- Studying new classical algorithms.
- Installing OpenCV using Qt (instead of GTK).
- Continuing with the replicate of Vanessa’s master’s degree thesis.
Studying new classical algorithms.
During the week I have selected two classic Reinforcement Learning algorithms in which to focus the study and deepen the operation. These have been: Dynamic Programming and Monte Carlo.
To understand how they work, I had as a reference book: “Reinforcement Learning: An Introduction” by Richard S. Sutton and Andrew G. Barto.
In addition, I have been supporting the reading with other articles:
- Reinforcement Learning Demystified: Solving MDPs with Dynamic Programming.
- Model Free Reinforcement learning algorithms (Monte Carlo, SARSA, Q-learning).
- Monte Carlo in Reinforcement Learning, the Easy Way.
Installing OpenCV using Qt (instead of GTK).
I have tried compile OpenCV on the computer using flag to create using PyQT instead of GTK since in the Ubuntu version I use (Kubuntu: Ubuntu + KDE Plasma) sometimes has problems to raise a window.
In the process of trying to compile it I found the following error:
Makefile:162: recipe for target 'all' failed make: *** [all] Error 2
We are still working to find a way to compile the program.
[UPDATE] - Finally, i’m working with Ubuntu Operating System. The problem was solved.
I will continue to try to replicate Vanessa’s master’s dissertation to try to close that road.
On the other hand, continue studying classic Reinforcement Learning methods and try the first Deep Reinforcement Learning algorithm in Alberto’s repository,
puppis directory. It would be to solve the game of Pong using the DQN algorithm.
Knowing the different classical methods gives an idea of ćomo work before different problems to find the best option. Although this week I have known better these algorithms, I think that to continue studying them is the way to settle and assimilate all the concepts.