diff options
Diffstat (limited to 'doc/doc.tex')
| -rw-r--r-- | doc/doc.tex | 2 |
1 files changed, 2 insertions, 0 deletions
diff --git a/doc/doc.tex b/doc/doc.tex index 71e5e1e..aa85061 100644 --- a/doc/doc.tex +++ b/doc/doc.tex @@ -35,6 +35,8 @@ In the graphs, which show the total reward earned as a function of the episodes Note that the nn and nn2 runs get a reward of only $0.5$ for reaching the goal, while the old and new runs get $10$. Therefore, the nn/nn2 results have been multiplied by $20$ to make them comparable. +Then, there has been done a run with \verb+self.NN.randomize_weights+ in the range of $(-0.27, 0.27)$ instead of $(1,1)$. + \gnuplotloadfile[terminal=pdf]{array_naive_vs_episodic.gnuplot} \gnuplotloadfile[terminal=pdf]{array_naive_vs_episodic_deriv.gnuplot} |