PySR / pysr /sr.py
MilesCranmer's picture
Meta-program branching table rather than array of operators
e2ae8ef
raw
history blame
19.3 kB
import os
from argparse import ArgumentParser, ArgumentDefaultsHelpFormatter
from collections import namedtuple
import pathlib
import numpy as np
import pandas as pd
import sympy
from sympy import sympify, Symbol, lambdify
import subprocess
global_equation_file = 'hall_of_fame.csv'
global_n_features = None
global_variable_names = []
global_extra_sympy_mappings = {}
sympy_mappings = {
'div': lambda x, y : x/y,
'mult': lambda x, y : x*y,
'plus': lambda x, y : x + y,
'sub': lambda x, y : x - y,
'neg': lambda x : -x,
'pow': lambda x, y : sympy.sign(x)*abs(x)**y,
'cos': lambda x : sympy.cos(x),
'sin': lambda x : sympy.sin(x),
'tan': lambda x : sympy.tan(x),
'cosh': lambda x : sympy.cosh(x),
'sinh': lambda x : sympy.sinh(x),
'tanh': lambda x : sympy.tanh(x),
'exp': lambda x : sympy.exp(x),
'acos': lambda x : sympy.acos(x),
'asin': lambda x : sympy.asin(x),
'atan': lambda x : sympy.atan(x),
'acosh':lambda x : sympy.acosh(x),
'asinh':lambda x : sympy.asinh(x),
'atanh':lambda x : sympy.atanh(x),
'abs': lambda x : abs(x),
'mod': lambda x, y : sympy.Mod(x, y),
'erf': lambda x : sympy.erf(x),
'erfc': lambda x : sympy.erfc(x),
'logm': lambda x : sympy.log(abs(x)),
'logm10':lambda x : sympy.log10(abs(x)),
'logm2': lambda x : sympy.log2(abs(x)),
'log1p': lambda x : sympy.log(x + 1),
'floor': lambda x : sympy.floor(x),
'ceil': lambda x : sympy.ceil(x),
'sign': lambda x : sympy.sign(x),
'round': lambda x : sympy.round(x),
}
def pysr(X=None, y=None, weights=None,
procs=4,
populations=None,
niterations=100,
ncyclesperiteration=300,
binary_operators=["plus", "mult"],
unary_operators=["cos", "exp", "sin"],
alpha=0.1,
annealing=True,
fractionReplaced=0.10,
fractionReplacedHof=0.10,
npop=1000,
parsimony=1e-4,
migration=True,
hofMigration=True,
shouldOptimizeConstants=True,
topn=10,
weightAddNode=1,
weightInsertNode=3,
weightDeleteNode=3,
weightDoNothing=1,
weightMutateConstant=10,
weightMutateOperator=1,
weightRandomize=1,
weightSimplify=0.01,
perturbationFactor=1.0,
nrestarts=3,
timeout=None,
extra_sympy_mappings={},
equation_file='hall_of_fame.csv',
test='simple1',
verbosity=1e9,
maxsize=20,
fast_cycle=False,
maxdepth=None,
variable_names=[],
batching=False,
batchSize=50,
select_k_features=None,
warmupMaxsize=0,
threads=None, #deprecated
julia_optimization=3,
):
"""Run symbolic regression to fit f(X[i, :]) ~ y[i] for all i.
Note: most default parameters have been tuned over several example
equations, but you should adjust `threads`, `niterations`,
`binary_operators`, `unary_operators` to your requirements.
:param X: np.ndarray or pandas.DataFrame, 2D array. Rows are examples,
columns are features. If pandas DataFrame, the columns are used
for variable names (so make sure they don't contain spaces).
:param y: np.ndarray, 1D array. Rows are examples.
:param weights: np.ndarray, 1D array. Each row is how to weight the
mean-square-error loss on weights.
:param procs: int, Number of processes (=number of populations running).
:param populations: int, Number of populations running; by default=procs.
:param niterations: int, Number of iterations of the algorithm to run. The best
equations are printed, and migrate between populations, at the
end of each.
:param ncyclesperiteration: int, Number of total mutations to run, per 10
samples of the population, per iteration.
:param binary_operators: list, List of strings giving the binary operators
in Julia's Base, or in `operator.jl`.
:param unary_operators: list, Same but for operators taking a single `Float32`.
:param alpha: float, Initial temperature.
:param annealing: bool, Whether to use annealing. You should (and it is default).
:param fractionReplaced: float, How much of population to replace with migrating
equations from other populations.
:param fractionReplacedHof: float, How much of population to replace with migrating
equations from hall of fame.
