<sc>Dagma-DCE</sc>: Interpretable, Non-Parametric Differentiable Causal Discovery
We introduce <sc>Dagma-DCE</sc>, an interpretable and model-agnostic scheme for differentiable causal discovery. Current non- or over-parametric methods in differentiable causal discovery use opaque proxies of “independence” to justify the inclusion or exclusion of...
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2024-01-01
|
| Series: | IEEE Open Journal of Signal Processing |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/10384714/ |
| _version_ | 1797335539146293248 |
|---|---|
| author | Daniel Waxman Kurt Butler Petar M. Djuric |
| author_facet | Daniel Waxman Kurt Butler Petar M. Djuric |
| author_sort | Daniel Waxman |
| collection | DOAJ |
| description | We introduce <sc>Dagma-DCE</sc>, an interpretable and model-agnostic scheme for differentiable causal discovery. Current non- or over-parametric methods in differentiable causal discovery use opaque proxies of “independence” to justify the inclusion or exclusion of a causal relationship. We show theoretically and empirically that these proxies may be arbitrarily different than the actual causal strength. Juxtaposed with existing differentiable causal discovery algorithms, <sc>Dagma-DCE</sc> uses an interpretable measure of causal strength to define weighted adjacency matrices. In a number of simulated datasets, we show our method achieves state-of-the-art level performance. We additionally show that <sc>Dagma-DCE</sc> allows for principled thresholding and sparsity penalties by domain-experts. The code for our method is available open-source at <uri>https://github.com/DanWaxman/DAGMA-DCE</uri>, and can easily be adapted to arbitrary differentiable models. |
| first_indexed | 2024-03-08T08:38:51Z |
| format | Article |
| id | doaj.art-f4ae1f02b2dd4acb82f3205acd051a9f |
| institution | Directory Open Access Journal |
| issn | 2644-1322 |
| language | English |
| last_indexed | 2024-03-08T08:38:51Z |
| publishDate | 2024-01-01 |
| publisher | IEEE |
| record_format | Article |
| series | IEEE Open Journal of Signal Processing |
| spelling | doaj.art-f4ae1f02b2dd4acb82f3205acd051a9f2024-02-02T00:04:39ZengIEEEIEEE Open Journal of Signal Processing2644-13222024-01-01539340110.1109/OJSP.2024.335159310384714<sc>Dagma-DCE</sc>: Interpretable, Non-Parametric Differentiable Causal DiscoveryDaniel Waxman0https://orcid.org/0009-0004-0168-5547Kurt Butler1https://orcid.org/0000-0002-1520-4909Petar M. Djuric2https://orcid.org/0000-0001-7791-3199Department of Electrical and Computer Engineering, Stony Brook University, Stony Brook, NY, USADepartment of Electrical and Computer Engineering, Stony Brook University, Stony Brook, NY, USADepartment of Electrical and Computer Engineering, Stony Brook University, Stony Brook, NY, USAWe introduce <sc>Dagma-DCE</sc>, an interpretable and model-agnostic scheme for differentiable causal discovery. Current non- or over-parametric methods in differentiable causal discovery use opaque proxies of “independence” to justify the inclusion or exclusion of a causal relationship. We show theoretically and empirically that these proxies may be arbitrarily different than the actual causal strength. Juxtaposed with existing differentiable causal discovery algorithms, <sc>Dagma-DCE</sc> uses an interpretable measure of causal strength to define weighted adjacency matrices. In a number of simulated datasets, we show our method achieves state-of-the-art level performance. We additionally show that <sc>Dagma-DCE</sc> allows for principled thresholding and sparsity penalties by domain-experts. The code for our method is available open-source at <uri>https://github.com/DanWaxman/DAGMA-DCE</uri>, and can easily be adapted to arbitrary differentiable models.https://ieeexplore.ieee.org/document/10384714/Causal discoverydifferential causal effectinterpretable machine learning |
| spellingShingle | Daniel Waxman Kurt Butler Petar M. Djuric <sc>Dagma-DCE</sc>: Interpretable, Non-Parametric Differentiable Causal Discovery IEEE Open Journal of Signal Processing Causal discovery differential causal effect interpretable machine learning |
| title | <sc>Dagma-DCE</sc>: Interpretable, Non-Parametric Differentiable Causal Discovery |
| title_full | <sc>Dagma-DCE</sc>: Interpretable, Non-Parametric Differentiable Causal Discovery |
| title_fullStr | <sc>Dagma-DCE</sc>: Interpretable, Non-Parametric Differentiable Causal Discovery |
| title_full_unstemmed | <sc>Dagma-DCE</sc>: Interpretable, Non-Parametric Differentiable Causal Discovery |
| title_short | <sc>Dagma-DCE</sc>: Interpretable, Non-Parametric Differentiable Causal Discovery |
| title_sort | sc dagma dce sc interpretable non parametric differentiable causal discovery |
| topic | Causal discovery differential causal effect interpretable machine learning |
| url | https://ieeexplore.ieee.org/document/10384714/ |
| work_keys_str_mv | AT danielwaxman scdagmadcescinterpretablenonparametricdifferentiablecausaldiscovery AT kurtbutler scdagmadcescinterpretablenonparametricdifferentiablecausaldiscovery AT petarmdjuric scdagmadcescinterpretablenonparametricdifferentiablecausaldiscovery |