This paper develops tests for the correct specification of the conditional variance function in GARCH models when the true parameter may lie on the boundary of the parameter space. The test statistics considered are of Kolmogorov-Smirnov and Cram\'{e}r-von Mises type, and are based on a certain empirical process marked by centered squared residuals. The limiting distributions of the test statistics are not free from (unknown) nuisance parameters, and hence critical values cannot be tabulated. A novel bootstrap procedure is proposed to implement the tests; it is shown to be asymptotically valid under general conditions, irrespective of the presence of nuisance parameters on the boundary. The proposed bootstrap approach is based on shrinking of the parameter estimates used to generate the bootstrap sample toward the boundary of the parameter space at a proper rate. It is simple to implement and fast in applications, as the associated test statistics have simple closed form expressions. A simulation study demonstrates that the new tests: (i) have excellent finite sample behavior in terms of empirical rejection probabilities under the null as well as under the alternative; (ii) provide a useful complement to existing procedures based on Ljung-Box type approaches. Two data examples are considered to illustrate the tests.
Modern task-oriented semantic parsing approaches typically use seq2seq transformers to map textual utterances to semantic frames comprised of intents and slots. While these models are empirically strong, their specific strengths and weaknesses have largely remained unexplored. In this work, we study BART and XLM-R, two state-of-the-art parsers, across both monolingual and multilingual settings. Our experiments yield several key results: transformer-based parsers struggle not only with disambiguating intents/slots, but surprisingly also with producing syntactically-valid frames. Though pre-training imbues transformers with syntactic inductive biases, we find the ambiguity of copying utterance spans into frames often leads to tree invalidity, indicating span extraction is a major bottleneck for current parsers. However, as a silver lining, we show transformer-based parsers give sufficient indicators for whether a frame is likely to be correct or incorrect, making them easier to deploy in production settings.
As a prevailing task in video surveillance and forensics field, person re-identification (re-ID) aims to match person images captured from non-overlapped cameras. In unconstrained scenarios, person images often suffer from the resolution mismatch problem, i.e., \emph{Cross-Resolution Person Re-ID}. To overcome this problem, most existing methods restore low resolution (LR) images to high resolution (HR) by super-resolution (SR). However, they only focus on the HR feature extraction and ignore the valid information from original LR images. In this work, we explore the influence of resolutions on feature extraction and develop a novel method for cross-resolution person re-ID called \emph{\textbf{M}ulti-Resolution \textbf{R}epresentations \textbf{J}oint \textbf{L}earning} (\textbf{MRJL}). Our method consists of a Resolution Reconstruction Network (RRN) and a Dual Feature Fusion Network (DFFN). The RRN uses an input image to construct a HR version and a LR version with an encoder and two decoders, while the DFFN adopts a dual-branch structure to generate person representations from multi-resolution images. Comprehensive experiments on five benchmarks verify the superiority of the proposed MRJL over the relevent state-of-the-art methods.
Knowledge retrieval and reasoning are two key stages in multi-hop question answering (QA) at web scale. Existing approaches suffer from low confidence when retrieving evidence facts to fill the knowledge gap and lack transparent reasoning process. In this paper, we propose a new framework to exploit more valid facts while obtaining explainability for multi-hop QA by dynamically constructing a semantic graph and reasoning over it. We employ Abstract Meaning Representation (AMR) as semantic graph representation. Our framework contains three new ideas: (a) {\tt AMR-SG}, an AMR-based Semantic Graph, constructed by candidate fact AMRs to uncover any hop relations among question, answer and multiple facts. (b) A novel path-based fact analytics approach exploiting {\tt AMR-SG} to extract active facts from a large fact pool to answer questions. (c) A fact-level relation modeling leveraging graph convolution network (GCN) to guide the reasoning process. Results on two scientific multi-hop QA datasets show that we can surpass recent approaches including those using additional knowledge graphs while maintaining high explainability on OpenBookQA and achieve a new state-of-the-art result on ARC-Challenge in a computationally practicable setting.
We propose a method for conducting asymptotically valid inference for treatment effects in a multi-valued treatment framework where the number of units in the treatment arms can be small and do not grow with the sample size. We accomplish this by casting the model as a semi-/non-parametric conditional quantile model and using known finite sample results about the law of the indicator function that defines the conditional quantile. Our framework allows for structural functions that are non-additively separable, with flexible functional forms and heteroskedasticy in the residuals, and it also encompasses commonly used designs like difference in difference. We study the finite sample behavior of our test in a Monte Carlo study and we also apply our results to assessing the effect of weather events on GDP growth.