:param npop: int, Number of individuals in each population
:param parsimony: float, Multiplicative factor for how much to punish complexity.
:param migration: bool, Whether to migrate.
:param hofMigration: bool, Whether to have the hall of fame migrate.
:param shouldOptimizeConstants: bool, Whether to numerically optimize
constants (Nelder-Mead/Newton) at the end of each iteration.
:param topn: int, How many top individuals migrate from each population.
:param nrestarts: int, Number of times to restart the constant optimizer
:param perturbationFactor: float, Constants are perturbed by a max
factor of (perturbationFactor*T + 1). Either multiplied by this
or divided by this.
:param weightAddNode: float, Relative likelihood for mutation to add a node
:param weightInsertNode: float, Relative likelihood for mutation to insert a node
:param weightDeleteNode: float, Relative likelihood for mutation to delete a node
:param weightDoNothing: float, Relative likelihood for mutation to leave the individual
:param weightMutateConstant: float, Relative likelihood for mutation to change
the constant slightly in a random direction.
:param weightMutateOperator: float, Relative likelihood for mutation to swap
an operator.
:param weightRandomize: float, Relative likelihood for mutation to completely
delete and then randomly generate the equation
:param weightSimplify: float, Relative likelihood for mutation to simplify
constant parts by evaluation
:param timeout: float, Time in seconds to timeout search
:param equation_file: str, Where to save the files (.csv separated by |)
:param test: str, What test to run, if X,y not passed.
:param maxsize: int, Max size of an equation.
:param maxdepth: int, Max depth of an equation. You can use both maxsize and maxdepth.
maxdepth is by default set to = maxsize, which means that it is redundant.
:param fast_cycle: bool, (experimental) - batch over population subsamples. This
is a slightly different algorithm than regularized evolution, but does cycles
15% faster. May be algorithmically less efficient.
:param variable_names: list, a list of names for the variables, other
than "x0", "x1", etc.
:param batching: bool, whether to compare population members on small batches
during evolution. Still uses full dataset for comparing against
hall of fame.
:param batchSize: int, the amount of data to use if doing batching.
:param select_k_features: (None, int), whether to run feature selection in
Python using random forests, before passing to the symbolic regression
code. None means no feature selection; an int means select that many
features.
:param warmupMaxsize: int, whether to slowly increase max size from
a small number up to the maxsize (if greater than 0).
If greater than 0, says how many cycles before the maxsize
is increased.
:param julia_optimization: int, Optimization level (0, 1, 2, 3)
:returns: pd.DataFrame, Results dataframe, giving complexity, MSE, and equations
(as strings).
"""
if threads is not None:
raise ValueError("The threads kwarg is deprecated. Use procs.")
if maxdepth is None:
maxdepth = maxsize
if isinstance(X, pd.DataFrame):
variable_names = list(X.columns)
X = np.array(X)
use_custom_variable_names = (len(variable_names) != 0)
# Check for potential errors before they happen
assert len(unary_operators) + len(binary_operators) > 0
assert len(X.shape) == 2
assert len(y.shape) == 1
assert X.shape[0] == y.shape[0]
if weights is not None:
assert len(weights.shape) == 1
assert X.shape[0] == weights.shape[0]
if use_custom_variable_names:
assert len(variable_names) == X.shape[1]
if select_k_features is not None:
selection = run_feature_selection(X, y, select_k_features)
print(f"Using features {selection}")
X = X[:, selection]
if use_custom_variable_names:
variable_names = variable_names[selection]
if populations is None:
populations = procs
rand_string = f'{"".join([str(np.random.rand())[2] for i in range(20)])}'
if isinstance(binary_operators, str): binary_operators = [binary_operators]
if isinstance(unary_operators, str): unary_operators = [unary_operators]
if X is None:
if test == 'simple1':
eval_str = "np.sign(X[:, 2])*np.abs(X[:, 2])**2.5 + 5*np.cos(X[:, 3]) - 5"
elif test == 'simple2':
eval_str = "np.sign(X[:, 2])*np.abs(X[:, 2])**3.5 + 1/(np.abs(X[:, 0])+1)"
elif test == 'simple3':
eval_str = "np.exp(X[:, 0]/2) + 12.0 + np.log(np.abs(X[:, 0])*10 + 1)"
elif test == 'simple4':
eval_str = "1.0 + 3*X[:, 0]**2 - 0.5*X[:, 0]**3 + 0.1*X[:, 0]**4"
elif test == 'simple5':
eval_str = "(np.exp(X[:, 3]) + 3)/(np.abs(X[:, 1]) + np.cos(X[:, 0]) + 1.1)"
X = np.random.randn(100, 5)*3
y = eval(eval_str)
