few typos everywhere

Author: qgeissmann

report/discussion.tex

@@ -24,7 +24,7 @@ \subsection{Software package for feature computation}
 
 Several \texttt{PyEEG} functions were also found to be inconsistent with mathematical
 definitions (see \pr{} documentation, appendix).
-This unfortunatly apperas to be a common issue for academic software.
+This unfortunately appears to be a common issue for academic software.
 The general status of the peer-review process and the reproducibility of programs and algorithms have
 recently drawn attention (see \cite{morin_shining_2012,crick_can_2014} for
 discussions about this issue).

report/matmet.tex

@@ -33,7 +33,7 @@ \subsection{Data preprocessing}
 This frequency is typically uses as a cut-off value between theta and delta
 waves \cite{vyazovskiy_nrem_2014}.
 In addition, \gls{eeg} and \gls{emg} signals were standardised ($E[x] = 0, Var[x] = 1$) to account for the variability in baseline amplitude due to acquisition.
-Vigilance state anotations were resampled at exactly 0.20Hz using nearest neighbour interpolation.
+Vigilance state annotations were resampled at exactly 0.20Hz using nearest neighbour interpolation.
 
 \subsection{Feature extraction from time series}
 \label{sub:features}
@@ -152,7 +152,7 @@ \subsection{Statistical analysis}
 
 
 In order to assess the significance of the effect of the addition of temporal variables on the cross-validation
-error (fig.~\ref{fig:temporal_integration}), repreated t-tests were performed.
+error (fig.~\ref{fig:temporal_integration}), repeated t-tests were performed.
 Bonferronni correction was applied.
 
 In order to determine whether state prevalences were different between predicted and ground truth time series (fig.~\ref{fig:struct_assess}A),

report/results.tex

@@ -10,7 +10,7 @@ \subsection{A new efficient \texttt{python} package}
 Very significant improvement in performance were achieved for almost all functions(table~\ref{tab:benchmark}).
 Critically, sample and approximate entropies\cite{richman_physiological_2000} became usable in
 reasonable time.
-For instance, 40h of CPU time were originally  requiered to compute sample entropy on a all 5 second epochs in a 24h recording.
+For instance, 40h of CPU time were originally required to compute sample entropy on a all 5 second epochs in a 24h recording.
 The improved implementation cut this down to 55min.
 \input{./tables/benchmark}
 
@@ -59,7 +59,7 @@ \subsection{Including temporal information improves prediction}
 Significant improvement was achieved by both methods by including even little temporal information ($\tau = 1$, $n=\{1,3\}$, $p-value < 5.10^{-3}$).
 Although both methods improve performance, it appears that the inter-recording variablility is systematically lower using convolution method.
 For instance, the standard variation in error between all recordings is always lower than 0.03 for convolutions of window size greater than two,
-whilst it is systematicaly above 0.031 for all the values of $\tau$.
+whilst it is systematically above 0.031 for all the values of $\tau$.
 The most significant difference with the original set of variables was achieved using $n = \{1,3,7,15,31\}$ ($p-value < 10^{-6}$).
 Therefore, this new set of 105 variables was used to train the final predictor.
 Combining both convolution and lag approaches did not improve prediction any further (data not shown).

report/tables/importances.tex

@@ -2,7 +2,7 @@
 \begin{center}
 \caption{\ctit{Relative variable importance of the 21 selected features.}
 Random forest algorithm can produce a value to quantify variable importance.
-Variable importance corresponds to how much, statistically, a variable contributes to reducing the prediction inacuracy (or, to be more precise, the Gini impurity).
+Variable importance corresponds to how much, statistically, a variable contributes to reducing the prediction inaccuracy (or, to be more precise, the Gini impurity).
 Starting from 164 variables, the least important variables were recursively eliminated.
 This table represents the 21 most important remaining features.
 Features from both \gls{eeg} and \gls{emg} are important for accurate prediction.