Skip to content
GitLab
Explore
Sign in
Primary navigation
Search or go to…
Project
S
schaertl_andreas
Manage
Activity
Members
Labels
Plan
Issues
Issue boards
Milestones
Wiki
Code
Merge requests
Repository
Branches
Commits
Tags
Repository graph
Compare revisions
Snippets
Build
Pipelines
Jobs
Pipeline schedules
Artifacts
Deploy
Releases
Container registry
Model registry
Operate
Environments
Monitor
Incidents
Analyze
Value stream analytics
Contributor analytics
CI/CD analytics
Repository analytics
Model experiments
Help
Help
Support
GitLab documentation
Compare GitLab plans
Community forum
Contribute to GitLab
Provide feedback
Terms and privacy
Keyboard shortcuts
?
Snippets
Groups
Projects
Show more breadcrumbs
supervision
schaertl_andreas
Commits
b165ae21
Commit
b165ae21
authored
4 years ago
by
Andreas Schärtl
Browse files
Options
Downloads
Patches
Plain Diff
add some actually interesting slides
parent
49747a58
No related branches found
No related tags found
No related merge requests found
Changes
1
Hide whitespace changes
Inline
Side-by-side
Showing
1 changed file
doc/slides/slides.tex
+57
-15
57 additions, 15 deletions
doc/slides/slides.tex
with
57 additions
and
15 deletions
doc/slides/slides.tex
+
57
−
15
View file @
b165ae21
...
@@ -181,17 +181,38 @@
...
@@ -181,17 +181,38 @@
\mathcal
{
L
}
\rightarrow
\mathcal
{
E
}
\rightarrow
\mathcal
{
D
}
\mathcal
{
L
}
\rightarrow
\mathcal
{
E
}
\rightarrow
\mathcal
{
D
}
\end{align*}
\end{align*}
from
\emph
{
third party library~
$
\mathcal
{
L
}$}
to an
\emph
{
XML
from
\emph
{
third party library~
$
\mathcal
{
L
}$}
to an
\emph
{
XML
representation
~
$
\mathcal
{
E
}$}
to
\emph
{
database storage~
$
\mathcal
{
D
}$}
.
export
~
$
\mathcal
{
E
}$}
to
\emph
{
database storage~
$
\mathcal
{
D
}$}
.
\item
When
$
\mathcal
{
E
}$
changes, so does~
$
\mathcal
{
E
}$
when a new
\item
When
$
\mathcal
{
L
}$
changes, so does~
$
\mathcal
{
E
}$
when a new
export is run. This should result in~
$
\mathcal
{
D
}$
changing as well.
export is run. This should result in~
$
\mathcal
{
D
}$
changing as well.
\end{itemize}
\end{itemize}
\begin{alertblock}
{
Huh!
}
\begin{alertblock}
{
Open Question
}
Not trivial! A work around is (1)~re-creating databases from
Not trivial! A work around could be (1)~re-creating databases from
scratch and (2)~splitting up knowledge into smaller repositories.
scratch and (2)~splitting up knowledge into smaller database
repositories.
\end{alertblock}
\end{alertblock}
\end{frame}
\end{frame}
\begin{frame}
{
Implementation: Versioning
}
\begin{alertblock}
{
Open Question
}
Not trivial! A work around could be (1)~re-creating databases from
scratch and (2)~splitting up knowledge into smaller database
repositories.
\end{alertblock}
\begin{itemize}
\item
Updating existing RDF data sets is unrealistic. Adding
versioning information to~
$
n
$
triplets can add up to
$
\mathcal
{
O
}
(
n
)
$
triplets.
\item
Constantly re-creating the entire data sets introduces
lots of latency between updates.
\item
Distributing every given query to a set of smaller
repositories and then joining the result seems like a long term
solution.
\end{itemize}
\end{frame}
% [Applications & Questions]
% [Applications & Questions]
\section
{
Exploration
}
\section
{
Exploration
}
...
@@ -224,8 +245,8 @@
...
@@ -224,8 +245,8 @@
ULO~predicates. Many predicates are used by neither exports.
ULO~predicates. Many predicates are used by neither exports.
\end{itemize}
\end{itemize}
\begin{alertblock}
{
Huh!
}
\begin{alertblock}
{
Open Question
}
Queries formulated for a tetrapodal search system have to
Queries formulated for a tetrapodal search system
might
have to
account for these ``holes'' in existing data sets.
account for these ``holes'' in existing data sets.
