PySwarms

1.3.0 verified Fri May 01 auth: no python maintenance

PySwarms is a Python library for Particle Swarm Optimization (PSO). It provides a simple and intuitive API for implementing PSO algorithms, including global best (gbest), local best (lbest), and binary PSO, as well as tools for visualization and hyperparameter tuning. The current version is 1.3.0, released in 2021. The project is in maintenance mode with quarterly release cadence.

pip install pyswarms

Common errors

error AttributeError: module 'pyswarms' has no attribute 'GlobalBestPSO' ↓

cause Trying to import optimizer from top-level pyswarms.

fix

Use: from pyswarms.single import GlobalBestPSO

error ValueError: operands could not be broadcast together with shapes (10,2) (2,) ↓

cause Objective function returns wrong shape (e.g., returns a single value for all particles).

fix

Ensure objective function returns an array of shape (n_particles,), i.e., one cost per particle.

error KeyError: 'c1' ↓

cause Missing required key in options dict for optimizer.

fix

Provide options as dict: {'c1': 0.5, 'c2': 0.3, 'w': 0.9}

error TypeError: 'NoneType' object is not callable ↓

cause Using the old animate module incorrectly or passing None as a callback.

fix

Use the new plotters API: from pyswarms.utils.plotters import plot_cost_history

Warnings

deprecated The 'animate' module and some plotting functions have been deprecated in favor of plotters. Check documentation for the correct API. ↓

fix Use pyswarms.utils.plotters instead of pyswarms.utils.animate.

gotcha The optimizer modifies the objective function's arguments; ensure the function signature is correct (returns array of shape (n_particles,)). Mistakes lead to shape errors. ↓

fix Your objective function must accept a 2D array (rows=particles, cols=dimensions) and return a 1D array of costs.

gotcha Always pass options as a dict with keys 'c1', 'c2', 'w' for PSO variants. Missing keys cause KeyError or unexpected behavior. ↓

fix options = {'c1': 0.5, 'c2': 0.3, 'w': 0.9}

breaking Version 1.0.0 changed the backend API and removed many old functions. Code written for v0.x may not work. ↓

fix Review migration guide: https://pyswarms.readthedocs.io/en/latest/migration.html

Imports

GlobalBestPSO
wrong
```
import pyswarms
```
correct
```
from pyswarms.single import GlobalBestPSO
```
Importing the package alone does not give access to optimizer classes; you must import from submodules.
LocalBestPSO
wrong
```
from pyswarms import LocalBestPSO
```
correct
```
from pyswarms.single import LocalBestPSO
```
LocalBestPSO is not in the top-level pyswarms namespace.

BinaryPSO

wrong

from pyswarms.single import BinaryPSO

correct

from pyswarms.discrete import BinaryPSO

BinaryPSO is in the discrete submodule, not single.

plot_cost_history

wrong

from pyswarms import plot_cost_history

correct

from pyswarms.utils.plotters import plot_cost_history

Plotting functions are under utils.plotters.

Quickstart

Minimal example optimizing the sphere function with global best PSO.

import numpy as np
from pyswarms.single import GlobalBestPSO

def sphere(x):
    return np.sum(x**2, axis=1)

optimizer = GlobalBestPSO(n_particles=10, dimensions=2, options={'c1': 0.5, 'c2': 0.3, 'w':0.9})
cost, pos = optimizer.optimize(sphere, iters=100)
print("Best cost:", cost)
print("Best position:", pos)