DeepFeature: Iterative Context-aware Feature Generation for Wearable Biosignals

Quick start

1. Create a new Python environment

conda create -n featminer python=3.8
conda activate featminer

2. Install the necessary Python packages

pip install -r requirements.txt

3. Configure your API keys

To get started, you'll need to obtain API keys from the following providers:

• LLM API: Choose either Volcengine or Azure
• Text Embedding Model: Cohere

Once you have your API keys, configure them in config file:

# src/config_*.yaml
llm_api_settings:
  platform: # choose your LLM API platform here: [azure, volcengine]

  azure:
    api_key: #
    api_version: #
    azure_endpoint: #
    llm_name: #
    temperature: 0.5
  
  volcengine:
    temperature: 0.5
    api_key: #
    model: #

embedding_settings:
  cohere_api_key: #

4. Prepare the local knowledge base

Please download the local knowledge base files of the tasks from Google Drive and extract them in the embedding_db directory:

• ppg_htn.zip

After downloading, extract all zip files to the embedding_db folder:

# Create the embedding_db directory if it doesn't exist
mkdir -p embedding_db

# Extract the downloaded files
unzip ppg_htn.zip -d embedding_db/

The directory structure should look like:

embedding_db/
└── ppg_bp_htn/

5. Run the example

# run each cell in this file
example_ppg_htn.ipynb

intermediate_results stores the detailed results of each iteration.