Design of Experiments (DoE)
===========================

When running experiments, it is important to stimulate the target environment in a way that we can extract
as much information as possible from it.

Good experiments shall stress the target environment as much as possible under different conditions.
This is important because a model is as trustworthy as the dataset used for validating it is *informative*.

    **Example**

    If we are developing the model of a car, we want to log sensors measurements 
    while driving within a wide range of speeds, with different accelerations profiles, 
    in different road and weather conditions and so on and so forth.

    If we log sensors measurements only when we are driving on a flat road and in the range 0-10 km/h 
    and by doing exactly the same manoeuvres over and over, then it would be hard to disagree 
    on that the collected dataset is poorly informative. 

 

In the current release, *Dymoval* only stores the coverage region of the dataset and compute 
some statistics on it.
In this way, the user have an understanding under which conditions the developed model is trustworthy, 
provided that the validation results provides good figures.

.. note::
   In future releases we plan to further provide measures (Cramer-Rao Theorem? Fisher Matrix?) on the
   information level contained in a given dataset in within its coverage region.


It is worth noting that a dataset covering a fairly large region 
won't necessarily imply *information richness.*
This happens for example when you take a wide range of values but you stimulate your target environment 
only with constant inputs in within such a range. 
You would certainly have a dataset with a fairly large covered region but... it would contain little information.  

    **Example**

    With reference to the example above, you can imagine to drive the car only at constant speeds 
    in a range from 0 to 180 km/h on a flat road without never accelerating of braking.
    That is, you make a first run driving (and logging) data at a constant speed of 10 km/h, without accelerating nor braking and by staying on a flat road. 
    Then you perform a second run at a constant speed of 20 km/h, still with the same condition as the previous run, and so on, until reaching the last run when you drove at 180 km/h.
    Your dataset will have a fairly large coverage region, but it will contain little information as in all the runs you drove at a constant speed without accelerating or braking and by staying on a flat road. 



How to design experiments that produce sufficiently informative datasets then?

Well, such an issue cannot be automatized but there is some theory behind it in the field of *Design of experiments (DoE)*,
feel free to google it for more details.
Due to that *DoE* cannot be automatized, it will not be included in *Dymoval*, at least for now.