Students Session

Are you working in your MSc or PhD dissertation? Do you want to present your ongoing work to the AI4G school lecturers, and get some valuable insight on your work?

If so, drop us an e-mail with your name, affiliation, supervisor, your (tentative) dissertation title and a brief summary of your goals. We'll give you 5 to 10 minutes to present your work!

Note: Students from outside of IPCA that are pesenting their work on the Students Session can register with the lowest registration fee!

Artificial Intelligence on AAA games: Applied AI in Empire of Sin, a modern strategy game

Miguel Campião

This talk will present a high level overview of the different AI systems implemented in Empire of Sin game. It will then focus on one of those systems with more detail, giving more emphasis to scenarios, plans and behaviors, used in in-world characters and missions.

How to apply to a Digital Games industry position?

Miguel Campião

This presentation will focus on the process of applying for a job or position in the Digital Games Industry. We will discuss some of the different things that companies are looking for in a prospective employee and how one could focus on what actually matters and what doesn't.

Runtime Procedural Generation: Real-life cases and issues

Tiago Loureiro

From modern cities to 2D game levels, runtime procedural generation has been used throughout multiple industries to create seamless and tailor-made experiences to users from multiple backgrounds. In this session we will look at some representative examples and show different approaches to creating these systems.

Tiago Loureiro

Advances in computational power have enabled more complex decision structures and pathways. Despite the historical dominance of Finite State Machines, commercial products have started using other real-time decision support strategies and techniques like Dynamic Scripting or Behavior Trees. In this session, we will discuss the appropriate usage of such systems and watch realtime examples of said applications.

From traditional Q-Learning to Deep Reinforcement Learning: an Introduction.

João Dias

This talk will depart from traditional Reinforcement Learning approaches to introduce key concepts of Deep Reinforcement Learning, and briefly explain the conceptual ideas behind several state of the art algorithms that achieved outstanding results in the last few years: DQN, A3C and PPO.

Tutorial in Deep Reinforcement Learning with Unity Machine Learning Agents Toolkit.

João Dias

This tutorial explores Unity 3D's Machine Learning Agents Toolkit and its application to a concrete game-like scenario. It will begin with an introduction to the ML Toolkit, and how to define an agent, the agent's actions and observations and its corresponding reward function. Afterwards, we will learn how to analyse the TensorBoard which displays important information about the training process, to decide on how to adjust learning parameters that will affect the learning process. Finally, we will look at the application of the Curriculum Learning technique.

Infusing life into Watch Dogs

Vincent Breton

This presentation will discuss how we brought life to Chicago and San Francisco in Watch Dogs 1 and 2. We will look into the different systems and tricks used to create a believable population that moves around like real human beings.

An Introduction to Story Generation

Gabriella Barros

Story generation is a vast and diverse field of research. There are several different approaches to procedural story generation, each with its advantages and disadvantages. This talk will present a non-exhaustive overview of different techniques used for automatically or semi-automatically generating stories, focusing on general methods that can be applied to interactive and/or non-interactive mediums.