src.baseline.metabbo.symbol

Module Contents

Classes

Critic
Memory
SYMBOL	Introduction The paper “SYMBOL: Generating Flexible Black-Box Optimizers through Symbolic Equation Learning” introduces a novel framework, SYMBOL, designed to automate the discovery of advanced black-box optimizers using symbolic equation learning. Unlike traditional Meta-Black-Box Optimization (MetaBBO) methods that rely on predefined, hand-crafted optimizers, SYMBOL employs a Symbolic Equation Generator (SEG) to dynamically produce closed-form optimization rules tailored to specific tasks and optimization steps. The framework incorporates reinforcement learning-based strategies to efficiently meta-learn these symbolic rules.
LSTM
LSTM_Memory
Tokenizer
MyTokenizer

Functions

get_binary
get_choice
get_c
construct_action
get_mask
which_layer
left_or_right
have_continous_const
continus_mul_c
get_along_continuous_mul
test_pre
get_along_continuous_plus_with_minus
get_along_continuous_plus
along_continuous_plus
find_prefix_of_token_ancestor
get_prefix
get_prefix_with_consts
get_next_position
get_str_prefix
prefix_to_infix
expr_to_func

API

class src.baseline.metabbo.symbol.Critic(fea_dim, value_dim)

Bases: torch.nn.Module

Initialization

forward(x)

class src.baseline.metabbo.symbol.Memory

Initialization

clear_memory()

class src.baseline.metabbo.symbol.SYMBOL(config)

Bases: src.rl.ppo.PPO_Agent

Introduction

The paper “SYMBOL: Generating Flexible Black-Box Optimizers through Symbolic Equation Learning” introduces a novel framework, SYMBOL, designed to automate the discovery of advanced black-box optimizers using symbolic equation learning. Unlike traditional Meta-Black-Box Optimization (MetaBBO) methods that rely on predefined, hand-crafted optimizers, SYMBOL employs a Symbolic Equation Generator (SEG) to dynamically produce closed-form optimization rules tailored to specific tasks and optimization steps. The framework incorporates reinforcement learning-based strategies to efficiently meta-learn these symbolic rules.

Original Paper

“Symbol: Generating Flexible Black-Box Optimizers through Symbolic Equation Learning.” The Twelfth International Conference on Learning Representations. (2024)

Official Implementation

SYMBOL

Application Scenario

single-object optimization problems(SOOP)

Args:

`config`: Configuration object containing all necessary parameters for experiment.For details you can visit config.py.

Attributes:

config (object): Configuration object with various hyperparameters and settings.
tokenizer (MyTokenizer): Tokenizer instance used for processing sequences.
actor (LSTM): Actor network for generating actions.
critic (Critic): Critic network for evaluating states.
optimizer (torch.optim.Optimizer): Optimizer for training the actor and critic networks.
learning_time (int): Counter for the number of learning steps performed.
cur_checkpoint (int): Counter for the current checkpoint during training.

Methods:

__str__():
    Returns the string representation of the class.
train_episode(envs, seeds, para_mode='dummy', compute_resource={}, tb_logger=None, required_info={}):
    Trains the agent for one episode using the provided environments.
    Args:
        envs (list): List of environments for training.
        seeds (Optional[Union[int, List[int], np.ndarray]]): Seeds for environment initialization.
        para_mode (Literal['dummy', 'subproc', 'ray', 'ray-subproc']): Parallelization mode.
        compute_resource (dict): Resources for computation (e.g., CPUs, GPUs).
        tb_logger (object): TensorBoard logger for logging training metrics.
        required_info (dict): Additional information required from the environment.
    Returns:
        Tuple[bool, dict]: A tuple containing a boolean indicating if training has ended and a dictionary with training information.
rollout_episode(env, seed=None, required_info={}):
    Executes a single rollout episode in the given environment.
    Args:
        env (object): Environment for the rollout.
        seed (Optional[int]): Seed for environment initialization.
        required_info (dict): Additional information required from the environment.
    Returns:
        dict: A dictionary containing rollout results.

