Metadata-Version: 2.4
Name: ascent-science-aco
Version: 0.1.3
Summary: Ant colony optimizer plugin for Ascent Science SDK
Author: Ascent Science
License-Expression: LicenseRef-Proprietary
Classifier: Programming Language :: Python :: 3
Classifier: Programming Language :: Python :: 3 :: Only
Classifier: Typing :: Typed
Requires-Python: >=3.9
Description-Content-Type: text/markdown
Requires-Dist: ascent-science<0.10,>=0.9.3

# Ascent Science ACO plugin

Install version 0.1.3 with
`python -m pip install ascent-science-aco==0.1.3`, then use
`ascent.combinatorial.ACOOptimizer(...)`.

ACO runs on the authenticated platform server by default, where the deployment
selects the Haifeike Compass accelerator on host 244. Pass `execution="local"`
only when local execution is explicitly required.

The solver implements the Ant System update rule for directed and symmetric
travelling-salesperson instances. The result exposes three convergence series:

- `history`: best-so-far distance (monotonically non-increasing)
- `iteration_best_history`: best distance found in each iteration
- `mean_history`: mean distance of all ants in each iteration

## 0.1.3

- Preserves the real accelerator runtime returned by the platform.
- Falls back to the response envelope runtime for compatibility with older servers.
- Rejects missing or invalid runtime metadata instead of reporting a misleading zero.

## 0.1.2

- Runs on the platform GPU server by default and preserves explicit local mode.
- Returns all three convergence histories from Haifeike accelerator executions.

## 0.1.1

- Uses the Ant System transition probability with pheromone and inverse-distance
  heuristics.
- Applies evaporation and `Q / L` deposits after every iteration.
- Preserves directed pheromone updates for asymmetric matrices and mirrors
  deposits for symmetric matrices.
- Returns the best-so-far convergence history used by the experiment chart.
