“The algorithm repeatedly sent officers to neighborhoods with a high proportion of racial minotrities regardless of the true crime rate of the areas.”
“The image recognition algorithms in Google Photos were classifying black people as gorillas.”
“The algorithm reportedly downgraded resumes containing the words "women's" and filtered out candidates who had attended women-only colleges.”
Automate testing of ML systems
Share and track your results
Run locally and in the cloud
Get reliable and actionable results
Plug & Play
First, use Algoneer's data schema description capabilities to annotate your data.
from algoneer import DataSchema, AttributeSchema as AS class MySchema(DataSchema): temperature = AS(type=AS.Numerical, roles=["x"]) humidity = AS(type=AS.Numerical, roles=["x"]) count = AS(type=AS.Numerical, roles=["y"])
Next up you load your dataset. Algoneer supports pandas dataframes (more formats coming soon).
from algoneer.dataset.pandas import PandasDataset import pandas as pd #Import your dataset df = pd.read_csv("my-data.csv") dataset = PandasDataset(df)
To manage your ML models and algorithms with Algoneer. Currently, Algoneer supports sklearn algorithms, with more libraries (tensorflow, pytorch) coming soon.
from algoneer.algorithm.sklearn import SklearnAlgorithm from sklearn.ensemble import RandomForest algo = SklearnAlgorithm(RandomForest, n_estimators=100) model = algo.train(dataset)
Using the data and algorithm schemas that you defined, Algoneer can automatically run tests on your models and datasets. Results can be viewed and analyzed offline or sent to the Algonaut API.
# Test machine learning models model_results = model.test() # Test datasets dataset_results = dataset.test() for name, result in model_results.items(): print(result.format('text'))
Explainability: LIME, SHAP, PDP, ALE
Robustness, Bias, Security: coming soon
Supported model types
Supported data types