Publications

Bayesian Meta-Learning with Expert Feedback for Task-Shift Adaptation through Causal Embeddings

This paper provides an approach to incorporate expert knowledge to adapt to task-shift in meta learners when predicting in out-of-distribution, by using causal graphs.

Recommended citation: Mäkinen, L., Loría, J., Kaski, S. (2026+). " Bayesian Meta-Learning with Expert Feedback for Task-Shift Adaptation through Causal Embeddings." in The International Conference of Machine Learning. https://proceedings.mlr.press/v306/makinen26a.html

Incorporating Expert Knowledge into Bayesian Causal Discovery of Mixtures of Directed Acyclic Graphs

We provide a meaningful way to incorporate domain knowledge into causal discovery, in the specially hard case of mixtures of graphs. Incorporating the expert knowledge allows inferring correct mixture components, without having observed them.

Recommended citation: Björkman, Z., Loría, J., Wharrie, S., Kaski, S. (2026). " Incorporating Expert Knowledge into Bayesian Causal Discovery of Mixtures of Directed Acyclic Graphs." In Proceedings of the 28th International Conference on Artificial Intelligence and Statistics, volume 300 of Proceedings of Machine Learning Research. https://proceedings.mlr.press/v300/bjorkman26a.html

Deep Kernel Posterior Learning under Infinite Variance Prior Weights

We propose a method to efficiently predict on neural networks that are infinitely wide, which take the form of deep kernel processes, using priors that have unbounded variance.

Recommended citation: Loría, J, Bhadra, A. (2025). "Deep Kernel Posterior Learning under Infinite Variance Prior Weights ." (The 13th International Conference on Learning Representations (ICLR 2025)). https://openreview.net/forum?id=usFdPd4Ghs

Inferring Effect Ordering Without Causal Effect Estimation

Decision makers often care about the causal effect to select the top customers in a specific treatment, we propose a framework on how to do this even when the data does not lend itself to accurately estimate causal effects.

Recommended citation: Fernández-Loría, C, Loría J. (2024+). " Inferring Effect Ordering Without Causal Effect Estimation ." submitted. https://arxiv.org/pdf/2206.12532.pdf

Posterior Inference on Shallow Infinitely Wide Bayesian Neural Networks under Weights with Unbounded Variance

We propose a method to do posterior inference on neural networks that are infinitely wide, using priors that have unbounded variance.

Recommended citation: Loría, J, Bhadra, A. (2024). "Posterior Inference on Shallow Infinitely Wide Bayesian Neural Networks under Weights with Unbounded Variance ." 40th Conference on Uncertainty in Artificial Intelligence (UAI 2024). https://openreview.net/forum?id=J97bdMR7Lv

SURE-tuned Bridge Regression

Bridge is a regularization technique that can consume a lot of time when the number of covariates is much larger than the number of observations. We propose a non-iterative method to reduce this timing without losing statistical prediction power.

Recommended citation: Loria, J, Bhadra, A. (2024). "SURE-tuned Bridge Regression." Statistics and Computing 34, 30. https://link.springer.com/article/10.1007/s11222-023-10350-z

Demographic Modeling via 3-dimensional Markov Chains

We propose a new model for understanding and prediction of demographic modeling, using Markov Chains. We apply it to an institution in Costa Rica.

Recommended citation: Víquez, J. J., Víquez, J. A., Campos, A., Loría, J., & Mendoza, L. A. (2018). Modelación de poblaciones vía cadenas de Markov tridimensionales. Revista De Matemática: Teoría Y Aplicaciones, 25(2), 185–214. https://doi.org/10.15517/rmta.v25i2.33608. https://revistas.ucr.ac.cr/index.php/matematica/article/view/33608/34172

Jorge Loría