🧠Active Learning

Once the unlabeled data is in Dioptra, running Active Learning Miners can be done in just a few lines of code.

Step 1: define your filters

Define the filters to select the data to mine from. It's like SQL.

We support standard operators =, !=, >, <, >=, <=, in, not in, like, not like

Python

filters = [{
    'left': 'tags.name',
    'op': '=',
    'right': 'model_id'
 }, {
    'left': 'tags.value',
    'op': '=',
    'right': 'anomaly_detection'
 }, {
    'left': 'tags.name',
    'op': '=',
    'right': 'model_version'
 }, {
    'left': 'tags.value',
    'op': '=',
    'right': 'v1'
 }]

Step 2: Create your miner, run it and wait for the results

Python

import time
from dioptra.miners.entropy_miner import EntropyMiner

my_miner = EntropyMiner('my entropy miner', budget=100, filters=filters)
my_miner.run()

datapoint_ids = my_miner.get_results()

Step 3: get the resulting datapoints from the lake

Python

from dioptra.lake.utils import select_datapoints

datapoint_ids = my_miner.get_results()

results_df = select_datapoints(filters=[{
    'left': 'id',
    'op': 'in',
    'right': ids
}])

# do stuff with the results

📖Miners basics

⛏️Miners SDK

🚗[Experimental] Mining on the edge