print("Running on", eval_str)
pkg_directory = '/'.join(__file__.split('/')[:-2] + ['julia'])
def_hyperparams = ""
# Add pre-defined functions to Julia
for op_list in [binary_operators, unary_operators]:
for i in range(len(op_list)):
op = op_list[i]
if '(' not in op:
continue
def_hyperparams += op + "\n"
# Cut off from the first non-alphanumeric char:
first_non_char = [
j for j in range(len(op))
if not (op[j].isalpha() or op[j].isdigit())][0]
function_name = op[:first_non_char]
op_list[i] = function_name
def_hyperparams += f"""include("{pkg_directory}/operators.jl")
const binops = {'[' + ', '.join(binary_operators) + ']'}
const unaops = {'[' + ', '.join(unary_operators) + ']'}
const ns=10;
const parsimony = {parsimony:f}f0
const alpha = {alpha:f}f0
const maxsize = {maxsize:d}
const maxdepth = {maxdepth:d}
const fast_cycle = {'true' if fast_cycle else 'false'}
const migration = {'true' if migration else 'false'}
const hofMigration = {'true' if hofMigration else 'false'}
const fractionReplacedHof = {fractionReplacedHof}f0
const shouldOptimizeConstants = {'true' if shouldOptimizeConstants else 'false'}
const hofFile = "{equation_file}"
const nprocs = {procs:d}
const npopulations = {populations:d}
const nrestarts = {nrestarts:d}
const perturbationFactor = {perturbationFactor:f}f0
const annealing = {"true" if annealing else "false"}
const weighted = {"true" if weights is not None else "false"}
const batching = {"true" if batching else "false"}
const batchSize = {min([batchSize, len(X)]) if batching else len(X):d}
const useVarMap = {"true" if use_custom_variable_names else "false"}
const mutationWeights = [
{weightMutateConstant:f},
{weightMutateOperator:f},
{weightAddNode:f},
{weightInsertNode:f},
{weightDeleteNode:f},
{weightSimplify:f},
{weightRandomize:f},
{weightDoNothing:f}
]
const warmupMaxsize = {warmupMaxsize:d}
"""
op_runner = ""
if len(binary_operators) > 0:
op_runner += f"""
@inline function BINOP(i::Int, x::Float32, y::Float32)::Float32
if i == 1
return @fastmath {binary_operators[0]}(x, y)
"""
for i in range(1, len(binary_operators)):
op_runner += f"""
elseif i == {i+1}
return @fastmath {binary_operators[i]}(x, y)
"""
op_runner += """
end
end
"""
if len(unary_operators) > 0:
op_runner += f"""
@inline function UNAOP(i::Int, x::Float32)::Float32
if i == 1
return @fastmath {unary_operators[0]}(x)
"""
for i in range(1, len(unary_operators)):
op_runner += f"""
elseif i == {i+1}
return @fastmath {unary_operators[i]}(x)
"""
op_runner += """
end
end
"""
def_hyperparams += op_runner
if X.shape[1] == 1:
X_str = 'transpose([' + str(X.tolist()).replace(']', '').replace(',', '').replace('[', '') + '])'
else:
X_str = str(X.tolist()).replace('],', '];').replace(',', '')
y_str = str(y.tolist())
def_datasets = """const X = convert(Array{Float32, 2}, """f"{X_str})""""
const y = convert(Array{Float32, 1}, """f"{y_str})"
if weights is not None:
weight_str = str(weights.tolist())
def_datasets += """
const weights = convert(Array{Float32, 1}, """f"{weight_str})"
if use_custom_variable_names:
def_hyperparams += f"""
const varMap = {'["' + '", "'.join(variable_names) + '"]'}"""
with open(f'/tmp/.hyperparams_{rand_string}.jl', 'w') as f:
print(def_hyperparams, file=f)
with open(f'/tmp/.dataset_{rand_string}.jl', 'w') as f:
print(def_datasets, file=f)
with open(f'/tmp/.runfile_{rand_string}.jl', 'w') as f:
print(f'@everywhere include("/tmp/.hyperparams_{rand_string}.jl")', file=f)
print(f'@everywhere include("/tmp/.dataset_{rand_string}.jl")', file=f)
print(f'@everywhere include("{pkg_directory}/sr.jl")', file=f)
print(f'fullRun({niterations:d}, npop={npop:d}, ncyclesperiteration={ncyclesperiteration:d}, fractionReplaced={fractionReplaced:f}f0, verbosity=round(Int32, {verbosity:f}), topn={topn:d})', file=f)
print(f'rmprocs(nprocs)', file=f)
command = [
f'julia', f'-O{julia_optimization:d}',
f'-p', f'{procs}',
f'/tmp/.runfile_{rand_string}.jl',
]
if timeout is not None:
command = [f'timeout', f'{timeout}'] + command
global global_n_features
global global_equation_file
global global_variable_names
global global_extra_sympy_mappings
global_n_features = X.shape[1]
global_equation_file = equation_file
global_variable_names = variable_names
global_extra_sympy_mappings = extra_sympy_mappings
print("Running on", ' '.join(command))
process = subprocess.Popen(command)
try:
process.wait()
except KeyboardInterrupt:
process.kill()
return get_hof()
def run_feature_selection(X, y, select_k_features):
"""Use a gradient boosting tree regressor as a proxy for finding
the k most important features in X, returning indices for those
features as output."""
