combined into n-rules(errors) and p-rules (approaches) with machine learning. - - \item Patterns are useful, because in our experiment ... - \begin{itemize} - \item classification accuracy of Random Forest was on average 17\% higher than default accuracy. - \item n-rules explained over 70\% of incorrect submissions. - \item p-rules explained 62\% of correct programs. - \end{itemize} - \item However ... - \begin{itemize} - - \item In some domains, patterns were not informative (\textsf{ballistics} and \textsf{minimax}), therefore more sophisticated patterns are needed. - \item To construct new patterns, a tool for vizualization of patterns is needed. - \end{itemize} + \item AST patterns represent program features, while induced n-rules and p-rules encode errors and solution strategies. + + \item Patterns are useful, because in our experiment ...\\ + ~~... classification accuracy was on average 17\% higher using patterns;\\ + ~~... n-rules explained over 70\% of incorrect submissions;\\ + ~~... p-rules explained 62\% of correct submissions. + + \item However ...\\ + ~~... patterns were not always informative -- new kinds of patterns are needed;\\ + ~~... a visualization tool could support discovery of new kinds of useful patterns. \end{itemize} \end{myblock}\vfill } |