Conformal predictors are an important class of algorithms that allow predictions to be made with a user-defined confidence level. They are able to do this by outputting prediction sets, rather than simple point predictions. The conformal predictor is valid in the sense that the accuracy of its predictions is guaranteed to meet the confidence level, only assuming exchangeability in the data. Since accuracy is guaranteed, the performance of a conformal predictor is measured through the efficiency of the prediction sets. Typically, a conformal predictor is built on an underlying machine learning algorithm and hence its predictive power is inherited from this algorithm. However, since the underlying machine learning algorithm is not trained with the objective of minimizing predictive efficiency it means that the resulting conformal predictor may be sub-optimal and not aligned sufficiently to this objective. Hence, in this study we consider an approach to train the conformal predictor directly with maximum predictive efficiency as the optimization objective, and we focus specifically on the inductive conformal predictor for classification. To do this, the conformal predictor is approximated by a differentiable objective function and gradient descent used to optimize it. The resulting parameter estimates are then passed to a proper inductive conformal predictor to give valid prediction sets. We test the method on several real world data sets and find that the method is promising and in most cases gives improved predictive efficiency against a baseline conformal predictor.
Due to the practical importance of vehicle routing problems (VRP), there exists an ever-growing body of research in algorithms and (meta)heuristics for solving such problems. However, the diversity of VRP domains creates the separate problem of modeling such problems -- describing the domain entities (and, in particular, the planning decisions), the set of valid planning decisions, and the preferences between different plans. In this paper, we review the approaches for modeling vehicle routing problems. To make the comparison more straightforward, we formulate several criteria for evaluating modeling methods reflecting the practical requirements of the development of optimization algorithms for such problems. Finally, as a result of this comparison, we discuss several future research avenues in the field of modeling VRP domains.
We introduce ASQ, a tool to automatically mine questions and answers from a sentence using the Abstract Meaning Representation (AMR). Previous work has used question-answer pairs to specify the predicate-argument structure of a sentence using natural language, which does not require linguistic expertise or training, and created datasets such as QA-SRL and QAMR, for which the question-answer pair annotations were crowdsourced. Our goal is to build a tool (ASQ) that maps from the traditional meaning representation AMR to a question-answer meaning representation (QMR). This enables construction of QMR datasets automatically in various domains using existing high-quality AMR parsers, and provides an automatic mapping AMR to QMR for ease of understanding by non-experts. A qualitative evaluation of the output generated by ASQ from the AMR 2.0 data shows that the question-answer pairs are natural and valid, and demonstrate good coverage of the content. We run ASQ on the sentences from the QAMR dataset, to observe that the semantic roles in QAMR are also captured by ASQ. We intend to make this tool and the results publicly available for others to use and build upon.
This article is motivated by the objective of providing a new analytically tractable and fully frequentist framework to characterize and implement regression trees while also allowing a multivariate (potentially high dimensional) response. The connection to regression trees is made by a high dimensional model with dynamic mean vectors over multi-dimensional change axes. Our theoretical analysis is carried out under a single two dimensional change point setting. An optimal rate of convergence of the proposed estimator is obtained, which in turn allows existence of limiting distributions. Distributional behavior of change point estimates are split into two distinct regimes, the limiting distributions under each regime is then characterized, in turn allowing construction of asymptotically valid confidence intervals for $2d$-location of change. All results are obtained under a high dimensional scaling $s\log^2 p=o(T_wT_h),$ where $p$ is the response dimension, $s$ is a sparsity parameter, and $T_w,T_h$ are sampling periods along change axes. We characterize full regression trees by defining a multiple multi-dimensional change point model. Natural extensions of the single $2d$-change point estimation methodology are provided. Two applications, first on segmentation of {\it Infra-red astronomy satellite (IRAS)} data and second to segmentation of digital images are provided. Methodology and theoretical results are supported with monte-carlo simulations.
This paper investigates whether computer usage profiles comprised of process-, network-, mouse-, and keystroke-related events are unique and consistent over time in a naturalistic setting, discussing challenges and opportunities of using such profiles in applications of continuous authentication. We collected ecologically-valid computer usage profiles from 31 MS Windows 10 computer users over 8 weeks and submitted this data to comprehensive machine learning analysis involving a diverse set of online and offline classifiers. We found that: (i) profiles were mostly consistent over the 8-week data collection period, with most (83.9%) repeating computer usage habits on a daily basis; (ii) computer usage profiling has the potential to uniquely characterize computer users (with a maximum F-score of 99.90%); (iii) network-related events were the most relevant features to accurately recognize profiles (95.69% of the top features distinguishing users were network-related); and (iv) binary models were the most well-suited for profile recognition, with better results achieved in the online setting compared to the offline setting (maximum F-score of 99.90% vs. 95.50%).
