aind_dynamic_foraging_models.generative_model.params package

Submodules

aind_dynamic_foraging_models.generative_model.params.forager_loss_counting_params module

Generate pydantic models for Loss Count agent parameters.

aind_dynamic_foraging_models.generative_model.params.forager_loss_counting_params.generate_pydantic_loss_counting_params(win_stay_lose_switch: Literal[False, True] = False, choice_kernel: Literal['none', 'one_step', 'full'] = 'none') → Tuple[Type[BaseModel], Type[BaseModel]]

Generate Pydantic models for Loss-counting agent parameters.

All default values are hard-coded in this function. But when instantiating the model, you can always override the default values, both the params and the fitting bounds.

Parameters:

• win_stay_lose_switch (bool, optional) – If True, the agent will be a win-stay-lose-shift agent (loss_count_threshold_mean and loss_count_threshold_std are fixed at 1 and 0), by default False

• choice_kernel (Literal["none", "one_step", "full"], optional) –

  Choice kernel type, by default “none” If “none”, no choice kernel will be included in the model. If “one_step”, choice_kernel_step_size will be set to 1.0, i.e., only the previous choice

  affects the choice kernel. (Bari2019)

  If “full”, both choice_kernel_step_size and choice_kernel_relative_weight will be included

aind_dynamic_foraging_models.generative_model.params.forager_q_learning_params module

Dynamically generate pydantic models for Q-learning agent parameters.

aind_dynamic_foraging_models.generative_model.params.forager_q_learning_params.generate_pydantic_q_learning_params(number_of_learning_rate: Literal[1, 2] = 2, number_of_forget_rate: Literal[0, 1] = 1, choice_kernel: Literal['none', 'one_step', 'full'] = 'none', action_selection: Literal['softmax', 'epsilon-greedy'] = 'softmax') → Tuple[Type[BaseModel], Type[BaseModel]]

Dynamically generate Pydantic models for Q-learning agent parameters.

All default values are hard-coded in this function. But when instantiating the model, you can always override the default values, both the params_fields and the fitting bounds.

Parameters:

• number_of_learning_rate (Literal[1, 2], optional) – Number of learning rates, by default 2 If 1, only one learn_rate will be included in the model. If 2, learn_rate_rew and learn_rate_unrew will be included in the model.

• number_of_forget_rate (Literal[0, 1], optional) – Number of forget_rates, by default 1. If 0, forget_rate_unchosen will not be included in the model. If 1, forget_rate_unchosen will be included in the model.

• choice_kernel (Literal["none", "one_step", "full"], optional) –

  Choice kernel type, by default “none” If “none”, no choice kernel will be included in the model. If “one_step”, choice_kernel_step_size will be set to 1.0, i.e., only the previous choice

  affects the choice kernel. (Bari2019)

  If “full”, both choice_kernel_step_size and choice_kernel_relative_weight will be included

• action_selection (Literal["softmax", "epsilon-greedy"], optional) – Action selection type, by default “softmax”

aind_dynamic_foraging_models.generative_model.params.util module

aind_dynamic_foraging_models.generative_model.params.util.create_pydantic_models_dynamic(params_fields: Dict[str, Any], fitting_bounds: Dict[str, Tuple[float, float]])

Create Pydantic models dynamically based on the input fields and fitting bounds.

aind_dynamic_foraging_models.generative_model.params.util.get_params_options(params_model, default_range=[-inf, inf], para_range_override={}) → dict

Get options for the params fields.

Useful for the Streamlit app.

Parameters:

• params_model (Pydantic model) – The Pydantic model for the parameters.

• default_range (list, optional) – The default range for the parameters, by default [-np.inf, np.inf] If the range is not specified in the Pydantic model, this default range will be used.

• para_range_override (dict, optional) – The range override for user-specified parameters, by default {}

• Example

• ParamsModel (>>>) –

  number_of_learning_rate=1,

  number_of_forget_rate=1, choice_kernel=”one_step”, action_selection=”softmax”,

  )

• generate_pydantic_q_learning_params( (FittingBoundsModel =) –

  number_of_learning_rate=1,

  number_of_forget_rate=1, choice_kernel=”one_step”, action_selection=”softmax”,

  )

• get_params_options(ParamsModel) (>>> params_options =)

• {'learn_rate' ({'para_range': [0.0, 1.0],) – ‘para_default’: 0.5, ‘para_symbol’: <ParamsSymbols.learn_rate: ‘$alpha$’>, ‘para_desc’: ‘Learning rate’}, …

• }

Module contents

Package for params

class aind_dynamic_foraging_models.generative_model.params.ParamsSymbols(value)

Bases: str, Enum

Symbols for the parameters.

The order determined the default order of parameters when output as a string.

biasL = '$b_L$'

choice_kernel_relative_weight = '$w_{ck}$'

choice_kernel_step_size = '$\\alpha_{ck}$'

epsilon = '$\\epsilon$'

forget_rate_unchosen = '$\\delta$'

learn_rate = '$\\alpha$'

learn_rate_rew = '$\\alpha_{rew}$'

learn_rate_unrew = '$\\alpha_{unr}$'

loss_count_threshold_mean = '$\\mu_{LC}$'

loss_count_threshold_std = '$\\sigma_{LC}$'

reset_to_threshold = '$\\mathrm{reset}$'

softmax_inverse_temperature = '$\\beta$'

threshold = '$\\rho$'