{
  "id": "2210.15515",
  "version": "v1",
  "published": "2022-10-27T14:54:06.000Z",
  "updated": "2022-10-27T14:54:06.000Z",
  "title": "Meta-Reinforcement Learning Using Model Parameters",
  "authors": [
    "Gabriel Hartmann",
    "Amos Azaria"
  ],
  "comment": "8 pages",
  "categories": [
    "cs.LG"
  ],
  "abstract": "In meta-reinforcement learning, an agent is trained in multiple different environments and attempts to learn a meta-policy that can efficiently adapt to a new environment. This paper presents RAMP, a Reinforcement learning Agent using Model Parameters that utilizes the idea that a neural network trained to predict environment dynamics encapsulates the environment information. RAMP is constructed in two phases: in the first phase, a multi-environment parameterized dynamic model is learned. In the second phase, the model parameters of the dynamic model are used as context for the multi-environment policy of the model-free reinforcement learning agent.",
  "revisions": [
    {
      "version": "v1",
      "updated": "2022-10-27T14:54:06.000Z"
    }
  ],
  "analyses": {
    "keywords": [
      "model parameters",
      "meta-reinforcement learning",
      "predict environment dynamics encapsulates",
      "model-free reinforcement learning agent",
      "multi-environment parameterized dynamic model"
    ],
    "note": {
      "typesetting": "TeX",
      "pages": 8,
      "language": "en",
      "license": "arXiv",
      "status": "editable"
    }
  }
}