Probabilistic Circuits (PCs) have emerged as an efficient framework for representing and learning complex probability distributions. Nevertheless, the existing body of research on PCs predominantly concentrates on data-driven parameter learning, often neglecting the potential of knowledge-intensive learning, a particular issue in data-scarce/knowledge-rich domains such as healthcare. To bridge this gap, we propose a novel unified framework that can systematically integrate diverse domain knowledge into the parameter learning process of PCs. Experiments on several benchmarks as well as real world datasets show that our proposed framework can both effectively and efficiently leverage domain knowledge to achieve superior performance compared to purely data-driven learning approaches.

Human-in-the-loop (HIL) systems have emerged as a promising approach for combining the strengths of data-driven machine learning models with the contextual understanding of human experts. However, a deeper look into several of these systems reveals that calling them HIL would be a misnomer, as they are quite the opposite, namely AI-in-the-loop (AI2L) systems: the human is in control of the system, while the AI is there to support the human. We argue that existing evaluation methods often overemphasize the machine (learning) component's performance, neglecting the human expert's critical role. Consequently, we propose an AI2L perspective, which recognizes that the human expert is an active participant in the system, significantly influencing its overall performance. By adopting an AI2L approach, we can develop more comprehensive systems that faithfully model the intricate interplay between the human and machine components, leading to more effective and robust AI systems.

We consider the problem of late multimodal fusion for discriminative learning. Motivated by noisy, multi-source domains that require understanding the reliability of each data source, we explore the notion of credibility in the context of multimodal fusion. We propose a combination function that uses probabilistic circuits (PCs) to combine predictive distributions over individual modalities. We also define a probabilistic measure to evaluate the credibility of each modality via inference queries over the PC. Our experimental evaluation demonstrates that our fusion method can reliably infer credibility while being competitive with the state-of-the-art.

Multimodal fusion is important for building intelligent systems that exploit patterns across diverse data sources for improved decision-making. However, the reliability and robustness of these systems in safety-critical domains are often compromised by the inherent noise and incompleteness of data. Probabilistic Circuits (PCs) have recently emerged as a promising approach for late (or decision) fusion. Their strength lies in being both expressive and capable of inferring source credibility due to their ability to tractably perform exact probabilistic inference. However, their ability to handle missing data and their reliability in practical scenarios remains underexplored. This work investigates the robustness of PCs as fusion functions in scenarios with missing and noisy data; particularly by examining their impact on the calibration and reliability of the resulting classifiers. Our findings show that PCs not only enable the modeling of complex correlations across modalities but also lead to calibrated and reliable classifiers, highlighting their potential as a robust fusion mechanism in multimodal systems.

Bayesian networks are a popular class of directed probabilistic graphical models that allow for closed-form learning of the local parameters if complete data are available. However, learning the parameters is challenging when the data are sparse, incomplete, and uncertain. In this work, we present an approach to this problem based on credal networks, a generalization of Bayesian networks based on set-valued local parameters. We derive an algorithm to learn such set-valued parameters from data using qualitative knowledge in the form of monotonic influence statements. Our empirical evaluation shows that using qualitative knowledge reduces uncertainty about the parameters without significant loss in accuracy.

Cutset networks (CNs) are interpretable probabilistic representations that combine probability trees and tree Bayesian networks, to model and reason about large multi-dimensional probability distributions. Motivated by high-stakes applications in domains such as healthcare where (a) rich domain knowledge in the form of qualitative influences is readily available and (b) use of interpretable models that the user can efficiently probe and infer over is often necessary, we focus on learning CNs in the presence of qualitative influences. We propose a penalized objective function that uses the influences as constraints, and develop a gradient-based learning algorithm, KICN. We show that because CNs are tractable, KICN is guaranteed to converge to a local maximum of the penalized objective function. Our experiments on several benchmark data sets show that our new algorithm is superior to the state-of-the-art, especially when the data is scarce or noisy.

Summary Deep learning has revolutionized the field of conversation modeling. A lot of the research has been toward making the conversational agent more human‐like. As a result, overall the model size increases. Bigger models require more data and are costly to build and maintain. Often, for some tasks, high‐quality responses are not necessary. In this paper, a model that consumes fewer resources and a way to augment conversation data without increasing the size of the vocabulary is proposed. The proposed model uses a modified version of the GRU instead of the LSTM to encode and decode sequences of text.