There is an emerging interest in generating robust counterfactual explanations that would remain valid if the model is updated or changed even slightly. Towards finding robust counterfactuals, existing literature often assumes that the original model $m$ and the new model $M$ are bounded in the parameter space, i.e., $\|\text{Params}(M){-}\text{Params}(m)\|{<}\Delta$. However, models can often change significantly in the parameter space with little to no change in their predictions or accuracy on the given dataset. In this work, we introduce a mathematical abstraction termed $\textit{naturally-occurring}$ model change, which allows for arbitrary changes in the parameter space such that the change in predictions on points that lie on the data manifold is limited. Next, we propose a measure -- that we call $\textit{Stability}$ -- to quantify the robustness of counterfactuals to potential model changes for differentiable models, e.g., neural networks. Our main contribution is to show that counterfactuals with sufficiently high value of $\textit{Stability}$ as defined by our measure will remain valid after potential $\textit{naturally-occurring}$ model changes with high probability (leveraging concentration bounds for Lipschitz function of independent Gaussians). Since our quantification depends on the local Lipschitz constant around a data point which is not always available, we also examine practical relaxations of our proposed measure and demonstrate experimentally how they can be incorporated to find robust counterfactuals for neural networks that are close, realistic, and remain valid after potential model changes. This work also has interesting connections with model multiplicity, also known as, the Rashomon effect.
Natural language understanding includes the tasks of intent detection (identifying a user's objectives) and slot filling (extracting the entities relevant to those objectives). Prior slot filling methods assume that each intent type cannot occur more than once within a message, however this is often not a valid assumption for real-world settings. In this work, we generalize slot filling by removing the constraint of unique intents in a message. We cast this as a JSON generation task and approach it using a language model. We create a pre-training dataset by combining DBpedia and existing slot filling datasets that we convert for JSON generation. We also generate an in-domain dataset using GPT-3. We train T5 models for this task (with and without exemplars in the prompt) and find that both training datasets improve performance, and that the model is able to generalize to intent types not seen during training.
Networked sensing refers to the capability of properly orchestrating multiple sensing terminals to enhance specific figures of merit, e.g., positioning accuracy or imaging resolution. Regarding radio-based sensing, it is essential to understand \textit{when} and \textit{how} sensing terminals should be orchestrated, namely the best cooperation that trades between performance and cost (e.g., energy consumption, communication overhead, and complexity). This paper addresses networked sensing from a physics-driven perspective, aiming to provide a general theoretical benchmark to evaluate its \textit{imaging} performance bounds and to guide the sensing orchestration accordingly. Diffraction tomography theory (DTT) is the method to quantify the imaging resolution of any radio sensing experiment from inspection of its spectral (or wavenumber) content. In networked sensing, the image formation is based on the back-projection integral, valid for any network topology and physical configuration of the terminals. The \textit{wavefield networked sensing} is a framework in which multiple sensing terminals are orchestrated during the acquisition process to maximize the imaging quality (resolution and grating lobes suppression) by pursuing the deceptively simple \textit{wavenumber tessellation principle}. We discuss all the cooperation possibilities between sensing terminals and possible killer applications. Remarkably, we show that the proposed method allows obtaining high-quality images of the environment in limited bandwidth conditions, leveraging the coherent combination of multiple multi-static low-resolution images.
We propose a new approach for sampling and Bayesian computation that uses the score of the target distribution to construct a transport from a given reference distribution to the target. Our approach is an infinite-dimensional Newton method, involving a linear PDE, for finding a zero of a ``score-residual'' operator. We prove sufficient conditions for convergence to a valid transport map. Our Newton iterates can be computed by exploiting fast solvers for elliptic PDEs, resulting in new algorithms for Bayesian inference and other sampling tasks. We identify elementary settings where score-operator Newton transport achieves fast convergence while avoiding mode collapse.
Generative models for molecules based on sequential line notation (e.g. SMILES) or graph representation have attracted an increasing interest in the field of structure-based drug design, but they struggle to capture important 3D spatial interactions and often produce undesirable molecular structures. To address these challenges, we introduce Lingo3DMol, a pocket-based 3D molecule generation method that combines language models and geometric deep learning technology. A new molecular representation, fragment-based SMILES with local and global coordinates, was developed to assist the model in learning molecular topologies and atomic spatial positions. Additionally, we trained a separate noncovalent interaction predictor to provide essential binding pattern information for the generative model. Lingo3DMol can efficiently traverse drug-like chemical spaces, preventing the formation of unusual structures. The Directory of Useful Decoys-Enhanced (DUD-E) dataset was used for evaluation. Lingo3DMol outperformed state-of-the-art methods in terms of drug-likeness, synthetic accessibility, pocket binding mode, and molecule generation speed.
