[info] Remove initial results.

to sqaush

Author: Arthur Douillard

README.md

@@ -16,34 +16,9 @@ Every model must inherit `inclearn.models.base.IncrementalLearner`.
 
 :white_check_mark: iCaRL\
 :construction: Learning without Forgetting (LwF)\
-:white_check_mark: End-to-End Incremental Learning (E2E)
+:white_check_mark: End-to-End Incremental Learning (E2E)\
 :x: Overcoming catastrophic forgetting (EWC)
 
-## Results
-
-Every experiments have been runned at least 20 times, each with a different class
-ordering. The class ordering is defined by random using a different seed. I'm
-using the seed from 1 to 20.
-
-```bash
-python3 inclearn --model <model> --seed-range 1 20 --name <exp_name> <other options>
-```
-
-The metric used is what iCaRL defined the `average incremental accuracy`. It's
-what is plotted on every graph. In addition the in-parenthesis metric is the
-average of those average incremental accuracy. You can see in the notebook
-[here](results.ipynb) how it is done.
-
-I'll always precise whether the results are from a paper `[paper]` or myself `[me]`.
-
-
-### iCIFAR100, 10-split
-
-![icifar100, 10 split](figures/icifar100_10split.png)
-
-### iCIFAR100, 2-split
-
-TODO
 
 
 ## TODO