\end{alertblock}
\end{alertblock}
\end{frame}
\end{frame}
...
@@ -247,7 +268,7 @@
...
@@ -247,7 +268,7 @@
\texttt
{
unimportant
}
\texttt
{
universe
}
\texttt
{
uses
}
\texttt
{
unimportant
}
\texttt
{
universe
}
\texttt
{
uses
}
\end{flushleft}
\end{flushleft}
\end{scriptsize}
\end{scriptsize}
\caption
{
Used predicates in an Isabelle export.
}
\caption
{
Used predicates in an Isabelle export
~
\cite
{
uloisabelle
}
.
}
\end{figure}
\end{figure}
\begin{figure}
\begin{figure}
...
@@ -279,10 +300,32 @@
...
@@ -279,10 +300,32 @@
\texttt
{
uses-implementation
}
\texttt
{
uses-interface
}
\texttt
{
uses-implementation
}
\texttt
{
uses-interface
}
\end{flushleft}
\end{flushleft}
\end{scriptsize}
\end{scriptsize}
\caption
{
\emph
{
Unused
}
predicates in an Isabelle export.
}
\caption
{
Unused predicates in an Isabelle export
~
\cite
{
uloisabelle
}
.
}
\end{figure}
\end{figure}
\end{frame}
\end{frame}
\subsection
{
Scoring Items
}
\begin{frame}
{
Exploration: Scoring
}
\begin{itemize}
\item
One query for a tetrapodal search system is the following:
``Find theorems with non-elementary proofs.''~
\cite
{
tetra
}
.
problems''~
\cite
{
tetra
}
.
\begin{itemize}
\item
Assuming ``elementary'' means easy~
\cite
{
elempro
}
, we
need to find a way of rating the difficulty of proofs.
\item
We can rate proofs by length, check time and so on.
\item
Computing metrics can even be formulated in a SPARQL query.
\end{itemize}
\end{itemize}
\begin{alertblock}
{
Open Question
}
Scanning the entire data set every time we get a tetrapodal query
for a given score is unrealistic. Such queries need to be cached,
either on the organizational or tetrapodal level.
\end{alertblock}
\end{frame}
\subsection
{
Algorithms and Problems
}
\subsection
{
Algorithms and Problems
}
\begin{frame}
{
Exploration: Algorithms and Problems
}
\begin{frame}
{
Exploration: Algorithms and Problems
}
...
@@ -295,12 +338,15 @@
...
@@ -295,12 +338,15 @@
Tempting but potentially very complicated.
Tempting but potentially very complicated.
\item
Algorithms aren't programs! Programs implement
\item
Algorithms aren't programs! Programs implement
algorithms that solve problems.
algorithms that solve problems.
\item
Maybe it would be interesting to collect a database of
algorithms and proofs (OEIS~
\cite
{
oeis
}
for algorithms).
\end{itemize}
\end{itemize}
\end{itemize}
\end{itemize}
\begin{alertblock}
{
Huh!
}
\begin{alertblock}
{
Open Question
}
This illustrates the difficulty in designing an ontology
This illustrates the difficulty in designing an ontology
(schema) that is both expressive and concise.
(schema) that is both expressive and concise. Should an universal
ontology be as concise as possible?
\end{alertblock}
\end{alertblock}
\end{frame}
\end{frame}
...
@@ -320,10 +366,6 @@
...
@@ -320,10 +366,6 @@
subsets of~
{
ULO
}
. On the other hand, representing algorithms and
subsets of~
{
ULO
}
. On the other hand, representing algorithms and
algorithmic problems might require us to extend~
{
ULO
}
? Maybe instead
algorithmic problems might require us to extend~
{
ULO
}
? Maybe instead
of tetrapodal search we need
$
n
$
-podal search.
of tetrapodal search we need
$
n
$
-podal search.
\begin{itemize}
\item
Maybe it would be interesting to collect a database of
algorithms and proofs (OEIS~
\cite
{
oeis
}
for algorithms).
\end{itemize}
\end{itemize}
\end{itemize}
\end{frame}
\end{frame}
...
...
This diff is collapsed.
Click to expand it.
Preview
0%
Loading
Try again
or
attach a new file
.
Cancel
You are about to add
0
people
to the discussion. Proceed with caution.
Finish editing this message first!
Save comment
Cancel
Please
register
or
sign in
to comment