Attach external data to experiments

In this guide, we will instruct you how to attach external data to an experiment.

Load your data

We will use fake data for this example. You will be responsible for loading your data from your own source.

import pandas as pd

df = pd.DataFrame({
    "dof1": [1, 2, 3, 4, 5],
    "dof2": [1, 2, 3, 4, 5],
    "dof3": [1, 2, 3, 4, 5],
    "objective1": [1, 2, 3, 4, 5],
    "objective2": [1, 2, 3, 4, 5],
})

Transform your data to the correct format

data = df.to_dict(orient="records")

Configure an agent

The DOF and Objective names must match the keys in the data dictionaries.

from blop.ax import Agent, RangeDOF, Objective

dofs = [
    RangeDOF(movable=dof1, bounds=(-5.0, 5.0), parameter_type="float"),
    RangeDOF(movable=dof2, bounds=(-5.0, 5.0), parameter_type="float"),
    RangeDOF(movable=dof3, bounds=(-5.0, 5.0), parameter_type="float"),
]

objectives = [
    Objective(name="objective1", minimize=True),
    Objective(name="objective2", minimize=False),
]

def evaluation_function(uid: str, suggestions: list[dict]) -> list[dict]:
    """Replace this with your own evaluation function."""
    outcomes = []
    for suggestion in suggestions:
        outcome = {
            "_id": suggestion["_id"],
            "objective1": 0.1,
            "objective2": 0.2,
        }
        outcomes.append(outcome)
    return outcomes

agent = Agent(
    readables=[readable1, readable2],
    dofs=dofs,
    objectives=objectives,
    evaluation=evaluation_function,
)

Ingest your data

After this, the next time you get a suggestion from the agent it will re-train the model(s) with the new data.

agent.ingest(data)

(Optional) Configure the generation strategy

If no trials have been run yet, you must configure the generation strategy before summarizing the data.

agent.ax_client.configure_generation_strategy()

Sanity check the data you attached

Verify the data you attached is correct.

agent.ax_client.summarize()