Update the test case section

[jaghili]

Add the following figures :

1. A figure where we plot $x_1,\ldots x_m$ close to the dataset $\cal{D}$ (but differents !) and also $xi \oplus\theta xj$ for each pair $(i,j)$ and also the quantity $f\theta (f_\theta^{-1}(xi) + f\theta^{-1}(x_j))$ for each pair
2. the same for $\odot_\theta$
3. and plot of the images $f_\theta(x_i)$ for each $x_i$
4. plot several lines $\cal{D}_1,\ldots,\cal{D}4$ around $(Ox)$ and plot the image by $f\theta$.

[chahid-rahouti]

hi sir, please can you verify these plot to see if the result we have is correct. the prediction changes everytime, but it is the same. Is it possible to have a meeting tomorrow to discuss about it @jaghili, @Blessed-joseph

[chahid-rahouti]

Y - Xi is our loss

[jaghili]

Replot as a table in the following way :

For any couple of points $(y_i, t_i)$ compute the associated error $||xi \oplus\theta xj - f\theta (f_\theta^{-1}(xi) + f\theta^{-1}(x_j))||$ list them as a table

|y_i|t_i|err_i|