ml-agents/docs/Training-Curriculum-Learning.md

Training with Curriculum Learning

Sample Environment

Imagine a task in which an agent needs to scale a wall to arrive at a goal. The starting point when training an agent to accomplish this task will be a random policy. That starting policy will have the agent running in circles, and will likely never, or very rarely scale the wall properly to the achieve the reward. If we start with a simpler task, such as moving toward an unobstructed goal, then the agent can easily learn to accomplish the task. From there, we can slowly add to the difficulty of the task by increasing the size of the wall, until the agent can complete the initially near-impossible task of scaling the wall. We are including just such an environment with the ML-Agents toolkit 0.2, called Wall Jump.

Demonstration of a curriculum training scenario in which a progressively taller wall obstructs the path to the goal.

To see this in action, observe the two learning curves below. Each displays the reward over time for an agent trained using PPO with the same set of training hyperparameters. The difference is that one agent was trained using the full-height wall version of the task, and the other agent was trained using the curriculum version of the task. As you can see, without using curriculum learning the agent has a lot of difficulty. We think that by using well-crafted curricula, agents trained using reinforcement learning will be able to accomplish tasks otherwise much more difficult.

How-To

Each group of Agents under the same Behavior Name in an environment can have a corresponding curriculum. These curriculums are held in what we call a metacurriculum. A metacurriculum allows different groups of Agents to follow different curriculums within the same environment.

Specifying a Metacurriculum

We first create a folder inside config/curricula/ for the environment we want to use curriculum learning with. For example, if we were creating a metacurriculum for Wall Jump, we would create the folder config/curricula/wall-jump/. We will place our curriculums inside this folder.

Specifying a Curriculum

In order to define a curriculum, the first step is to decide which parameters of the environment will vary. In the case of the Wall Jump environment, what varies is the height of the wall. We define this as a Shared Float Property that can be accessed in Academy.FloatProperties, and by doing so it becomes adjustable via the Python API. Rather than adjusting it by hand, we will create a JSON file which describes the structure of the curriculum. Within it, we can specify which points in the training process our wall height will change, either based on the percentage of training steps which have taken place, or what the average reward the agent has received in the recent past is. Below is an example curriculum for the BigWallBehavior in the Wall Jump environment.

{
    "measure" : "progress",
    "thresholds" : [0.1, 0.3, 0.5],
    "min_lesson_length" : 100,
    "signal_smoothing" : true,
    "parameters" :
    {
        "big_wall_min_height" : [0.0, 4.0, 6.0, 8.0],
        "big_wall_max_height" : [4.0, 7.0, 8.0, 8.0]
    }
}

• measure - What to measure learning progress, and advancement in lessons by.

  • reward - Uses a measure received reward.
  • progress - Uses ratio of steps/max_steps.

• thresholds (float array) - Points in value of measure where lesson should be increased.

• min_lesson_length (int) - The minimum number of episodes that should be completed before the lesson can change. If measure is set to reward, the average cumulative reward of the last min_lesson_length episodes will be used to determine if the lesson should change. Must be nonnegative.

  Important: the average reward that is compared to the thresholds is different than the mean reward that is logged to the console. For example, if min_lesson_length is 100, the lesson will increment after the average cumulative reward of the last 100 episodes exceeds the current threshold. The mean reward logged to the console is dictated by the summary_freq parameter in the trainer configuration file.

• signal_smoothing (true/false) - Whether to weight the current progress measure by previous values.

  • If true, weighting will be 0.75 (new) 0.25 (old).

• parameters (dictionary of key:string, value:float array) - Corresponds to Academy reset parameters to control. Length of each array should be one greater than number of thresholds.

Once our curriculum is defined, we have to use the reset parameters we defined and modify the environment from the Agent's AgentReset() function. See WallJumpAgent.cs for an example.

We will save this file into our metacurriculum folder with the name of its corresponding Behavior Name. For example, in the Wall Jump environment, there are two different Behaviors Name set via script in WallJumpAgent.cs ---BigWallBrainLearning and SmallWallBrainLearning. If we want to define a curriculum for the BigWallBrainLearning, we will save BigWallBrainLearning.json into config/curricula/wall-jump/.

Training with a Curriculum

Once we have specified our metacurriculum and curriculums, we can launch mlagents-learn using the –curriculum flag to point to the metacurriculum folder and PPO will train using Curriculum Learning. For example, to train agents in the Wall Jump environment with curriculum learning, we can run

mlagents-learn config/trainer_config.yaml --curriculum=config/curricula/wall-jump/ --run-id=wall-jump-curriculum --train

We can then keep track of the current lessons and progresses via TensorBoard.