sample

Sampling methods.

gym_socks.sampling.sample.grid_sampler(grid_points)

Sample from a set of pre-defined grid points.

Hint

Use boxgrid() or cartesian() from gym_socks.utils.grid to generate a grid of points.

Parameters

grid_points (numpy.ndarray) – A collection of grid points in a list or array.

Yields

A point in the grid. Note that points are yielded in the order they are given.

Example

>>> import numpy as np
>>> from gym_socks.utils.grid import cartesian
>>> from gym_socks.envs.sample import grid_sampler
>>> grid = cartesian(np.linspace(-1, 1, 3), np.linspace(-2, 2, 3))
>>> sampler = grid_sampler(grid)
>>> S = sampler.sample(size=100)

gym_socks.sampling.sample.observation_sampler(env, state_sampler, action_sampler)

Observation sampler.

Parameters

env (gym.Env) –

gym_socks.sampling.sample.sample_fn(fn)

Sample function decorator.

Converts a sample function into a sample generator.

Parameters

fn – Sample function that returns or yields an observation.

Returns

A function that can be used to islice a sample from the sample generator.

Example

>>> from gym_socks.envs.sample import sample_fn
>>> @sample_fn
... def custom_sampler(env, policy):
...     state = env.reset()
...     action = policy(state=state)
...     observation, *_ = env.step(action)
...     yield state, action, observation
>>> sampler = custom_sampler(env, policy)
>>> S = sampler.sample(size=100)

gym_socks.sampling.sample.space_sampler(space)

Randomly sample from a space.

Parameters

space (gym.spaces.Space) – A gym.space.Space that implements a sample() function.

Yields

A random sample from the space.

Example

>>> from gym.spaces import Box
>>> from gym_socks.envs.sample import space_sampler
>>> sampler = space_sampler(Box(low=-1, high=1, shape=(2,), dtype=float))
>>> S = sampler.sample(size=100)

gym_socks.sampling.sample.trajectory_sampler(env, state_sampler, action_sampler, time_horizon=1)

Trajectory sampler.

Parameters

• env (gym.Env) –

• time_horizon (int) –

gym_socks.sampling.sample.transition_sampler(env, state_sampler, action_sampler)

Transition sampler.

Parameters

env (gym.Env) –