SELA : SELA enhances automated machine learning by combining Monte Carlo Tree Search and LLM-based agents.

SELA

Model Training and Deployment Research Tools #Automated Machine Learning #Monte Carlo Tree Search #Large Language Models #Model Selection #Model Integration Standard Picks Open Source

Overview :

SELA is an innovative system that enhances automated machine learning (AutoML) by integrating Monte Carlo Tree Search (MCTS) with LLM-based agents. Traditional AutoML methods often produce low-diversity and suboptimal code, limiting their effectiveness in model selection and integration. SELA represents pipeline configurations as trees, enabling agents to intelligently explore the solution space and iteratively refine strategies based on experimental feedback.

Target Users :

SELA's target audience includes machine learning researchers and developers, particularly professionals seeking to enhance model selection and integration efficiency through automated machine learning processes. SELA offers a novel approach to exploring and optimizing machine learning pipelines, suitable for those who need to handle extensive data and model selection tasks.

Total Visits： 474.6M

Top Region： US(19.34%)

Website Views ： 48.0K

Use Cases

Using SELA for model selection and integration on the Titanic dataset.

Utilizing SELA to optimize model performance in a housing price prediction task.

Conducting ablation studies with SELA to compare the impact of different search strategies on model performance.

Features

Data Preparation: Supports downloading datasets from links or preparing datasets from scratch.

Flexible Configuration: Users can modify data and LLM configurations as needed.

Running SELA: Provides detailed steps for running SELA, including setup, experiment execution, and parameter configuration.

Experiment Modes: Supports both MCTS and Random Search (RS) experimental modes.

Parameter Tuning: Users can adjust parameters such as rollouts and timeout as required.

Resume from Checkpoint: Supports loading existing MCTS trees to continue experiments after interruptions.

Ablation Study: Supports conducting ablation studies to compare the effects of different search strategies.

How to Use

1. Data Preparation: Download or prepare datasets according to SELA's guidelines.

2. Configuration Settings: Modify the configurations in the `datasets.yaml` and `data.yaml` files as needed.

3. Install Dependencies: Run `pip install -r requirements.txt` in the SELA directory to install the required dependencies.

4. Run SELA: Use the command `python run_experiment.py` along with the appropriate parameters to run an experiment.

5. Parameter Tuning: Adjust parameters such as rollouts and timeout as per experimental requirements.

6. Results Analysis: Analyze the experimental results and iteratively improve strategies based on feedback.

7. Resume from Checkpoint: If the experiment is interrupted, use the `--load_tree` parameter to load the previous MCTS tree and continue the experiment.