diff --git a/code/examples_pyYggdrasil/exercise/exercise/yggdrasil_exercise.tex b/code/examples_pyYggdrasil/exercise/exercise/yggdrasil_exercise.tex
index 3ff73bf2a376074484617082734b9062a2e0881a..7a1911f17021999fb0025caffbedecb5031941bc 100644
--- a/code/examples_pyYggdrasil/exercise/exercise/yggdrasil_exercise.tex
+++ b/code/examples_pyYggdrasil/exercise/exercise/yggdrasil_exercise.tex
@@ -274,12 +274,12 @@ Note doing the \CPP{} \emph{exceptionally} well will result in full points on th
Try to be as efficient as possible.
\item
- Create a one dimensional distribution of your choice, and generate a few sample.
- Again test if the distribution is consistent also verify that the pdf function of the distribution is consistent with our choice.
+ Create a one dimensional distribution of your choice, and generate a few samples.
+ Again test if the distribution is consistent, also verify that the pdf function of the distribution is consistent with our choice.
\item
Redo the last exercise but with a two modal distribution.
- The mean of the different distributions must be different from each other.
+ The mean of the modes must be different from each other.
\end{enumerate}
@@ -295,11 +295,22 @@ Note doing the \CPP{} \emph{exceptionally} well will result in full points on th
\item
Create a container of your choice, and fill it with samples from your distributions.
- You are not allowed to use sampling methods that returns containers only the one that return samples.
+ You are only allowed to use the generation function that generate one sample at a time.
%
\\
%
- To pass this exercise you must be as efficient as possible, so choose the \emph{right} functions.
+ To pass this exercise you must be as efficient as possible, so choose the \emph{right} functions.
+
+
+ \item
+ Writing a generic function that operates on a container.
+ This exercise is primary for checking if you have mastered the container interface.
+ The function has to at least use a NumPy related function.
+ Also a copy operation (type preserving and type changing) of the container has to be used.
+ %
+ \\
+ %
+ As a recommendation the function should do, first copy its argument, modify it and then copy it again, but this type the container type should change.
\end{enumerate}
@@ -307,6 +318,7 @@ Note doing the \CPP{} \emph{exceptionally} well will result in full points on th
%%%%%%%%%%%%%%%%%%%%%%%%%%
+
\subsection{Tree}
\subsubsection{Warm up}
@@ -325,7 +337,8 @@ Note doing the \CPP{} \emph{exceptionally} well will result in full points on th
%
\\
%
- Justify that your results are accurate, no mathematics is needed just argue.
+ Justify your results.
+ Argue why your results are correct, no mathematics is needed just argue.
\item
Visualize the domains, also how the domain is partitioned.
@@ -342,8 +355,9 @@ Note doing the \CPP{} \emph{exceptionally} well will result in full points on th
\item
Generate a two dimensional distributions, with four modes.
- The only restriction is that the mean of the distribution must be $(2, 5)^{T}$ and the domain must be centred on that point, and have an extension in each dimension of $\pm10$ units.
- All mode must have different parameters from each other and no mode is allowed to have the mean given above.
+ The only restriction is that the mean of the (full) distribution must be $(2, 5)^{T}$ and the domain must be centred at that point.
+ The extension of the domain in each dimension must be $\pm10$ units.
+ All modes must have different parameters from each other and no mode is allowed to have the mean given above.
%
\\
%
@@ -408,37 +422,38 @@ Note doing the \CPP{} \emph{exceptionally} well will result in full points on th
\item
From the same tree generate a new tree sampler object, but this time apply a condition to it.
Verify that the sampler still generates samples that are distributed correctly.
- For that you have to compute the covariance, you can use Mathematica for that.
+ For that you have to compute the mean and covariance, you can use Mathematica to compute the analytical values.
\item
- Perform a scaling like experiment.
+ Perform a scaling experiment.
Fit a tree with a certain, in the beginning small amount of samples.
Use the tree to get a tree sampler object and use it to judge the accuracy of the quality of the sampler, as it was shown in the notebooks.
Increase the amount of samples and show the results.
+ %
+ \\
+ %
+ Use the same guides as above for this experiment.
\item
Generate a tree dimensional Dirichlet distribution, this distribution will have $4$ parameters, of your choice.
Generate samples from it and fit a tree.
From the tree construct a tree sampler that restricts one dimension.
The tree sampler will have ``dimension 2'' visualize the pdf of the sampler.
+ Compare it with the analytical value.
\end{enumerate}
-
-
-
% END: TREE SAMPLER
- %%%%%%%%%%%%%%%%%%%%%%%%
-
-
+ %%%%%%%%%%%%%%%%%%%%%%%%
% END: P Y T H O N
%%%%%%%%%%%%%%%%%%%%%%%%
\section{\CPP }
-This is the test on \CPP{} it is not expected that after studding the manual a user, beside some software developer, could solve them.
+This is the test on \CPP , it is not expected that after studding the manual a user, beside some software developer, could solve them.
Also whereas the Python part needs a few hour to do, this part will take several days, at least.
+The task here are more like a list of things that could be implemented.
%
\\
%
@@ -459,6 +474,17 @@ Also note that every new feature you add must also find its way to \pyYggdrasil
\item
Implement references into the sample array and the sample collection.
+ \item
+ Write a single function small function of your choice and integrate this function into the code base.
+ The code of this function such be created in a new folder called \texttt{extra}.
+ Also the function should be inside its own namespace which is also called \texttt{extra}.
+ Also export this function to \pyYggdrasil .
+ The function must be in its own submodule called \texttt{extra}.
+ %
+ \\
+ %
+ Do not forget to integrate the your code into CMake.
+
\item
Extend the sample list such that it is impossible to insert a sample of the wrong dimension.
@@ -470,7 +496,7 @@ Also note that every new feature you add must also find its way to \pyYggdrasil
This is considered rather easy.
\item
- Extend the container such that they have a ``natural array'' constructor.
+ Extend the containers such that they have a ``natural array'' constructor.
This is a constructor which takes an array that is tailored for their internal format.
\item
@@ -478,20 +504,22 @@ Also note that every new feature you add must also find its way to \pyYggdrasil
\item
Implement a new model and perform the integration in the code base.
- Since this part only goes about how well you can integrate a new model, you are allowed to copy one of the existing ones and rename it.
+ Since this part only goes about how well you can integrate a new model, you are allowed to copy one of the existing models and rename it.
\item
Implement a new test, the same as for the model applies.
\item
Implemented a task based parallelism into the tree.
+ For an idea about it consult the technical specifications.
\item
Implement a new sample container.
- That behaves as the array container, but is only a reference to an other array.
+ That behaves as the array container, but is only a reference to an array.
+ Similar to a \texttt{Eigen::Map} type.
\item
- Implement the inserting scheme.
+ Implement the inserting scheme into the tree.
Note this is considered hard.
\item
@@ -502,6 +530,11 @@ Also note that every new feature you add must also find its way to \pyYggdrasil
Implement the parallel generation of random samples from the tree and the distributions.
Note that a simple openMP \texttt{pargma} is not enough, the reason is that the generator is not thread safe.
Implement a generator that bypass that problem.
+
+ \item
+ Extend \emph{all} container such that they support slicing operations.
+ You can also select only a subset of them and integrate them.
+ Also integrate this feature into Python.
\end{enumerate}