Python code implementing the algorithms presented in our AISTATS-24 paper titled "Clustering Items From Adaptively Collected Inconsistent Feedback." These algorithms cluster a set of items using pairwise similarity feedback from an oracle.

Installing the required packages

Run poetry install. This installs numpy, scipy, scikit-learn, and tqdm. Alternatively, you can install these packages manually.

Running the code

1. Create a configuration file like example_config.json
2. Run poetry run python run.py path/to/config.json

The output, consisting of key-value pairs, will be stored as a *.pkl file that can be read using the pickle package. The keys are named so that their purpose is self-explanatory. To load the output file, use something like:

import pickle

output = pickle.load(open("output.pkl", "rb"))

Code structure

|-- clban
|     |-- algorithms.py    # Implements the algorithms
|     |-- oracle.py        # Implements the oracle
|     |-- experiment.py    # Implements the experimental setup
|     |-- utils.py         # Utility functions for evaluation
|
|-- run.py                 # For running a single configuration
|-- example_config.json    # An example configuration file