Module moog_demos.example_configs.red_green

Red-Green Task.

This task is a variant of the task used in this paper: Smith, K. A., Peres, F., Vul, E., & Tenebaum, J. (2017). Thinking inside the box: Motion prediction in contained spaces uses simulation. In CogSci.

In this task, there is a blue ball that bounces in an enclosed rectangular arena. The arena may have gray rectangular obstacles that the blue ball bounces off. The arena has one green box and one red box. The subject's goal is to predict which of the green/red boxes the blue ball will contact first.

In this particular implementation, the subject moves a token at the bottom of the screen left or right to indicate its choice.

The main entrypoint is the get_config() function at the bottom of this file.

Functions

def get_config(level)

Expand source code

def get_config(level):
    """Get config to pass to environment constructor.

    Args:
        level: Int. Number of obstacles in arena.
    """
    if not isinstance(level, int):
        raise ValueError(f'level is {level}, but must be an integer.')
    return _get_config(num_obstacles=level, valid_step_range=(50, 150))

Get config to pass to environment constructor.

Args

level

Int. Number of obstacles in arena.

Classes

class RadialVelocity (speed)

Expand source code

class RadialVelocity(distribs.AbstractDistribution):
    """Radial velocity distribution."""

    def __init__(self, speed):
        """Constructor.

        Args:
            speed: Float. Speed of the sampled velocities.
        """
        self._speed = speed

    def sample(self, rng):
        rng = self._get_rng(rng)
        theta = rng.uniform(0., 2 * np.pi)
        x_vel = self._speed * np.cos(theta)
        y_vel = self._speed * np.sin(theta)
        return {'x_vel': x_vel, 'y_vel': y_vel}

    def contains(self, spec):
        if 'x_vel' not in spec or 'y_vel' not in spec:
            return False

        vel_norm = np.linalg.norm([spec['x_vel'], spec['y_vel']])
        if np.abs(vel_norm - self._speed) < _EPSILON:
            return True
        else:
            return False

    def to_str(self, indent):
        s = 'RadialVelocity({})'.format(self._speed)
        return indent * '  ' + s

    @property
    def keys(self):
        return set(['x_vel', 'y_vel'])

Radial velocity distribution.

Constructor.

Args

speed

Float. Speed of the sampled velocities.

Ancestors

• moog.state_initialization.distributions.AbstractDistribution
• abc.ABC

Instance variables

prop keys

Expand source code

@property
def keys(self):
    return set(['x_vel', 'y_vel'])

The set of keys in specs sampled from this distribution.

Methods

def contains(self, spec)

Expand source code

def contains(self, spec):
    if 'x_vel' not in spec or 'y_vel' not in spec:
        return False

    vel_norm = np.linalg.norm([spec['x_vel'], spec['y_vel']])
    if np.abs(vel_norm - self._speed) < _EPSILON:
        return True
    else:
        return False

Return whether distribution contains spec dictionary.

def sample(self, rng)

Expand source code

def sample(self, rng):
    rng = self._get_rng(rng)
    theta = rng.uniform(0., 2 * np.pi)
    x_vel = self._speed * np.cos(theta)
    y_vel = self._speed * np.sin(theta)
    return {'x_vel': x_vel, 'y_vel': y_vel}

Sample a spec from this distribution. Returns a dictionary.

Args

rng

Random number generator. Fed into self._get_rng(), if None defaults to np.random.

def to_str(self, indent)

Expand source code

def to_str(self, indent):
    s = 'RadialVelocity({})'.format(self._speed)
    return indent * '  ' + s

Recursive string description of this distribution.