from sklearn.ensemble import RandomForestRegressor, GradientBoostingRegressor
from sklearn.feature_selection import SelectFromModel, SelectKBest
clf = GradientBoostingRegressor(n_estimators=100, learning_rate=0.1, max_depth=1, random_state=0, loss='ls') #RandomForestRegressor()
clf.fit(X, y)
selector = SelectFromModel(clf, threshold=-np.inf,
max_features=select_k_features, prefit=True)
return selector.get_support(indices=True)
def get_hof(equation_file=None, n_features=None, variable_names=None, extra_sympy_mappings=None):
"""Get the equations from a hall of fame file. If no arguments
entered, the ones used previously from a call to PySR will be used."""
global global_n_features
global global_equation_file
global global_variable_names
global global_extra_sympy_mappings
if equation_file is None: equation_file = global_equation_file
if n_features is None: n_features = global_n_features
if variable_names is None: variable_names = global_variable_names
if extra_sympy_mappings is None: extra_sympy_mappings = global_extra_sympy_mappings
global_equation_file = equation_file
global_n_features = n_features
global_variable_names = variable_names
global_extra_sympy_mappings = extra_sympy_mappings
try:
output = pd.read_csv(equation_file + '.bkup', sep="|")
except FileNotFoundError:
print("Couldn't find equation file!")
return pd.DataFrame()
scores = []
lastMSE = None
lastComplexity = 0
sympy_format = []
lambda_format = []
use_custom_variable_names = (len(variable_names) != 0)
local_sympy_mappings = {
**extra_sympy_mappings,
**sympy_mappings
}
if use_custom_variable_names:
sympy_symbols = [sympy.Symbol(variable_names[i]) for i in range(n_features)]
else:
sympy_symbols = [sympy.Symbol('x%d'%i) for i in range(n_features)]
for i in range(len(output)):
eqn = sympify(output.loc[i, 'Equation'], locals=local_sympy_mappings)
sympy_format.append(eqn)
lambda_format.append(lambdify(sympy_symbols, eqn))
curMSE = output.loc[i, 'MSE']
curComplexity = output.loc[i, 'Complexity']
if lastMSE is None:
cur_score = 0.0
else:
cur_score = - np.log(curMSE/lastMSE)/(curComplexity - lastComplexity)
scores.append(cur_score)
lastMSE = curMSE
lastComplexity = curComplexity
output['score'] = np.array(scores)
output['sympy_format'] = sympy_format
output['lambda_format'] = lambda_format
return output[['Complexity', 'MSE', 'score', 'Equation', 'sympy_format', 'lambda_format']]
def best_row(equations=None):
"""Return the best row of a hall of fame file using the score column.
By default this uses the last equation file.
"""
if equations is None: equations = get_hof()
best_idx = np.argmax(equations['score'])
return equations.iloc[best_idx]
def best_tex(equations=None):
"""Return the equation with the best score, in latex format
By default this uses the last equation file.
"""
if equations is None: equations = get_hof()
best_sympy = best_row(equations)['sympy_format']
return sympy.latex(best_sympy.simplify())
def best(equations=None):
"""Return the equation with the best score, in latex format
By default this uses the last equation file.
"""
if equations is None: equations = get_hof()
best_sympy = best_row(equations)['sympy_format']
return best_sympy.simplify()
def best_tex(equations=None):
"""Return the equation with the best score, in latex format
By default this uses the last equation file.
"""
if equations is None: equations = get_hof()
best_sympy = best_row(equations)['sympy_format']
return sympy.latex(best_sympy.simplify())
def best_callable(equations=None):
"""Return the equation with the best score, in callable format
By default this uses the last equation file.
"""
if equations is None: equations = get_hof()
return best_row(equations)['lambda_format']