In partial label learning (PLL), each training sample is associated with a set of candidate labels, among which only one is valid. The core of PLL is to disambiguate the candidate labels to get the ground-truth one. In disambiguation, the existing works usually do not fully investigate the effectiveness of the non-candidate label set (a.k.a. complementary labels), which accurately indicates a set of labels that do not belong to a sample. In this paper, we use the non-candidate labels to induce a complementary classifier, which naturally forms an adversarial relationship against the traditional PLL classifier, to eliminate the false-positive labels in the candidate label set. Besides, we assume the feature space and the label space share the same local topological structure captured by a dynamic graph, and use it to assist disambiguation. Extensive experimental results validate the superiority of the proposed approach against state-of-the-art PLL methods on 4 controlled UCI data sets and 6 real-world data sets, and reveal the usefulness of complementary learning in PLL. The code has been released in the link https://github.com/Chongjie-Si/PL-CL.
This paper addresses the $\epsilon$-close parameter tuning problem for Bayesian Networks (BNs): find a minimal $\epsilon$-close amendment of probability entries in a given set of (rows in) conditional probability tables that make a given quantitative constraint on the BN valid. Based on the state-of-the-art "region verification" techniques for parametric Markov chains, we propose an algorithm whose capabilities go beyond any existing techniques. Our experiments show that $\epsilon$-close tuning of large BN benchmarks with up to 8 parameters is feasible. In particular, by allowing (i) varied parameters in multiple CPTs and (ii) inter-CPT parameter dependencies, we treat subclasses of parametric BNs that have received scant attention so far.
Recently, Koopman operator theory has become a powerful tool for developing linear representations of non-linear dynamical systems. However, existing data-driven applications of Koopman operator theory, including both traditional and deep learning approaches, perform poorly on non-linear network dynamics problems as they do not address the underlying geometric structure. In this paper we present a novel approach based on Koopman operator theory and message passing networks that finds a linear representation for the dynamical system which is globally valid at any time step. The linearisations found by our method produce predictions on a suite of network dynamics problems that are several orders of magnitude better than current state-of-the-art techniques. We also apply our approach to the highly non-linear training dynamics of neural network architectures, and obtain linear representations which can generate network parameters with comparable performance to networks trained by classical optimisers.
The Smatch metric is a popular method for evaluating graph distances, as is necessary, for instance, to assess the performance of semantic graph parsing systems. However, we observe some issues in the metric that jeopardize meaningful evaluation. E.g., opaque pre-processing choices can affect results, and current graph-alignment solvers do not provide us with upper-bounds. Without upper-bounds, however, fair evaluation is not guaranteed. Furthermore, adaptions of Smatch for extended tasks (e.g., fine-grained semantic similarity) are spread out, and lack a unifying framework. For better inspection, we divide the metric into three modules: pre-processing, alignment, and scoring. Examining each module, we specify its goals and diagnose potential issues, for which we discuss and test mitigation strategies. For pre-processing, we show how to fully conform to annotation guidelines that allow structurally deviating but valid graphs. For safer and enhanced alignment, we show the feasibility of optimal alignment in a standard evaluation setup, and develop a lossless graph compression method that shrinks the search space and significantly increases efficiency. For improved scoring, we propose standardized and extended metric calculation of fine-grained sub-graph meaning aspects. Our code is available at https://github.com/flipz357/smatchpp
Spectral-temporal graph neural network is a promising abstraction underlying most time series forecasting models that are based on graph neural networks (GNNs). However, more is needed to know about the underpinnings of this branch of methods. In this paper, we establish a theoretical framework that unravels the expressive power of spectral-temporal GNNs. Our results show that linear spectral-temporal GNNs are universal under mild assumptions, and their expressive power is bounded by our extended first-order Weisfeiler-Leman algorithm on discrete-time dynamic graphs. To make our findings useful in practice on valid instantiations, we discuss related constraints in detail and outline a theoretical blueprint for designing spatial and temporal modules in spectral domains. Building on these insights and to demonstrate how powerful spectral-temporal GNNs are based on our framework, we propose a simple instantiation named Temporal Graph GegenConv (TGC), which significantly outperforms most existing models with only linear components and shows better model efficiency.
Large language models (LLMs) have been widely studied for their ability to store and utilize positive knowledge. However, negative knowledge, such as "lions don't live in the ocean", is also ubiquitous in the world but rarely mentioned explicitly in the text. What do LLMs know about negative knowledge? This work examines the ability of LLMs to negative commonsense knowledge. We design a constrained keywords-to-sentence generation task (CG) and a Boolean question-answering task (QA) to probe LLMs. Our experiments reveal that LLMs frequently fail to generate valid sentences grounded in negative commonsense knowledge, yet they can correctly answer polar yes-or-no questions. We term this phenomenon the belief conflict of LLMs. Our further analysis shows that statistical shortcuts and negation reporting bias from language modeling pre-training cause this conflict.
