REINFORCE Benchmark

Proposed by Williams in 1992, REINFORCE is one of the oldest algorithms, but is still popular. For example, it is used in Google's AlphaGo and its descendants.

TODO clear below; link original paper

Name: Reinforce Benchmark

Date completed: 2018_02_04_012905, ongoing

Description: Standard Benchmark on Reinforce

Hypotheses: N/A

Prerequisites: N/A

Algorithms: Reinforce

Environments: CartPole-v0, Acrobot-v1

Specs: ("reinforce.json", "reinforce_benchmark")

Running instructions:

{
  "reinforce.json": {
    "reinforce_benchmark": "benchmark"
  }
}

Commit: c4538fc9c6e6cd5f1fb91ba742d95225ca4ad4a1

Results summary:

data: Reinforce_CartPole-v0_2018_02_04_012905