AdaptivePy 1.0.0 — First stable release

AdaptivePy 1.0.0

The first stable release of AdaptivePy — adaptive sampling for molecular dynamics trajectories via clustering-based state-space partitioning and policy-driven seed selection. This marks a commitment to a stable public API under semantic versioning.

Highlights

• Seven built-in policies
  • least_counts — target under-sampled clusters
  • random — baseline random cluster sampling
  • fast — feature-directed exploration/exploitation (Zimmerman & Bowman 2015)
  • ma_reap — multi-agent REAP with learned CV weights and agent stakes (Kleiman & Shukla 2022)
  • knn_as — k-nearest-neighbors adaptive sampling over cluster representatives (Rovers et al. 2025)
  • maxent_vampnet — entropy-based frame selection via VAMPNet, no clustering required (Kleiman & Shukla 2023)
  • ts_dar — OOD-score frame selection via TS-DAR hyperspherical embeddings, no clustering required (Liu et al. 2025)
• Metapolicy ensembles — combine multiple policies with majority_polling or per-policy allocation.
• Flexible clustering — KMeans, MiniBatch KMeans, and regular-space, with clustering automatically skipped for frame-level-only runs.
• Reproducible outputs — seeds, cluster assignments, fitted model, per-policy score sidecars, logs, a config snapshot, and optional PDB export.
• CLI + Python API — adaptivepy run/validate/list-policies plus run_adaptive_sampling(...).
• Full documentation — https://shuklagroup.github.io/AdaptivePy/

Installation

pip install adaptivepy-sampling

v0.3.0

What's Changed

Added

• Added the knn_as adaptive sampling policy with configuration, output writing, documentation, and tests.
• Added the MaxEnt VAMPNet frame-level policy, including the optional maxent dependency extra.
• Added metapolicy support with majority polling and allocation across multiple policies.
• Expanded policy output metadata for new policy and metapolicy workflows.

Documentation

• Updated README and user guides for KNN-AS, MaxEnt VAMPNet, MA-REAP, FAST, and metapolicy configuration.
• Added or expanded configuration, output, API reference, and getting-started documentation.

Testing

• Added broad test coverage for the new policies, metapolicy behavior, configuration parsing, and output handling.

v0.2.0

Highlights
This release adds two goal-oriented sampling policies from the adaptive MD literature, extends configuration with per-policy parameters, and formalizes project licensing.

New policies
FAST (fast) — Fluctuation Amplification of Specific Traits (Zimmerman & Bowman, 2015)

Balances feature-directed exploitation with exploration of poorly sampled clusters
Configurable feature columns, maximize/minimize directions, weights, and alpha
Writes fast/scores.csv with directed, exploration, and total reward components
MA-REAP (ma_reap) — Multiagent REAP (Kleiman & Shukla, 2022)

Multi-agent extension of REAP with per-agent stakes and learned CV weights
Assign trajectories to agents via policy_params.ma_reap.agents
Supports stakes methods (percentage, equal, max, logistic) and aggregation regimes (collaborative, noncollaborative, competitive)
Writes sidecar outputs: scores.csv, agent_weights.csv, stakes.csv, executors.csv

v0.1.2

Full Changelog: v0.1.0...v0.1.2

Highlights

• Documentation site — https://hnadeem2.github.io/AdaptivePy/
• README refresh — badges, overview table, and guide links
• .pkl feature support — per-trajectory .npy or .pkl inputs

Install

pip install adaptivepy-sampling==0.1.2

v0.1.1

Full Changelog: v0.1.0...v0.1.1

v0.1.0

Full Changelog: https://github.com/hnadeem2/AdaptivePy/commits/v0.1.0
First release