High complexity models are notorious in machine learning for overfitting, a phenomenon in which models well represent data but fail to generalize an underlying data generating process. A typical procedure for circumventing overfitting computes empirical risk on a holdout set and halts once (or flags that/when) it begins to increase. Such practice often helps in outputting a well-generalizing model, but justification for why it works is primarily heuristic. We discuss the overfitting problem and explain why standard asymptotic and concentration results do not hold for evaluation with training data. We then proceed to introduce and argue for a hypothesis test by means of which both model performance may be evaluated using training data, and overfitting quantitatively defined and detected. We rely on said concentration bounds which guarantee that empirical means should, with high probability, approximate their true mean to conclude that they should approximate each other. We stipulate conditions under which this test is valid, describe how the test may be used for identifying overfitting, articulate a further nuance according to which distributional shift may be flagged, and highlight an alternative notion of learning which usefully captures generalization in the absence of uniform PAC guarantees.
Language models are powerful tools for molecular design. Currently, the dominant paradigm is to parse molecular graphs into linear string representations that can easily be trained on. This approach has been very successful, however, it is limited to chemical structures that can be completely represented by a graph -- like organic molecules -- while materials and biomolecular structures like protein binding sites require a more complete representation that includes the relative positioning of their atoms in space. In this work, we show how language models, without any architecture modifications, trained using next-token prediction -- can generate novel and valid structures in three dimensions from various substantially different distributions of chemical structures. In particular, we demonstrate that language models trained directly on sequences derived directly from chemical file formats like XYZ files, Crystallographic Information files (CIFs), or Protein Data Bank files (PDBs) can directly generate molecules, crystals, and protein binding sites in three dimensions. Furthermore, despite being trained on chemical file sequences -- language models still achieve performance comparable to state-of-the-art models that use graph and graph-derived string representations, as well as other domain-specific 3D generative models. In doing so, we demonstrate that it is not necessary to use simplified molecular representations to train chemical language models -- that they are powerful generative models capable of directly exploring chemical space in three dimensions for very different structures.
Automated feedback as students answer open-ended math questions has significant potential in improving learning outcomes at large scale. A key part of automated feedback systems is an error classification component, which identifies student errors and enables appropriate, predefined feedback to be deployed. Most existing approaches to error classification use a rule-based method, which has limited capacity to generalize. Existing data-driven methods avoid these limitations but specifically require mathematical expressions in student responses to be parsed into syntax trees. This requirement is itself a limitation, since student responses are not always syntactically valid and cannot be converted into trees. In this work, we introduce a flexible method for error classification using pre-trained large language models. We demonstrate that our method can outperform existing methods in algebra error classification, and is able to classify a larger set of student responses. Additionally, we analyze common classification errors made by our method and discuss limitations of automated error classification.
Comparative knowledge (e.g., steel is stronger and heavier than styrofoam) is an essential component of our world knowledge, yet understudied in prior literature. In this paper, we harvest the dramatic improvements in knowledge capabilities of language models into a large-scale comparative knowledge base. While the ease of acquisition of such comparative knowledge is much higher from extreme-scale models like GPT-4, compared to their considerably smaller and weaker counterparts such as GPT-2, not even the most powerful models are exempt from making errors. We thus ask: to what extent are models at different scales able to generate valid and diverse comparative knowledge? We introduce NeuroComparatives, a novel framework for comparative knowledge distillation overgenerated from language models such as GPT-variants and LLaMA, followed by stringent filtering of the generated knowledge. Our framework acquires comparative knowledge between everyday objects, producing a corpus of up to 8.8M comparisons over 1.74M entity pairs - 10X larger and 30% more diverse than existing resources. Moreover, human evaluations show that NeuroComparatives outperform existing resources in terms of validity (up to 32% absolute improvement). Our acquired NeuroComparatives leads to performance improvements on five downstream tasks. We find that neuro-symbolic manipulation of smaller models offers complementary benefits to the currently dominant practice of prompting extreme-scale language models for knowledge distillation.
Recent advances in self-supervised learning and neural network scaling have enabled the creation of large models, known as foundation models, which can be easily adapted to a wide range of downstream tasks. The current paradigm for comparing foundation models involves evaluating them with aggregate metrics on various benchmark datasets. This method of model comparison is heavily dependent on the chosen evaluation metric, which makes it unsuitable for situations where the ideal metric is either not obvious or unavailable. In this work, we present a methodology for directly comparing the embedding space geometry of foundation models, which facilitates model comparison without the need for an explicit evaluation metric. Our methodology is grounded in random graph theory and enables valid hypothesis testing of embedding similarity on a per-datum basis. Further, we demonstrate how our methodology can be extended to facilitate population level model comparison. In particular, we show how our framework can induce a manifold of models equipped with a distance function that correlates strongly with several downstream metrics. We remark on the utility of this population level model comparison as a first step towards a taxonomic science of foundation models.
In a response to the 2022 cost-of-living crisis in Europe, the German government implemented a three-month fuel excise tax cut and a public transport travel pass for 9 Euro per month valid on all local and regional services. Following this period, a public debate immediately emerged on a successor to the so-called "9-Euro-Ticket", leading to the political decision of introducing a similar ticket priced at 49 Euro per month in May 2023, the so-called "Deutschlandticket". We observe this introduction of the new public transport ticket with a sample of 818 participants using a smartphone-based travel diary with passive tracking and a two-wave survey. The sample comprises 510 remaining participants of our initial "9-Euro-Ticket study from 2022 and 308 participants recruited in March and early April 2023. In this report we report on the status of the panel before the introduction of the "Deutschlandticket".
Existing MWP solvers employ sequence or binary tree to present the solution expression and decode it from given problem description. However, such structures fail to handle the variants that can be derived via mathematical manipulation, e.g., $(a_1+a_2) * a_3$ and $a_1 * a_3+a_2 * a_3$ can both be possible valid solutions for a same problem but formulated as different expression sequences or trees. The multiple solution variants depicting different possible solving procedures for the same input problem would raise two issues: 1) making it hard for the model to learn the mapping function between the input and output spaces effectively, and 2) wrongly indicating \textit{wrong} when evaluating a valid expression variant. To address these issues, we introduce a unified tree structure to present a solution expression, where the elements are permutable and identical for all the expression variants. We propose a novel non-autoregressive solver, named \textit{MWP-NAS}, to parse the problem and deduce the solution expression based on the unified tree. For evaluating the possible expression variants, we design a path-based metric to evaluate the partial accuracy of expressions of a unified tree. The results from extensive experiments conducted on Math23K and MAWPS demonstrate the effectiveness of our proposed MWP-NAS. The codes and checkpoints are available at: \url{https://github.com/mengqunhan/MWP-NAS}.
Since 2010, multiple cyber incidents on industrial infrastructure, such as Stuxnet and CrashOverride, have exposed the vulnerability of Industrial Control Systems (ICS) to cyber threats. The industrial systems are commissioned for longer duration amounting to decades, often resulting in non-compliance to technological advancements in industrial cybersecurity mechanisms. The unavailability of network infrastructure information makes designing the security policies or configuring the cybersecurity countermeasures such as Network Intrusion Detection Systems (NIDS) challenging. An empirical solution is to self-learn the network infrastructure information of an industrial system from its monitored network traffic to make the network transparent for downstream analyses tasks such as anomaly detection. In this work, a Python-based industrial communication paradigm-aware framework, named PROFINET Operations Enumeration and Tracking (POET), that enumerates different industrial operations executed in a deterministic order of a PROFINET-based industrial system is reported. The operation-driving industrial network protocol frames are dissected for enumeration of the operations. For the requirements of capturing the transitions between industrial operations triggered by the communication events, the Finite State Machines (FSM) are modelled to enumerate the PROFINET operations of the device, connection and system. POET extracts the network information from network traffic to instantiate appropriate FSM models (Device, Connection or System) and track the industrial operations. It successfully detects and reports the anomalies triggered by a network attack in a miniaturized PROFINET-based industrial system, executed through valid network protocol exchanges and resulting in invalid PROFINET operation transition for the device.
Although large language models demonstrate remarkable question-answering performances, revealing the intermediate reasoning steps that the models faithfully follow remains challenging. In this paper, we propose FAME (FAithful question answering with MontE-carlo planning) to answer questions based on faithful reasoning steps. The reasoning steps are organized as a structured entailment tree, which shows how premises are used to produce intermediate conclusions that can prove the correctness of the answer. We formulate the task as a discrete decision-making problem and solve it through the interaction of a reasoning environment and a controller. The environment is modular and contains several basic task-oriented modules, while the controller proposes actions to assemble the modules. Since the search space could be large, we introduce a Monte-Carlo planning algorithm to do a look-ahead search and select actions that will eventually lead to high-quality steps. FAME achieves state-of-the-art performance on the standard benchmark. It can produce valid and faithful reasoning steps compared with large language models with a much smaller model size.