With the ever-increasing popularity and applications of graph neural networks, several proposals have been made to explain and understand the decisions of a graph neural network. Explanations for graph neural networks differ in principle from other input settings. It is important to attribute the decision to input features and other related instances connected by the graph structure. We find that the previous explanation generation approaches that maximize the mutual information between the label distribution produced by the model and the explanation to be restrictive. Specifically, existing approaches do not enforce explanations to be valid, sparse, or robust to input perturbations. In this paper, we lay down some of the fundamental principles that an explanation method for graph neural networks should follow and introduce a metric RDT-Fidelity as a measure of the explanation's effectiveness. We propose a novel approach Zorro based on the principles from rate-distortion theory that uses a simple combinatorial procedure to optimize for RDT-Fidelity. Extensive experiments on real and synthetic datasets reveal that Zorro produces sparser, stable, and more faithful explanations than existing graph neural network explanation approaches.
Machine learning is vulnerable to adversarial examples-inputs designed to cause models to perform poorly. However, it is unclear if adversarial examples represent realistic inputs in the modeled domains. Diverse domains such as networks and phishing have domain constraints-complex relationships between features that an adversary must satisfy for an attack to be realized (in addition to any adversary-specific goals). In this paper, we explore how domain constraints limit adversarial capabilities and how adversaries can adapt their strategies to create realistic (constraint-compliant) examples. In this, we develop techniques to learn domain constraints from data, and show how the learned constraints can be integrated into the adversarial crafting process. We evaluate the efficacy of our approach in network intrusion and phishing datasets and find: (1) up to 82% of adversarial examples produced by state-of-the-art crafting algorithms violate domain constraints, (2) domain constraints are robust to adversarial examples; enforcing constraints yields an increase in model accuracy by up to 34%. We observe not only that adversaries must alter inputs to satisfy domain constraints, but that these constraints make the generation of valid adversarial examples far more challenging.
Reconfiguration aims at recovering a system from a fault by automatically adapting the system configuration, such that the system goal can be reached again. Classical approaches typically use a set of pre-defined faults for which corresponding recovery actions are defined manually. This is not possible for modern hybrid systems which are characterized by frequent changes. Instead, AI-based approaches are needed which leverage on a model of the non-faulty system and which search for a set of reconfiguration operations which will establish a valid behavior again. This work presents a novel algorithm which solves three main challenges: (i) Only a model of the non-faulty system is needed, i.e. the faulty behavior does not need to be modeled. (ii) It discretizes and reduces the search space which originally is too large -- mainly due to the high number of continuous system variables and control signals. (iii) It uses a SAT solver for propositional logic for two purposes: First, it defines the binary concept of validity. Second, it implements the search itself -- sacrificing the optimal solution for a quick identification of an arbitrary solution. It is shown that the approach is able to reconfigure faults on simulated process engineering systems.
Motor imagery (MI)-based brain-computer interface (BCI) systems are being increasingly employed to provide alternative means of communication and control for people suffering from neuro-motor impairments, with a special effort to bring these systems out of the controlled lab environments. Hence, accurately classifying MI from brain signals, e.g., from electroencephalography (EEG), is essential to obtain reliable BCI systems. However, MI classification is still a challenging task, because the signals are characterized by poor SNR, high intra-subject and cross-subject variability. Deep learning approaches have started to emerge as valid alternatives to standard machine learning techniques, e.g., filter bank common spatial pattern (FBCSP), to extract subject-independent features and to increase the cross-subject classification performance of MI BCI systems. In this paper, we first present a review of the most recent studies using deep learning for MI classification, with particular attention to their cross-subject performance. Second, we propose DynamicNet, a Python-based tool for quick and flexible implementations of deep learning models based on convolutional neural networks. We show-case the potentiality of DynamicNet by implementing EEGNet, a well-established architecture for effective EEG classification. Finally, we compare its performance with FBCSP in a 4-class MI classification over public datasets. To explore its cross-subject classification ability, we applied three different cross-validation schemes. From our results, we demonstrate that DynamicNet-implemented EEGNet outperforms FBCSP by about 25%, with a statistically significant difference when cross-subject validation schemes are applied.
Previous transfer methods for anomaly detection generally assume the availability of labeled data in source or target domains. However, such an assumption is not valid in most real applications where large-scale labeled data are too expensive. Therefore, this paper proposes an importance weighted adversarial autoencoder-based method to transfer anomaly detection knowledge in an unsupervised manner, particularly for a rarely studied scenario where a target domain has no labeled normal/abnormal data while only normal data from a related source domain exist. Specifically, the method learns to align the distributions of normal data in both source and target domains, but leave the distribution of abnormal data in the target domain unchanged. In this way, an obvious gap can be produced between the distributions of normal and abnormal data in the target domain, therefore enabling the anomaly detection in the domain. Extensive experiments on multiple synthetic datasets and the UCSD benchmark demonstrate the effectiveness of our approach.
In this paper, we provide simplified semantics for the logic K45(G), i.e. the many-valued Godel counterpart of the classical modal logic K45. More precisely, we characterize K45(G) as the set of valid formulae of the class of possibilistic Godel Kripke Frames <W,\pi> where W is a non-empty set of worlds and \pi: W \to [0, 1] is a possibility distribution on W.
