utils

Utility and helper functions for the Genius Python SDK.

To import any utility functions run: from genius_client_sdk.utils import <function>

control_map

Maps from an integer dummy encoding of an action to the action's string name.

Parameter	Type	Default	Description
controls	list[str]	Required	A list of strings containing possible actions.
chosen_action	int	Required	An integer encoding of the selected action.

Returns:

Type	Description
str	The string representation of the action.

Example:

controls = ["LEFT", "RIGHT"]
chosen_action = 1

control_map(controls=controls, chosen_action=chosen_action)

# Returns: "RIGHT"

flatten_nested_list

Flatten a nested list.

Parameter	Type	Default	Description
nested_list	list[list]	Required	A nested list with a maximum depth of 1.

Returns:

Type	Description
list	A flattened list.

get_policy_space

Constructs the space of all policies possible for the agent to pursue. Note that this function can be used for multiple observation modalities or state factor which is why the inputs for n_states and n_actions are lists.

Parameter	Type	Default	Description
n_states	list[int]	Required	A list of the number of states for each state factor.
n_actions	list[int]	Required	A list of the number of actions for each state factor.
policy_len	int	Required	Length of policy (i.e. number of actions).
actions	list[str]	Optional	String representation of possible actions.

Returns:

Type	Description
list[list[list[str]]]	Either the string representation of the policy space (if actions are provided) or the dummy encoding of the policy space.

onehot

Generates a one hot numpy array of size size with a 1 inserted in position position.

Parameter	Type	Default	Description
size	int	Required	The size of the array.
position	int	Required	The position to insert a 1.

Returns:

Type	Description
np.ndarray	A one hot numpy array of dimension size with a 1 at the specified position and zeros elsewhere.

plot_categorical

Plot a categorical distribution as a bar plot.

The color parameter must either be a hex string or one of the standard Matplotlib CSS colors.

Parameter	Type	Default	Description
x	list	Required	A list of categories or x-axis values.
y	list	Required	A list of values corresponding to each category in x.
xlabel	str	X	A label for the x-axis.
ylabel	str	Y	A label for the y-axis.
title	str	"no title"	The title of the plot.
color	str	"dodgerblue"	The color of the bars in the plot.
rotation	int	0	The rotation angle of the x-axis labels.

policy_map

Maps from the integer dummy encoding of a policy in the policy space and the policy string.

Parameter	Type	Default	Description
policy_space	list[list[str]]	Required	The The space of possible policies pursued by the agent. Can be obtained by using get_policy_space().
chosen_policy	int	Required	An integer encoding the chosen policy in the policy space.

Returns:

Type	Description
list[str]	A policy list which contains a sequence of actions as a string.

Example:

policy_space = [["LEFT", "RIGHT"], ["RIGHT", "LEFT"], ["UP", "DOWN"]]
chosen_policy = 2

policy_map(policy_space=policy_space, chosen_policy=chosen_policy)

# Returns: ["UP", "DOWN"]