PLATO is a model architecture that identifies objects in its environment through a process of object individuation. It uses this knowledge to outperform its flat model counterparts when it comes to building a "common sense" understanding of its physical environment. The models are compared using the "Violation of Expectation" formalism; a concept inspired by developmental psychology.
📚 References
• Original publication: [ Ссылка ]
🎞 Checkout more videos
• GATO: An introduction: [ Ссылка ]
• When should agents explore?: [ Ссылка ]
• Discovering reinforcement learning algorithms: [ Ссылка ]
📱 Social media
[ Ссылка ]
[ Ссылка ]
#deepmind #agents #research #reinforcementlearning #NiklasOPF #NiklasOPF #Phd #Scientists #Theory #Research #ResearchPapers #ComputerScience #PLATO
![](https://s2.save4k.ru/pic/WFMuq0U5NJk/mqdefault.jpg)