Fix typos in Optimizer post

parent 0f553bce
Pipeline #154089025 passed with stage
in 5 minutes and 30 seconds
......@@ -22,7 +22,7 @@ description: Learn how to automatically adjust figure widths and more to fit you
## Introduction
Scientific paper submission is a sort of empirical risk minimization problem where we want to minimize the risk that our paper will be rejected. We don't have access the true risk, but have to measure this empirical risk in some other way.
Scientific paper submission is a sort of empirical risk minimization problem where we want to minimize the risk that our paper will be rejected. We don't have access to the true risk, but have to measure this empirical risk in some other way.
There are many factors that affect this risk - the most obvious being the quality of the content. However, with the increasing number of submissions the first impression of a reviewer is also increasingly important. In order for a reviewer to be able to assess the real quality of our paper, we must first avoid that the reviewer throws our paper into the recycling bin. A paper that successfully avoids the recycling bin should continue to convey a positive feeling so that the reviewer tries to find a reason to accept the paper rather than the opposite.
......@@ -58,12 +58,12 @@ At submission time we find ourselves fighting the automated PDF checks of the pu
<div class="row justify-content-md-center">
<div class="col-md-8">
<div><img src="http://www.phdcomics.com/comics/archive/phd090617s.gif" alt="Page limits from Piled Higher and Deeper by Jorge Cham" class="img-rounded"/></div>
<div class="caption">A use-case many of us recognize, regardless of where we are in our academic careers. With permission from "<a href="http://www.phdcomics.com">Piled Higher and Deeper</a>" by Jorge Cham.</div>
<div class="caption">A situation many of us recognize, regardless of where we are in our academic careers. With permission from "<a href="http://www.phdcomics.com">Piled Higher and Deeper</a>" by Jorge Cham.</div>
</div>
</div>
</div>
One large step towards a solution to this was posed by {% cite Acher2018 --file latex-packages %} in which the authors annotate the LaTeX source with variability information. This information can be numerical values on figure sizes, or boolean values on options or whether to include certain paragraphs. In their work they formulate the learning problem as a constrained binary classification problem to classify into acceptable and non-acceptable configurations so that acceptable solutions can be presented to the user.
One large step towards a solution to this was proposed by {% cite Acher2018 --file latex-packages %} in which the authors annotate the LaTeX source with variability information. This information can be numerical values on figure sizes, or boolean values on options or whether to include certain paragraphs. In their work they formulate the learning problem as a constrained binary classification problem to classify into acceptable and non-acceptable configurations so that acceptable solutions can be presented to the user.
Here, we instead formulate this problem as a constrained optimization problem, where the constraints are defined by the automated PDF checks and the optimization is defined by proxy metrics such as amount of white space on the last page.
......@@ -234,7 +234,7 @@ We will use Ray Tune {% cite liaw2018tune --file python-packages %} for searchin
1. Copy the LaTeX code to a temporary directory.
1. Sample variables and write to file `macros.tex`,
1. Clean document directory using <code>latexmk</code> and
1. Compile document with <code>latemk</code>.
1. Compile document with <code>latexmk</code>.
1. Measure quality proxy-metrics on PDF file.
The average execution time for each task is 5.46 seconds, and we can run 8 of these in parallel. This means that an experiment takes roughly 30 minutes on my laptop. With Ray Tune we also have the option to run this using a much larger set of machines if needed. Compilation using more than one CPU per worker is shorter, but since we can run fewer in parallel the total execution time is longer.
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment