Search

Squared tensor networks (TNs) and their generalization as parameterized computational graphs -- squared circuits -- have been recently used as expressive distribution estimators in high dimensions. However, the squaring operation introduces additional complexity when marginalizing variables or computing the partition function, which hinders their usage in machine learning applications. Canonical forms of popular TNs are parameterized via unitary matrices as to simplify the computation of particular marginals, but cannot be mapped to general circuits since these might not correspond to a known TN. Inspired by TN canonical forms, we show how to parameterize squared circuits to ensure they encode already normalized distributions. We then use this parameterization to devise an algorithm to compute any marginal of squared circuits that is more efficient than a previously known one. We conclude by formally showing the proposed parameterization comes with no expressiveness loss for many circuit classes.

This study explored a concept for using Augmented Reality (AR) glasses to customize augmented table setup and lighting in a restaurant. The aim was to provide insights into AR usage in restaurants and contribute to existing research by introducing an extendable and versatile concept for scholars and restaurateurs. A controlled laboratory study, using a within-subjects design, was conducted to investigate the effects of a customizable augmented table setup and lighting on user experience (UX), perceived waiting time, psychological ownership, and social acceptability. A simulated restaurant environment was created using a 360-degree image in Virtual Reality (VR). The study implemented default and customizable table setup and lighting. Results from a paired samples t-test showed a statistically significant effect of table setup and lighting on the pragmatic quality of UX, hedonic quality of UX, overall UX, valence, dominance, psychological ownership, and affect. Furthermore, table setup had a significant effect on arousal and perceived waiting time. Moreover, table setup significantly affected AR interaction, isolation, and safety acceptability, while lighting only affected AR interaction acceptability. Findings suggest that these investigated variables are worth considering for AR applications in a restaurant, especially when offering customizable augmented table setup and lighting., Comment: This paper was accepted as short paper at IEEE ISM 2024, and will be presented on the 13th of December 2024

Complex query answering (CQA) on knowledge graphs (KGs) is gaining momentum as a challenging reasoning task. In this paper, we show that the current benchmarks for CQA are not really complex, and the way they are built distorts our perception of progress in this field. For example, we find that in these benchmarks, most queries (up to 98% for some query types) can be reduced to simpler problems, e.g., link prediction, where only one link needs to be predicted. The performance of state-of-the-art CQA models drops significantly when such models are evaluated on queries that cannot be reduced to easier types. Thus, we propose a set of more challenging benchmarks, composed of queries that require models to reason over multiple hops and better reflect the construction of real-world KGs. In a systematic empirical investigation, the new benchmarks show that current methods leave much to be desired from current CQA methods.

Nowadays, Mixed Reality (MR) headsets are a game-changer for knowledge work. Unlike stationary monitors, MR headsets allow users to work with large virtual displays anywhere they wear the headset, whether in a professional office, a public setting like a cafe, or a quiet space like a library. This study compares four different layouts (eye level-close, eye level-far, below eye level-close, below eye level-far) of virtual displays regarding feelings of safety, perceived productivity, and social acceptability when working with MR in public. We test which layout is most preferred by users and seek to understand which factors affect users' layout preferences. The aim is to derive useful insights for designing better MR layouts. A field study in a public library was conducted using a within-subject design. While the participants interact with a layout, they are asked to work on a planning task. The results from a repeated measure ANOVA show a statistically significant effect on productivity but not on safety and social acceptability. Additionally, we report preferences expressed by the users regarding the layouts and using MR in public., Comment: Part of the 23rd annual IEEE International Symposium on Mixed and Augmented Reality (ISMAR 2024)

The EyeSight feature, introduced with the new Apple Vision Pro XR headset, promises to revolutionize user interaction by simulating real human eye expressions on a digital display. This feature could enhance XR devices' social acceptability and social presence when communicating with others outside the XR experience. In this pilot study, we explore the implications of the EyeSight feature by examining social acceptability, social presence, emotional responses, and technology acceptance. Eight participants engaged in conversational tasks in three conditions to contrast experiencing the Apple Vision Pro with EyeSight, the Meta Quest 3 as a reference XR headset, and a face-to-face setting. Our preliminary findings indicate that while the EyeSight feature improves perceptions of social presence and acceptability compared to the reference headsets, it does not match the social connectivity of direct human interactions., Comment: In 30th ACM Symposium on Virtual Reality Software and Technology (VRST 2024)

This paper establishes a rigorous connection between circuit representations and tensor factorizations, two seemingly distinct yet fundamentally related areas. By connecting these fields, we highlight a series of opportunities that can benefit both communities. Our work generalizes popular tensor factorizations within the circuit language, and unifies various circuit learning algorithms under a single, generalized hierarchical factorization framework. Specifically, we introduce a modular "Lego block" approach to build tensorized circuit architectures. This, in turn, allows us to systematically construct and explore various circuit and tensor factorization models while maintaining tractability. This connection not only clarifies similarities and differences in existing models, but also enables the development of a comprehensive pipeline for building and optimizing new circuit/tensor factorization architectures. We show the effectiveness of our framework through extensive empirical evaluations, and highlight new research opportunities for tensor factorizations in probabilistic modeling.

Large language models (LLMs) are a promising venue for natural language understanding and generation. However, current LLMs are far from reliable: they are prone to generating non-factual information and, more crucially, to contradicting themselves when prompted to reason about relations between entities of the world. These problems are currently addressed with large scale fine-tuning or by delegating reasoning to external tools. In this work, we strive for a middle ground and introduce a loss based on neuro-symbolic reasoning that teaches an LLM to be logically consistent with an external set of facts and rules and improves self-consistency even when the LLM is fine-tuned on a limited set of facts. Our approach also allows to easily combine multiple logical constraints at once in a principled way, delivering LLMs that are more consistent w.r.t. all constraints and improve over several baselines w.r.t. a given constraint. Moreover, our method allows LLMs to extrapolate to unseen but semantically similar factual knowledge, represented in unseen datasets, more systematically.

This study explores the user experience (UX) of virtual reality (VR) cloud gaming under simulated network degradation conditions. Two contrasting games (Beat Saber, Cubism) were streamed via Meta Air Link to a Quest 3 device in a laboratory setup. Packet loss and delay were introduced into the streaming network using NetEm for WiFi traffic manipulation. In a within-subjects experiment, 16 participants played both games under three network conditions (Loss, Delay, Baseline), followed by post-game questionnaires assessing their emotions, perceived quality, game experience, sense of presence, and cybersickness. Friedman's test and Dunn's post-hoc test for pairwise comparisons revealed that packet loss had a greater impact on UX than delay across almost all evaluated aspects. Notably, packet loss in Beat Saber led to a significant increase in cybersickness, whereas in Cubism, players experienced a significant reduction in their sense of presence. Additionally, both games exhibited statistically significant variations between conditions in most game experience dimensions, perceived quality, and emotional responses. This study highlights the critical role of network stability in VR cloud gaming, particularly in minimizing packet loss. The different dynamics between the games suggest the possibility of genre-specific optimization and novel game design considerations for VR cloud games., Comment: Accepted and will be presented in the ISMAR 2024 poster track

Designing expressive generative models that support exact and efficient inference is a core question in probabilistic ML. Probabilistic circuits (PCs) offer a framework where this tractability-vs-expressiveness trade-off can be analyzed theoretically. Recently, squared PCs encoding subtractive mixtures via negative parameters have emerged as tractable models that can be exponentially more expressive than monotonic PCs, i.e., PCs with positive parameters only. In this paper, we provide a more precise theoretical characterization of the expressiveness relationships among these models. First, we prove that squared PCs can be less expressive than monotonic ones. Second, we formalize a novel class of PCs -- sum of squares PCs -- that can be exponentially more expressive than both squared and monotonic PCs. Around sum of squares PCs, we build an expressiveness hierarchy that allows us to precisely unify and separate different tractable model classes such as Born Machines and PSD models, and other recently introduced tractable probabilistic models by using complex parameters. Finally, we empirically show the effectiveness of sum of squares circuits in performing distribution estimation.

Code-LLMs, LLMs pre-trained on large code corpora, have shown great progress in learning rich representations of the structure and syntax of code, successfully using it to generate or classify code fragments. At the same time, understanding if they are able to do so because they capture code semantics, and how well, is still an open question. In this paper, we tackle this problem by introducing SeqCoBench, a benchmark for systematically assessing how Code-LLMs can capture code functional equivalence. SeqCoBench contains over 20 code transformations that either preserve or alter the semantics of Python programs. We conduct extensive evaluations in different settings, including zero-shot and parameter-efficient finetuning methods on state-of-the-art (Code-)LLMs to see if they can discern semantically equivalent or different pairs of programs in SeqCoBench. We find that the performance gap between these LLMs and classical match-based retrieval scores is minimal, with both approaches showing a concerning lack of depth in understanding code semantics., Comment: 37 pages

The advent of powerful neural classifiers has increased interest in problems that require both learning and reasoning. These problems are critical for understanding important properties of models, such as trustworthiness, generalization, interpretability, and compliance to safety and structural constraints. However, recent research observed that tasks requiring both learning and reasoning on background knowledge often suffer from reasoning shortcuts (RSs): predictors can solve the downstream reasoning task without associating the correct concepts to the high-dimensional data. To address this issue, we introduce rsbench, a comprehensive benchmark suite designed to systematically evaluate the impact of RSs on models by providing easy access to highly customizable tasks affected by RSs. Furthermore, rsbench implements common metrics for evaluating concept quality and introduces novel formal verification procedures for assessing the presence of RSs in learning tasks. Using rsbench, we highlight that obtaining high quality concepts in both purely neural and neuro-symbolic models is a far-from-solved problem. rsbench is available at: https://unitn-sml.github.io/rsbench.

Probabilistic integral circuits (PICs) have been recently introduced as probabilistic models enjoying the key ingredient behind expressive generative models: continuous latent variables (LVs). PICs are symbolic computational graphs defining continuous LV models as hierarchies of functions that are summed and multiplied together, or integrated over some LVs. They are tractable if LVs can be analytically integrated out, otherwise they can be approximated by tractable probabilistic circuits (PC) encoding a hierarchical numerical quadrature process, called QPCs. So far, only tree-shaped PICs have been explored, and training them via numerical quadrature requires memory-intensive processing at scale. In this paper, we address these issues, and present: (i) a pipeline for building DAG-shaped PICs out of arbitrary variable decompositions, (ii) a procedure for training PICs using tensorized circuit architectures, and (iii) neural functional sharing techniques to allow scalable training. In extensive experiments, we showcase the effectiveness of functional sharing and the superiority of QPCs over traditional PCs.

This study investigates the feasibility of remote virtual reality (VR) studies conducted at home using VR headsets and video conferencing by deploying an experiment on emotion ratings. 20 participants used head-mounted displays to immerse themselves in 360{\deg} videos selected to evoke emotional responses. The research compares continuous ratings using a graphical interface to retrospective questionnaires on a digitized Likert Scale for measuring arousal and valence, both based on the self-assessment manikin (SAM). It was hypothesized that the two different rating methods would lead to significantly different values for both valence and arousal. The goal was to investigate whether continuous ratings during the experience would better reflect users' emotions compared to the post-questionnaire by mitigating biases such as the peak-end rule. The results show significant differences with moderate to strong effect sizes for valence and no significant differences for arousal with low to moderate effect sizes. This indicates the need for further investigation of the methods used to assess emotion ratings in VR studies. Overall, this study is an example of a remotely conducted VR experiment, offering insights into methods for emotion elicitation in VR by varying the timing and interface of the rating., Comment: The paper is accepted and will be presented at QoMEX 2024

One of the most promising technologies inside the Extended Reality (XR) spectrum is Augmented Reality. This technology is already in people's pockets regarding Mobile Augmented Reality with their smartphones. The scientific community still needs answers about how humans could and should interact in environments where perceived stimuli are different from fully physical or digital circumstances. Moreover, it is still being determined if people accept these new technologies in different social environments and interaction settings or if some obstacles could exist. This paper explores the impact of the Social Environment and the Focus of social interaction on users while playing a location-based augmented reality game, measuring it with user experience and social acceptance indicators. An empirical study in a within-subject fashion was performed in different social environments and under different settings of social interaction focus with N = 28 participants compiling self-reported questionnaires after playing a Scavenger Hunt in Augmented Reality. The measures from two different Social Environments (Crowded vs. Uncrowded) resulted in statistically relevant mean differences with indicators from the Social Acceptability dimension. Moreover, the analyses show statistically relevant differences between the variances from different degrees of Social Interaction Focus with Overall Social Presence, Perceived Psychological Engagement, Perceived Attentional Engagement, and Perceived Emotional Contagion. The results suggest that a location-based AR game played in different social environments and settings can influence the user experience's social dimension. Therefore, they should be carefully considered while designing immersive technological experiences in public spaces involving social interactions between players.

The auditory sense of humans is important when it comes to navigation. The importance is especially high in cases when an object of interest is visually partly or fully covered. Interactions with users of technology are mainly focused on the visual domain of navigation tasks. This paper presents the results of a literature review and user study exploring the impact of spatial auditory navigation on user experience during an augmented outdoor navigation task. For the user test, participants used an augmented reality app guiding them to different locations with different digital augmentation. We conclude that the utilization of the auditory sense is yet still underrepresented in augmented reality applications. In the future, more usage scenarios for audio-augmented reality such as navigation will enhance user experience and interaction quality.

Large language models (LLMs) are a promising venue for natural language understanding and generation tasks. However, current LLMs are far from reliable: they are prone to generate non-factual information and, more crucially, to contradict themselves when prompted to reason about beliefs of the world. These problems are currently addressed with large scale fine-tuning or by delegating consistent reasoning to external tools. In this work, we strive for a middle ground and introduce a training objective based on principled probabilistic reasoning that teaches a LLM to be consistent with external knowledge in the form of a set of facts and rules. Fine-tuning with our loss on a limited set of facts enables our LLMs to be more logically consistent than previous baselines and allows them to extrapolate to unseen but semantically similar factual knowledge more systematically., Comment: Accepted at ICLR 2024 Workshop on Reliable and Responsible Foundation Models

State-of-the-art neurosymbolic learning systems use probabilistic reasoning to guide neural networks towards predictions that conform to logical constraints over symbols. Many such systems assume that the probabilities of the considered symbols are conditionally independent given the input to simplify learning and reasoning. We study and criticise this assumption, highlighting how it can hinder optimisation and prevent uncertainty quantification. We prove that loss functions bias conditionally independent neural networks to become overconfident in their predictions. As a result, they are unable to represent uncertainty over multiple valid options. Furthermore, we prove that these loss functions are difficult to optimise: they are non-convex, and their minima are usually highly disconnected. Our theoretical analysis gives the foundation for replacing the conditional independence assumption and designing more expressive neurosymbolic probabilistic models., Comment: Accepted at ICML 2024

Inside-out tracking is growing popular in consumer VR, enhancing accessibility. It uses HMD camera data and neural networks for effective hand tracking. However, limited user experience studies have compared this method to traditional controllers, with no consensus on the optimal control technique. This paper investigates the impact of control methods and gaming duration on VR user experience, hypothesizing hand tracking might be preferred for short sessions and by users new to VR due to its simplicity. Through a lab study with twenty participants, evaluating presence, emotional response, UX quality, and flow, findings revealed control type and session length affect user experience without significant interaction. Controllers were generally superior, attributed to their reliability, and longer sessions increased presence and realism. The study found that individuals with more VR experience were more inclined to recommend hand tracking to others, which contradicted predictions., Comment: Copyright held by the owner/author(s). Publication rights licensed to ACM. This is the author's version of the work. It is posted here for your personal use. Not for redistribution. The definitive Version of Record was published in Digital Library, https://doi.org/10.1145/3652212.3652222

A virtual embodiment can benefit conversational agents, but it is unclear how their personalities and non-verbal behavior influence the User Experience and Social Presence in Augmented Reality (AR). We asked 30 users to converse with a virtual assistant who gives recommendations about city activities. The participants interacted with two different personalities: Sammy, a cheerful blue mouse, and Olive, a serious green human-like agent. Each was presented with two body languages - happy/friendly and annoyed/unfriendly. We conclude how agent representation and humor affect User Experience aspects, and that body language is significant in the evaluation and perception of the AR agent.

Millions of people have seen their daily habits transform, reducing physical activity and leading to mental health issues. This study explores how virtual characters impact motivation for well-being. Three prototypes with cartoon, robotic, and human-like avatars were tested by 22 participants. Results show that animated virtual avatars, especially with extended reality, boost motivation, enhance comprehension of activities, and heighten presence. Multiple output modalities, like audio and text, with character animations, improve the user experience. Notably, the cartoon-like character evoked positive responses. This research highlights virtual characters' potential to engage individuals in daily well-being activities.

Nowadays, Augmented Reality (AR) is available on almost all smartphones creating some exciting interaction opportunities but also challenges. For example, already after the famous AR app Pokemon GO was released in July 2016, numerous accidents related to the use of the app were reported by users. At the same time, the spread of AR can be noticed in the tourism industry, enabling tourists to explore their surroundings in new ways but also exposing them to safety issues. This preliminary study explores users' sense of safety when manipulating the amount and UI elements visualization parameters of Point of Interest (POI) markers in a developed AR application. The results show that the amount of POI markers that are displayed is significant for participants' sense of safety. The influence of manipulating UI elements in terms of transparency, color, and size cannot be proven. Nevertheless, most tested people stated that manipulating transparency and size somehow influences their sense of safety, so a closer look at them should be taken in future studies.

Neuro-Symbolic (NeSy) predictors that conform to symbolic knowledge - encoding, e.g., safety constraints - can be affected by Reasoning Shortcuts (RSs): They learn concepts consistent with the symbolic knowledge by exploiting unintended semantics. RSs compromise reliability and generalization and, as we show in this paper, they are linked to NeSy models being overconfident about the predicted concepts. Unfortunately, the only trustworthy mitigation strategy requires collecting costly dense supervision over the concepts. Rather than attempting to avoid RSs altogether, we propose to ensure NeSy models are aware of the semantic ambiguity of the concepts they learn, thus enabling their users to identify and distrust low-quality concepts. Starting from three simple desiderata, we derive bears (BE Aware of Reasoning Shortcuts), an ensembling technique that calibrates the model's concept-level confidence without compromising prediction accuracy, thus encouraging NeSy architectures to be uncertain about concepts affected by RSs. We show empirically that bears improves RS-awareness of several state-of-the-art NeSy models, and also facilitates acquiring informative dense annotations for mitigation purposes.

Recent work showed the possibility of building open-vocabulary large language models (LLMs) that directly operate on pixel representations. These models are implemented as autoencoders that reconstruct masked patches of rendered text. However, these pixel-based LLMs are limited to discriminative tasks (e.g., classification) and, similar to BERT, cannot be used to generate text. Therefore, they cannot be used for generative tasks such as free-form question answering. In this work, we introduce PIXAR, the first pixel-based autoregressive LLM that performs text generation. Consisting of only a decoder, PIXAR can perform free-form generative tasks while keeping the number of parameters on par with previous encoder-decoder models. Furthermore, we highlight the challenges of generating text as non-noisy images and show this is due to using a maximum likelihood objective. To overcome this problem, we propose an adversarial pretraining stage that improves the readability and accuracy of PIXAR by 8.1 on LAMBADA and 8.5 on bAbI -- making it comparable to GPT-2 on text generation tasks. This paves the way to build open-vocabulary LLMs that operate on perceptual input only and calls into question the necessity of the usual symbolic input representation, i.e., text as (sub)tokens.

Personal sensing, leveraging data passively and near-continuously collected with wearables from patients in their ecological environment, is a promising paradigm to monitor mood disorders (MDs), a major determinant of worldwide disease burden. However, collecting and annotating wearable data is very resource-intensive. Studies of this kind can thus typically afford to recruit only a couple dozens of patients. This constitutes one of the major obstacles to applying modern supervised machine learning techniques to MDs detection. In this paper, we overcome this data bottleneck and advance the detection of MDs acute episode vs stable state from wearables data on the back of recent advances in self-supervised learning (SSL). This leverages unlabelled data to learn representations during pre-training, subsequently exploited for a supervised task. First, we collected open-access datasets recording with an Empatica E4 spanning different, unrelated to MD monitoring, personal sensing tasks -- from emotion recognition in Super Mario players to stress detection in undergraduates -- and devised a pre-processing pipeline performing on-/off-body detection, sleep-wake detection, segmentation, and (optionally) feature extraction. With 161 E4-recorded subjects, we introduce E4SelfLearning, the largest to date open access collection, and its pre-processing pipeline. Second, we show that SSL confidently outperforms fully-supervised pipelines using either our novel E4-tailored Transformer architecture (E4mer) or classical baseline XGBoost: 81.23% against 75.35% (E4mer) and 72.02% (XGBoost) correctly classified recording segments from 64 (half acute, half stable) patients. Lastly, we illustrate that SSL performance is strongly associated with the specific surrogate task employed for pre-training as well as with unlabelled data availability.

Continuous latent variables (LVs) are a key ingredient of many generative models, as they allow modelling expressive mixtures with an uncountable number of components. In contrast, probabilistic circuits (PCs) are hierarchical discrete mixtures represented as computational graphs composed of input, sum and product units. Unlike continuous LV models, PCs provide tractable inference but are limited to discrete LVs with categorical (i.e. unordered) states. We bridge these model classes by introducing probabilistic integral circuits (PICs), a new language of computational graphs that extends PCs with integral units representing continuous LVs. In the first place, PICs are symbolic computational graphs and are fully tractable in simple cases where analytical integration is possible. In practice, we parameterise PICs with light-weight neural nets delivering an intractable hierarchical continuous mixture that can be approximated arbitrarily well with large PCs using numerical quadrature. On several distribution estimation benchmarks, we show that such PIC-approximating PCs systematically outperform PCs commonly learned via expectation-maximization or SGD.

Sigmoid output layers are widely used in multi-label classification (MLC) tasks, in which multiple labels can be assigned to any input. In many practical MLC tasks, the number of possible labels is in the thousands, often exceeding the number of input features and resulting in a low-rank output layer. In multi-class classification, it is known that such a low-rank output layer is a bottleneck that can result in unargmaxable classes: classes which cannot be predicted for any input. In this paper, we show that for MLC tasks, the analogous sigmoid bottleneck results in exponentially many unargmaxable label combinations. We explain how to detect these unargmaxable outputs and demonstrate their presence in three widely used MLC datasets. We then show that they can be prevented in practice by introducing a Discrete Fourier Transform (DFT) output layer, which guarantees that all sparse label combinations with up to $k$ active labels are argmaxable. Our DFT layer trains faster and is more parameter efficient, matching the F1@k score of a sigmoid layer while using up to 50% fewer trainable parameters. Our code is publicly available at https://github.com/andreasgrv/sigmoid-bottleneck., Comment: Published at AAAI24

Mixture models are traditionally represented and learned by adding several distributions as components. Allowing mixtures to subtract probability mass or density can drastically reduce the number of components needed to model complex distributions. However, learning such subtractive mixtures while ensuring they still encode a non-negative function is challenging. We investigate how to learn and perform inference on deep subtractive mixtures by squaring them. We do this in the framework of probabilistic circuits, which enable us to represent tensorized mixtures and generalize several other subtractive models. We theoretically prove that the class of squared circuits allowing subtractions can be exponentially more expressive than traditional additive mixtures; and, we empirically show this increased expressiveness on a series of real-world distribution estimation tasks.

