grizzly/deneir
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

light modification of text

Browse source
This commit is contained in:
Arthur Grisel-Davy 2022-10-12 16:42:20 -04:00
parent bc0c042845
commit 17527217ae
1 changed files with 11 additions and 10 deletions
Download patch file Download diff file Expand all files Collapse all files

19
DSD/first_formalization_presentation/presentation_formalization.tex
View file

@ -13,7 +13,7 @@
\usepackage{array} \usepackage{array}
\usepackage{xspace} \usepackage{xspace}
\title{Device State Detector (DSD): Formalization} \title{Device State Detector (DSD): Technical Presentation}
\subtitle{Trust me I've read a book.} \subtitle{Trust me I've read a book.}
\date{} \date{}
\author{Arthur Grisel-Davy} \author{Arthur Grisel-Davy}
@ -42,7 +42,7 @@ $\Longrightarrow$ Supervised, multi-class, mono-label, classification problem.
\end{frame} \end{frame}
\begin{frame}{Normalized distance} \begin{frame}{Normalized distance}
\begin{align}\label{eq:distance_original} \begin{align}
&nd: a\times b \in T^2 \rightarrow \mathbb{R} \nonumber\\ &nd: a\times b \in T^2 \rightarrow \mathbb{R} \nonumber\\
&nd(a,b) = \dfrac{Eucd(a,b)}{N_b} &nd(a,b) = \dfrac{Eucd(a,b)}{N_b}
\end{align} \end{align}
@ -63,7 +63,7 @@ $\Longrightarrow$ Supervised, multi-class, mono-label, classification problem.
\begin{frame}{Performance metric} \begin{frame}{Performance metric}
\begin{itemize} \begin{itemize}
\item Accuracy: $Acc = \dfrac{1}{N}\sum_{n\in N}1_{(l[n] = true[n])}$ \item Accuracy: $Acc = \dfrac{1}{N}\sum_{n\in N}1_{(l[n] = true[n])}$
\item State occurrence: Count state occurrence independently of correct position or size. \item Levenshtein Distance: number of miss/non detected states.
\end{itemize} \end{itemize}
\end{frame} \end{frame}
@ -168,12 +168,13 @@ Three distance measures:
\end{frame} \end{frame}
\begin{frame}{Conclusion} \begin{frame}{Conclusion}
DSD is:
\begin{itemize} \begin{itemize}
\item DSD is an implementation of 1-NN with custom distances and shrinkage. \item an implementation of 1-NN with custom distances and shrinkage.
\item Can easily be adapted to multiple protos per class. \item easily adaptable to multiple protos per class.
\item Data requirements and constraints minimal. \item extremely low on data requirement.
\item Detects $P+1$ classes. \item able to detects $P+1$ classes.
\item One hyperparameter $\alpha$ to controle miss/un -classification tradeoff. \item parametrized by only $\alpha$ to controle miss/un -classification tradeoff.
\end{itemize} \end{itemize}
\end{frame} \end{frame}
@ -181,7 +182,7 @@ Three distance measures:
\begin{itemize} \begin{itemize}
\item Capture and label more data. \item Capture and label more data.
\item Evaluate possibility of uneven shrinkage. \item Evaluate possibility of uneven shrinkage.
\item Detect attacks, publish paper, save the world, accept Nobel prize. \item Enable multiple protos per class i.e. \textit{more-than-one}-NN
\end{itemize} \end{itemize}
\end{frame} \end{frame}