Exploring MADDPG algorithm from OpeanAI to solve environments with multiple agents.

By Antonio Lisi

Intro

Hello everyone, we’re coming back to solving reinforcement learning environments after having a little fun exercising with classic deep learning applications.

Today we’re going to solve an environment with multiple agents using OpenAI’s MADDPG algorithm. This is going to be different and at the same time similar to what we’ve seen so far. With multiple agents we’re adding complexity and computational requirements but, as we’ll see, we’re going to use what we’ve learned solving single-agent environments.

As always, we’re going to implement everything from scratch using…


Using the IMDB dataset to train a movie recommendation engine

By Antonio Lisi

Intro

Hello everyone, we continue our series on how to train algorithms from scratch in classic deep learning applications.

In this post, we will see how to create from scratch and interpret a recommendation engine using a collaborative filtering algorithm.

A Recommendation engine is a tool that predicts what a user may or may not like. It’s typically used when you have a large number of users and products, and you want to recommend which products are most likely to be useful for which users.

There are several…


Unleashing the power of transfer learning using Hugging Face library

By Antonio Lisi

Intro

Hello everyone, after my previous post on “How to create a State of the Art Image Classifier in less than 10 minutes”, I decided to do something similar for the NLP space.

So in this post, we’re going to see how to easily create a state-of-the-art model classifier using the Hugging Face library and exploiting the power of transfer learning as we did for the Rover Classification challenge.

Dataset

The Dataset that we’re going to use to test our models is the classic IMDB dataset that we downloaded…


Teaching a robot to walk

By Antonio Lisi
To my grandma, I’ll miss you

Intro

Hello everyone, in this post, we’re going to teach a robot to walk using one of the latest state-of-the-art algorithms, called SAC.

As always, we implement everything from scratch using Tensorflow 2. We’ll use a lot of the code developed in the post about DDPG. So if you didn’t read it, I recommend doing it before going forward.

Environment

In the DDPG post, we solved two environments provided by OpenAI. But they were too easy, I wanted to try more challenging continuous action spaces environments.

One of…


Diving deep into Reinforcement Learning

By Antonio Lisi

Intro

Hello everyone, this is the third post on reinforcement learning and the start of a new series that is focusing on continuous action environments.

In this post, we will implement DDPG from scratch, and we will try to solve Pendulum and Lunar Lander.

Why the Continuous Action Spaces?

Before proceeding further, we need to motivate why we’re moving to solve continuous space environments. Until now, we always solved pong, but you could use the algorithms, the approach, and even all the code to solve any other Atari game.

But many real-world applications of reinforcement learning require an…


Continuing my journey into Reinforcement Learning

By Antonio Lisi

Intro

Hello everyone, this is my second article on medium, and it’s the sequel of the previous article. We will use a lot of the code developed in the first part. So if you didn’t read it, I recommend doing it before going forward.

In this post, we will implement A2C and PPO from scratch to beat the atari pong game, as we did in the first part with DDDQN.

Motivating A2C and PPO

Before going any further, we need to discuss why we’re focusing on these two algorithms. First of all, both belong to the Policy gradient family of algorithms. While…


The start of my journey in Reinforcement Learning

By Antonio Lisi

Intro

Hello everyone, this is my first technical post on this blog. I hope you’ll enjoy it.

Today we talk about reinforcement learning. In the last few months, with the lockdown and so on, I had for the first time in years some free time, so I decided to study reinforcement learning (RL). Game theory was one of my favorite subjects at the university, but at the same time, I always thought that it couldn’t be applied in real life as it was. …

Antonio Lisi

Data scientist by trade, I develop and deploy machine learning models working in different industries like finance, energy, insurance, etc.

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store