Polymers are widely-studied materials with diverse properties and applications determined by different molecular structures. It is essential to represent these structures clearly and explore the full space of achievable chemical designs. However, existing approaches are unable to offer comprehensive design models for polymers because of their inherent scale and structural complexity. Here, we present a parametric, context-sensitive grammar designed specifically for the representation and generation of polymers. As a demonstrative example, we implement our grammar for polyurethanes. Using our symbolic hypergraph representation and 14 simple production rules, our PolyGrammar is able to represent and generate all valid polyurethane structures. We also present an algorithm to translate any polyurethane structure from the popular SMILES string format into our PolyGrammar representation. We test the representative power of PolyGrammar by translating a dataset of over 600 polyurethane samples collected from literature. Furthermore, we show that PolyGrammar can be easily extended to the other copolymers and homopolymers such as polyacrylates. By offering a complete, explicit representation scheme and an explainable generative model with validity guarantees, our PolyGrammar takes an important step toward a more comprehensive and practical system for polymer discovery and exploration. As the first bridge between formal languages and chemistry, PolyGrammar also serves as a critical blueprint to inform the design of similar grammars for other chemistries, including organic and inorganic molecules.
Adder neural network (AdderNet) is a new kind of deep model that replaces the original massive multiplications in convolutions by additions while preserving the high performance. Since the hardware complexity of additions is much lower than that of multiplications, the overall energy consumption is thus reduced significantly. To further optimize the hardware overhead of using AdderNet, this paper studies the winograd algorithm, which is a widely used fast algorithm for accelerating convolution and saving the computational costs. Unfortunately, the conventional Winograd algorithm cannot be directly applied to AdderNets since the distributive law in multiplication is not valid for the l1-norm. Therefore, we replace the element-wise multiplication in the Winograd equation by additions and then develop a new set of transform matrixes that can enhance the representation ability of output features to maintain the performance. Moreover, we propose the l2-to-l1 training strategy to mitigate the negative impacts caused by formal inconsistency. Experimental results on both FPGA and benchmarks show that the new method can further reduce the energy consumption without affecting the accuracy of the original AdderNet.
To generate drug molecules of desired properties with computational methods is the holy grail in pharmaceutical research. Here we describe an AI strategy, retro drug design, or RDD, to generate novel small molecule drugs from scratch to meet predefined requirements, including but not limited to biological activity against a drug target, and optimal range of physicochemical and ADMET properties. Traditional predictive models were first trained over experimental data for the target properties, using an atom typing based molecular descriptor system, ATP. Monte Carlo sampling algorithm was then utilized to find the solutions in the ATP space defined by the target properties, and the deep learning model of Seq2Seq was employed to decode molecular structures from the solutions. To test feasibility of the algorithm, we challenged RDD to generate novel drugs that can activate {\mu} opioid receptor (MOR) and penetrate blood brain barrier (BBB). Starting from vectors of random numbers, RDD generated 180,000 chemical structures, of which 78% were chemically valid. About 42,000 (31%) of the valid structures fell into the property space defined by MOR activity and BBB permeability. Out of the 42,000 structures, only 267 chemicals were commercially available, indicating a high extent of novelty of the AI-generated compounds. We purchased and assayed 96 compounds, and 25 of which were found to be MOR agonists. These compounds also have excellent BBB scores. The results presented in this paper illustrate that RDD has potential to revolutionize the current drug discovery process and create novel structures with multiple desired properties, including biological functions and ADMET properties. Availability of an AI-enabled fast track in drug discovery is essential to cope with emergent public health threat, such as pandemic of COVID-19.
Sensor and actuator selection problems (SASP) are some of the core problems in dynamic systems design and control. These problems correspond to determining the optimal selection of sensors (measurements) or actuators (control nodes) such that certain estimation/control objectives can be achieved. While the literature on SASP is indeed inveterate, the vast majority of the work focuses on linear(ized) representation of the network dynamics, resulting in the placements of sensors or actuators (SA) that are valid for confined operating regions. As an alternative, herein we propose a new general framework for addressing SASP in nonlinear dynamic systems (NDS), assuming that the inputs and outputs are linearly coupled with the nonlinear dynamics. This is investigated through (i) classifying and parameterizing the NDS into various nonlinear function sets, (ii) utilizing rich Lyapunov theoretic formulations, and (iii) designing a new customized branch-and-bound (BnB) algorithm that exploits problem structure of the SASP. The newly designed BnB routines are computationally more attractive than the standard one and also directly applicable to solve SASP for linear systems. In contrast with contemporary approaches from the literature, our approach is suitable for finding the optimal SA combination for stable/unstable NDS that ensures stabilization of estimation error and closed-loop dynamics through a simple linear feedback control policy.