Q3C

KO | EN

gitlite — search

#javascript #python #hacktoberfest #react #ai #typescript #llm #go #golang #android #machine-learning #rust #deep-learning #linux

★ 14 Open GitHub ↗

Actor-Free Continuous Control via Structurally Maximizable Q-Functions

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# Actor-Free Continuous Control via Structurally Maximizable Q-Functions

[![Python](https://img.shields.io/badge/python-3.10%2B-blue.svg)](https://www.python.org/)
[![License: MIT](https://img.shields.io/badge/license-MIT-green.svg)](https://raw.githubusercontent.com/USC-Lira/Q3C/main/LICENSE)
[![Paper](https://img.shields.io/badge/arXiv-2510.18828-b31b1b.svg)](https://arxiv.org/abs/2510.18828)

This repository provides the official implementation of the paper:

> **Actor-Free Continuous Control via Structurally Maximizable Q-Functions**  
> Yigit Korkmaz*, Urvi Bhuwania*, Ayush Jain†, Erdem Bıyık†  
> NeurIPS 2025

We introduce **Q3C**, a framework for continuous control without an explicit actor network by leveraging structurally maximizable Q-functions for stable and efficient policy learning.

## Installation

### Create a new Conda environment
```bash
conda create --name q3c_env python=3.10
```

### Activate the environment
```bash
conda activate q3c_env
```

### Install dependencies
```bash
pip install -r requirements.txt
pip install -e .
```

**Note:**  MuJoCo must be installed separately, as it is required for all benchmark environments.

This repository uses [Hydra](https://hydra.cc/) for hierarchical configuration management.  
The file `configs/hydra_sb3_q3c_hyperparams.yaml` defines environment-specific hyperparameters.

## Running Experiments

To train and log results to Weights & Biases, run:

```bash
python scripts/run_experiment.py wandb.entity=YOUR_WANDB_USERNAME train.environment=ENV_NAME
```

- Replace `YOUR_WANDB_USERNAME` with your W&B account name.  
- Replace `ENV_NAME` with your target environment (e.g., `Hopper-v4`, `HalfCheetah-v4`, `Walker2d-v4`).

All metrics, losses, and rollout statistics will be automatically logged to your W&B project.

## Citation

If you find this repository useful for your research, please cite:

```bibtex
@inproceedings{korkmaz2025actor,
  title={Actor-Free Continuous Control via Structurally Maximizable Q-Functions},
  author={Korkmaz, Yigit and Bhuwania, Urvi and Jain, Ayush and Bıyık, Erdem},
  booktitle={Advances in Neural Information Processing Systems (NeurIPS)},
  year={2025}
}
```

Q3C

Explore Similar Repositories

Was this content helpful?

Recent Feedback

Download README