Knowledge graphs are powerful tools for representing and organising complex biomedical data. Several knowledge graph embedding algorithms have been proposed to learn from and complete knowledge graphs. However, a recent study demonstrates the limited efficacy of these embedding algorithms when applied to biomedical knowledge graphs, raising the question of whether knowledge graph embeddings have limitations in biomedical settings. This study aims to apply state-of-the-art knowledge graph embedding models in the context of a recent biomedical knowledge graph, BioKG, and evaluate their performance and potential downstream uses. We achieve a three-fold improvement in terms of performance based on the HITS@10 score over previous work on the same biomedical knowledge graph. Additionally, we provide interpretable predictions through a rule-based method. We demonstrate that knowledge graph embedding models are applicable in practice by evaluating the best-performing model on four tasks that represent real-life polypharmacy situations. Results suggest that knowledge learnt from large biomedical knowledge graphs can be transferred to such downstream use cases. Our code is available at https://github.com/aryopg/biokge.

Neuro-Symbolic (NeSy) predictive models hold the promise of improved compliance with given constraints, systematic generalization, and interpretability, as they allow to infer labels that are consistent with some prior knowledge by reasoning over high-level concepts extracted from sub-symbolic inputs. It was recently shown that NeSy predictors are affected by reasoning shortcuts: they can attain high accuracy but by leveraging concepts with unintended semantics, thus coming short of their promised advantages. Yet, a systematic characterization of reasoning shortcuts and of potential mitigation strategies is missing. This work fills this gap by characterizing them as unintended optima of the learning objective and identifying four key conditions behind their occurrence. Based on this, we derive several natural mitigation strategies, and analyze their efficacy both theoretically and empirically. Our analysis shows reasoning shortcuts are difficult to deal with, casting doubts on the trustworthiness and interpretability of existing NeSy solutions., Comment: 37th Conference on Neural Information Processing Systems (NeurIPS 2023)

Some of the most successful knowledge graph embedding (KGE) models for link prediction -- CP, RESCAL, TuckER, ComplEx -- can be interpreted as energy-based models. Under this perspective they are not amenable for exact maximum-likelihood estimation (MLE), sampling and struggle to integrate logical constraints. This work re-interprets the score functions of these KGEs as circuits -- constrained computational graphs allowing efficient marginalisation. Then, we design two recipes to obtain efficient generative circuit models by either restricting their activations to be non-negative or squaring their outputs. Our interpretation comes with little or no loss of performance for link prediction, while the circuits framework unlocks exact learning by MLE, efficient sampling of new triples, and guarantee that logical constraints are satisfied by design. Furthermore, our models scale more gracefully than the original KGEs on graphs with millions of entities.

Continual learning (CL) is one of the most promising trends in recent machine learning research. Its goal is to go beyond classical assumptions in machine learning and develop models and learning strategies that present high robustness in dynamic environments. The landscape of CL research is fragmented into several learning evaluation protocols, comprising different learning tasks, datasets, and evaluation metrics. Additionally, the benchmarks adopted so far are still distant from the complexity of real-world scenarios, and are usually tailored to highlight capabilities specific to certain strategies. In such a landscape, it is hard to objectively assess strategies. In this work, we fill this gap for CL on image data by introducing two novel CL benchmarks that involve multiple heterogeneous tasks from six image datasets, with varying levels of complexity and quality. Our aim is to fairly evaluate current state-of-the-art CL strategies on a common ground that is closer to complex real-world scenarios. We additionally structure our benchmarks so that tasks are presented in increasing and decreasing order of complexity -- according to a curriculum -- in order to evaluate if current CL models are able to exploit structure across tasks. We devote particular emphasis to providing the CL community with a rigorous and reproducible evaluation protocol for measuring the ability of a model to generalize and not to forget while learning. Furthermore, we provide an extensive experimental evaluation showing that popular CL strategies, when challenged with our benchmarks, yield sub-par performance, high levels of forgetting, and present a limited ability to effectively leverage curriculum task ordering. We believe that these results highlight the need for rigorous comparisons in future CL works as well as pave the way to design new CL strategies that are able to deal with more complex scenarios.