ACM Transactions on Probabilistic Machine Learning is open for submissions at https://mc.manuscriptcentral.com/topml.

ACM Transactions on Probabilistic Machine Learning focuses on probabilistic methods that learn from data to improve performance on decision-making or prediction tasks under uncertainty. Optimization, decision-theoretic or information-theoretic methods are within the remit if they are underpinned by a probabilistic structure. Probabilistic methods may be harnessed to address questions related to statistical inference, uncertainty quantification, predictive calibration, data generation and sampling, causal inference, stability, and scalability. Examples of approaches relevant to the scope include Bayesian modelling and inference, variational inference, Gaussian processes, Monte Carlo sampling, Stein-based methods, and ensemble modelling. Examples of models for which probabilistic approaches are sought include neural networks, kernel-based models, graph-based models, reinforcement learning models, recommender systems, and statistical and stochastic models. Ethical considerations of probabilistic machine learning, such as data privacy and algorithmic fairness, should be addressed in papers where there is a direct ethical connection or context for the work being described.

The journal welcomes theoretical, methodological, and applied contributions. Purely theoretical contributions are of interest if they introduce novel methodology. Methodological and applied contributions are of interest provided that proposed probabilistic approaches are motivated and empirically corroborated by non-trivial examples or applications. Multidisciplinary approaches with a probabilistic structure are within the scope.

TOPML focuses on probabilistic methods that learn from data to improve performance on decision-making or prediction tasks under uncertainty. Indicative research areas, approaches and models are provided below to outline relevant areas of submission.

Probabilistic methods may be harnessed to address questions related to

    statistical inference,
    uncertainty quantification,
    predictive calibration,
    data generation and sampling,
    interpretability and explainability,
    causal inference,
    stability and robustness,
    scalability.

Examples of methods relevant to the scope include

    Bayesian modelling and inference,
    variational inference,
    Gaussian processes,
    Monte Carlo sampling,
    Stein-based methods,
    ensemble modelling,
    Bayesian optimization. 

Examples of machine learning and statistical models for which probabilistic approaches are sought include

    neural networks,
    kernel-based models,
    graph-based models,
    reinforcement learning models,
    recommender systems,
    stochastic models.

Ethical considerations of probabilistic machine learning should be addressed in papers where there is a direct ethical connection or context for the work being described. For instance, ethical considerations of topical interest to TOPML span

    data privacy,
    algorithmic fairness,
    safety-critical applications,
    potential negative societal impacts of the work.

Validation and support for reproducibility are important components of machine learning. Papers should address this with an appropriate combination of

    assumptions and proofs of theoretical claims, 
    arguments supporting practical application,  
    discussion of limitations,  
    discussion of evaluation methods and metrics,
    experimental evaluation. 

Experimental work should support reproducibility as discussed below.