Profile cover photo
Profile photo
Massimiliano Patacchiola
154 followers -
#include<smartquotes.h>
#include<smartquotes.h>

154 followers
About
Communities and Collections
View all
Posts

Post has attachment

Hello Machine Learner! I published the eighth post of the "Dissecting Reinforcement Learning" series. In this episode I show how to use a feedforward neural network as function approximator in Reinforcement Learning. There is also ready-to-run Python code on the official GitHub repository of the project: https://github.com/mpatacchiola/dissecting-reinforcement-learning

Cheers!


https://mpatacchiola.github.io/blog/2018/12/28/dissecting-reinforcement-learning-8.html

Post has attachment

Hello folks! After almost one year of inactivity I finally found the time to release the eight episode of the "Dissecting Reinforcement Learning" series. We will see how to use a feedforward neural network as function approximator in RL. A generic Multi Layer Perceptron class built from scratch in Numpy is provided in the GitHub repository:

https://github.com/mpatacchiola/dissecting-reinforcement-learning

https://mpatacchiola.github.io/blog/2018/12/28/dissecting-reinforcement-learning-8.html

Post has attachment
Hello folks! The seventh episode of the "Dissecting Reinforcement Learning" series is out there. This time I (gently) introduce linear approximators, starting from a biological perspective and discussing the Grandmother's cell. As usual with Python code ready to run! Share if you like....

Post has attachment
Dear fellow I am back with the seventh episode of the "Dissecting Reinforcemnet Learning" series. This time I will gently intorduce linear function approximation starting from an intuitive introduction and a discussion of the "grandmother" cells. We are getting ready for using Neural Networks (in the next post). As usual there is ready-to-use Python code on the related repository on GitHub: https://github.com/mpatacchiola/dissecting-reinforcement-learning

Post has attachment
The sixth episode of the "Dissecting Reinforcement Learning" series is available on my Blog. In this episode get your hands dirty! How to use Reinforcement Learning for real applications: Multi-Armed Bandit, Drone Landing, Pole Balancing, etc. It includes Python code ready to run which can produce animations and video of the agent during training, check out the post and the GitHub repository:

https://github.com/mpatacchiola/dissecting-reinforcement-learning

Post has attachment
The sixth episode of the "Dissecting Reinforcement Learning" Blog series is out there. In this episode you will learn how to autonomously land a Drone on a platform, how to control a pole balancing robot and a mountain car. Moreover there is a lot of Python code ready to run which allows saving animated gif and videos of the agent...

Give a look to the official Github repository:

https://github.com/mpatacchiola/dissecting-reinforcement-learning

Post has attachment
Dear friends, the sixth episode of the "Dissecting Reinforcement Learning" series has been published. In this post I show how to tacke real problems using SARSA, Q-Learning, Monte Carlo Methods, Thompson Sampling, etc. Moreover there is a lot of Python code (ready to run) which allows creating plots, logs and animations. Have fun!

Post has attachment
Dear fellows, the sixth episode of the "Dissecting Reinforcement Learning" series is out there. In this episode get your hands dirty! Let's see how to use SARSA, Monte Carlo methods, and Q-Learning to solve problems such as: Multi-Armed Bandit, Mountain-Car, Inverted Pendulum, Drone Landing, etc. Moreover on the official repository there is a lot of Python code ready to run:

https://github.com/mpatacchiola/dissecting-reinforcement-learning

Post has attachment
The first episode of the BBC series "Hyper evolution: the rise of the robots" has been released on the official website. In this episode Ben Garrod visits our lab and we show him what the iCub humanoid robot can do!
Add a comment...

Post has attachment
Dear all, I would like to share with you the link to our recent work on head pose estimation based on Deep Neural Networks, which has been recently accepted on Pattern Recognition. The PDF can be downloaded for free in the next 50 days:

https://authors.elsevier.com/a/1VBdC77nKOnOt

The CNN used in this work has been implemented in the Deepgaze library:

https://github.com/mpatacchiola/deepgaze
Photo
Wait while more posts are being loaded