Dead or Alive ++

Street Fighter III

Tekken Tag Tournament

Ultimate Mortal Kombat

A.I. VS COM

A.I. algorithms fighting for the score record and to complete the game

A.I. VS A.I.

A.I. algorithms fighting one against the other

A.I. VS Human

A.I. algorithms fighting against human players

Imitation learning

Imitation learning is a technique that, given a collection of examples (usually provided by humans) representing a "desirable behavior" in solving a specific task, uses it to train an agent, with the aim of making it learn how to achieve the same human-level performances. It can be used, for example, as a starting point to speed-up training, providing a reference baseline from where the agent can start experimenting and exploring. With the aim of demonstrating as many interesting RL techniques as possible, we equipped DIAMBRA with Imitation Learning capabilities, we will show and walk you through them during our live events.

Self-play

While developing DIAMBRA, we wanted to make available most advanced and interesting Reinforcement Learning techniques, and self-play is for sure one of them. The platform allows the agent to play on both sides at the same time, being the challenger of himself. This technique is known to improve training, in particular because the agent will continuously face an opponent of the same ability level and also because for a single environment step it collects twice as much information.

The story behind DIAMBRA name

Naples, 1552. Two women, Isabella of Carazzi and DIAMBRA of Pottinella, in the presence of the Marquis of the Vast, dispute in a duel for the love of a man called Fabio Zeresola. This legendary episode, subject matter of the painting "Women Gladiators", by Jusepe de Ribera, Naples 1636, has also been held to be an allegory of the fight between Vice and Virtue.