this is for holding javascript data
Naets added Figure_ref_MedianImshowNormDataGMMPCA_shows_one__.tex
about 8 years ago
Commit id: 4ebf3b49d0293b873f4288b3c9bb53b072cf7b8e
deletions | additions
diff --git a/Figure_ref_MedianImshowNormDataGMMPCA_shows_one__.tex b/Figure_ref_MedianImshowNormDataGMMPCA_shows_one__.tex
new file mode 100644
index 0000000..bc75f71
--- /dev/null
+++ b/Figure_ref_MedianImshowNormDataGMMPCA_shows_one__.tex
...
Figure \ref{MedianImshowNormDataGMMPCA} shows one the respective medians for each of the five principal components of the car dataset for each of the clusters. This allows one to characterize the specificities of each of the two clusters.
First, one can see the biggest difference among the two clusters is made upon the first principal component.This first PC depicts, as explained in more details in the first report, the general quality (and price) of the car. Therefore, we can conclude the biggest difference between the two clusters concerns the quality and the price of the cars. However, one cannot conclude the first cluster depicts the good quality cars and the second one the bad quality cars based on the results of assignment 1. The reason for this is that for this particular exercise, the SVD decomposition and PCA calculation were computed again. Hence, we might end up with opposite orientation of the axes, i.e. positive values depicting expensive cars could portray the current set up of axes while in the assignment 1 setup, positive values may delineated bad quality cheap cars.
Second, the clusters are also quite different from PC2 (even though much less that for PC1). This means that one of the clusters has a tendency to correspond more to large family cars that retain their monetary value over time compared to the other cluster.
It seems the other PCs are not as good benchmarks to dissociate the two clusters as the two first PCs are. Therefore, we do not deepen any further our clusters analysis.