Returns:

None

Raises:

AssertionError: If a NaN value is found in the loss during training.

Initialization

Initializes the PPO agent with the given configuration, networks, and learning rates.Store the initial agent in the checkpoint directory.

Args:

• config: Configuration object containing all necessary parameters for the experiment.

• networks (dict): A dictionary of neural networks used by the agent.

• learning_rates (float): Learning rate for the optimizer.

__str__()

train_episode(envs, seeds: src.rl.ppo.Optional[src.rl.ppo.Union[int, List[int], numpy.ndarray]], para_mode: src.rl.ppo.Literal[dummy, subproc, ray, ray - subproc] = 'dummy', compute_resource={}, tb_logger=None, required_info={})

rollout_episode(env, seed=None, required_info={})

class src.baseline.metabbo.symbol.LSTM(max_layer, hidden_dim, num_layers, max_c, min_c, fea_dim, c_interval, tokenizer)

Bases: torch.nn.Module

Initialization

forward(x, save_data=False, fix_action=None)

get_random_seq(bs)

class src.baseline.metabbo.symbol.LSTM_Memory

Initialization

get_dict()

src.baseline.metabbo.symbol.get_binary(action, code_len)

src.baseline.metabbo.symbol.get_choice(output, mask, code_len, fix_choice=None)

src.baseline.metabbo.symbol.get_c(output, min_c, interval, fix_c=None)

class src.baseline.metabbo.symbol.Tokenizer

Initialization

SPECIAL_SYMBOLS

None

SPECIAL_FLOAT_SYMBOLS

None

SPECIAL_OPERATORS

None

SPECIAL_INTEGERS

None

abstractmethod encode(expr)

abstractmethod decode(expr)

abstractmethod is_unary(token)

abstractmethod is_binary(token)

abstractmethod is_leaf(token)

get_constant_ids()

class src.baseline.metabbo.symbol.MyTokenizer

Bases: src.baseline.metabbo.symbol.Tokenizer

Initialization

decode(expr)

encode(expr)

is_consts(id)

is_binary(token)

is_unary(token)

is_leaf(token)

is_var(token)

src.baseline.metabbo.symbol.construct_action(seq, const_seq, tokenizer)

src.baseline.metabbo.symbol.get_mask(pre_seq, tokenizer, position, max_layer)

src.baseline.metabbo.symbol.which_layer(position)

src.baseline.metabbo.symbol.left_or_right(position, root)

src.baseline.metabbo.symbol.have_continous_const(seq, position, tokenizer)

src.baseline.metabbo.symbol.continus_mul_c(seq, position, tokenizer)

src.baseline.metabbo.symbol.get_along_continuous_mul(tokenizer, seq, begin, visited)

src.baseline.metabbo.symbol.test_pre(target_vocab, pre_vocab, tokenizer)

src.baseline.metabbo.symbol.get_along_continuous_plus_with_minus(tokenizer, seq, begin, visited)

src.baseline.metabbo.symbol.get_along_continuous_plus(tokenizer, seq, begin, visited)

src.baseline.metabbo.symbol.along_continuous_plus(tokenizer, seq, neg_ancestor)

src.baseline.metabbo.symbol.find_prefix_of_token_ancestor(tokenizer, seq, position, token)

src.baseline.metabbo.symbol.get_prefix(seq, sub_root)

src.baseline.metabbo.symbol.get_prefix_with_consts(seq, consts, sub_root)

src.baseline.metabbo.symbol.get_next_position(seq, choice, position, tokenizer)

src.baseline.metabbo.symbol.get_str_prefix(seq, const_vals, tokenizer)

src.baseline.metabbo.symbol.prefix_to_infix(expr, constants, tokenizer: src.baseline.metabbo.symbol.Tokenizer)

src.baseline.metabbo.symbol.expr_to_func(sympy_expr, variables: List[str])