Incorporating probabilistic domain knowledge into deep multiple instance learning

Ghadi S. Al Hajj, Aliaksandr Hubin, Chakravarthi Kanduri, Milena Pavlovic, Knut Rand, Michael Widrich, Anne Solberg, Victor Greiff, Johan Pensar, Günter Klambauer, Geir Kjetil Sandve

The 41st International Conference on Machine Learning (ICML 2024)

Abstract

Deep learning methods, including deep multiple instance learning methods, have been criticized for their limited ability to incorporate domain knowledge. A reason that knowledge incorporation is challenging in deep learning is that the models usually lack a mapping between their model components and the entities of the domain, making it a non-trivial task to incorporate probabilistic prior information. In this work, we show that such a mapping between domain entities and model components can be defined for a multiple instance learning setting and propose a framework DeeMILIP that encompasses multiple strategies to exploit this mapping for prior knowledge incorporation. We motivate and formalize these strategies from a probabilistic perspective. Experiments on an immune-based diagnostics case show that our proposed strategies allow to learn generalizable models even in settings with weak signals, limited dataset size, and limited compute.

Proposed Strategies

Results Overview

Incorporating PDK improves performance at low witness rates

Incorporating PDK allows learning with less data

Incorporating PDK allows learning with less compute

Important links

Code	ICML 2024 Poster	OpenReview

Reference


            @inproceedings{
            hajj2024incorporating,
            title={Incorporating probabilistic domain knowledge into deep multiple instance learning},
            author={Ghadi S. Al Hajj and Aliaksandr Hubin and Chakravarthi Kanduri and Milena Pavlovic and Knut Dagestad Rand and Michael Widrich and Anne Schistad Solberg and Victor Greiff and Johan Pensar and G{\"u}nter Klambauer and Geir Kjetil Sandve},
            booktitle={Forty-first International Conference on Machine Learning},
            year={2024},
            url={https://openreview.net/forum?id=GfNyqrwECJ}
            }