To classify in-distribution samples, deep neural networks explore strongly label-related information and discard weakly label-related information according to the information bottleneck. Out-of-distribution samples drawn from distributions differing from that of in-distribution samples could be assigned with unexpected high-confidence predictions because they could obtain minimum strongly label-related information. To distinguish in- and out-of-distribution samples, Dual Representation Learning (DRL) makes out-of-distribution samples harder to have high-confidence predictions by exploring both strongly and weakly label-related information from in-distribution samples. For a pretrained network exploring strongly label-related information to learn label-discriminative representations, DRL trains its auxiliary network exploring the remaining weakly label-related information to learn distribution-discriminative representations. Specifically, for a label-discriminative representation, DRL constructs its complementary distribution-discriminative representation by integrating diverse representations less similar to the label-discriminative representation. Accordingly, DRL combines label- and distribution-discriminative representations to detect out-of-distribution samples. Experiments show that DRL outperforms the state-of-the-art methods for out-of-distribution detection.
The success of deep active learning hinges on the choice of an effective acquisition function, which ranks not yet labeled data points according to their expected informativeness. Many acquisition functions are (partly) based on the uncertainty that the current model has about the class label of a point, yet there is no generally agreed upon strategy for computing such uncertainty. This paper proposes a new and very simple approach to computing uncertainty in deep active learning with a Convolutional Neural Network (CNN). The main idea is to use the feature representation extracted by the CNN as data for training a Sum-Product Network (SPN). Since SPNs are typically used for estimating the distribution of a dataset, they are well suited to the task of estimating class probabilities that can be used directly by standard acquisition functions such as max entropy and variational ratio. The effectiveness of our method is demonstrated in an experimental study on several standard benchmark datasets for image classification, where we compare it to various state-of-the-art methods for assessing uncertainty in deep active learning.
Reasoning with preconditions such as "glass can be used for drinking water unless the glass is shattered" remains an open problem for language models. The main challenge lies in the scarcity of preconditions data and the model's lack of support for such reasoning. We present PInKS, Preconditioned Commonsense Inference with WeaK Supervision, an improved model for reasoning with preconditions through minimum supervision. We show, both empirically and theoretically, that PInKS improves the results on benchmarks focused on reasoning with the preconditions of commonsense knowledge (up to 40% Macro-F1 scores). We further investigate PInKS through PAC-Bayesian informativeness analysis, precision measures, and ablation study.
Learning vectorized embeddings is at the core of various recommender systems for user-item matching. To perform efficient online inference, representation quantization, aiming to embed the latent features by a compact sequence of discrete numbers, recently shows the promising potentiality in optimizing both memory and computation overheads. However, existing work merely focuses on numerical quantization whilst ignoring the concomitant information loss issue, which, consequently, leads to conspicuous performance degradation. In this paper, we propose a novel quantization framework to learn Binarized Graph Representations for Top-K Recommendation (BiGeaR). BiGeaR introduces multi-faceted quantization reinforcement at the pre-, mid-, and post-stage of binarized representation learning, which substantially retains the representation informativeness against embedding binarization. In addition to saving the memory footprint, BiGeaR further develops solid online inference acceleration with bitwise operations, providing alternative flexibility for the realistic deployment. The empirical results over five large real-world benchmarks show that BiGeaR achieves about 22%~40% performance improvement over the state-of-the-art quantization-based recommender system, and recovers about 95%~102% of the performance capability of the best full-precision counterpart with over 8x time and space reduction.
We propose a new paradigm for encoding the geometry of point cloud sequences, where the convolutional neural network (CNN) which estimates the encoding distributions is optimized on several frames of the sequence to be compressed. We adopt lightweight CNN structures, we perform training as part of the encoding process, and the CNN parameters are transmitted as part of the bitstream. The newly proposed encoding scheme operates on the octree representation for each point cloud, encoding consecutively each octree resolution layer. At every octree resolution layer, the voxel grid is traversed section-by-section (each section being perpendicular to a selected coordinate axis) and in each section the occupancies of groups of two-by-two voxels are encoded at once, in a single arithmetic coding operation. A context for the conditional encoding distribution is defined for each two-by-two group of voxels, based on the information available about the occupancy of neighbor voxels in the current and lower resolution layers of the octree. The CNN estimates the probability distributions of occupancy patterns of all voxel groups from one section in four phases. In each new phase the contexts are updated with the occupancies encoded in the previous phase, and each phase estimates the probabilities in parallel, providing a reasonable trade-off between the parallelism of processing and the informativeness of the contexts. The CNN training time is comparable to the time spent in the remaining encoding steps, leading to competitive overall encoding times. Bitrates and encoding-decoding times compare favorably with those of recently published compression schemes.
Extractive summaries are usually presented as lists of sentences with no expected cohesion between them and with plenty of redundant information if not accounted for. In this paper, we investigate the trade-offs incurred when aiming to control for inter-sentential cohesion and redundancy in extracted summaries, and their impact on their informativeness. As case study, we focus on the summarization of long, highly redundant documents and consider two optimization scenarios, reward-guided and with no supervision. In the reward-guided scenario, we compare systems that control for redundancy and cohesion during sentence scoring. In the unsupervised scenario, we introduce two systems that aim to control all three properties -- informativeness, redundancy, and cohesion -- in a principled way. Both systems implement a psycholinguistic theory that simulates how humans keep track of relevant content units and how cohesion and non-redundancy constraints are applied in short-term memory during reading. Extensive automatic and human evaluations reveal that systems optimizing for -- among other properties -- cohesion are capable of better organizing content in summaries compared to systems that optimize only for redundancy, while maintaining comparable informativeness. We find that the proposed unsupervised systems manage to extract highly cohesive summaries across varying levels of document redundancy, although sacrificing informativeness in the process. Finally, we lay evidence as to how simulated cognitive processes impact the trade-off between the analyzed summary properties.
Graphical structures estimated by causal learning algorithms from time series data can provide misleading causal information if the causal timescale of the generating process fails to match the measurement timescale of the data. Existing algorithms provide limited resources to respond to this challenge, and so researchers must either use models that they know are likely misleading, or else forego causal learning entirely. Existing methods face up-to-four distinct shortfalls, as they might 1) require that the difference between causal and measurement timescales is known; 2) only handle very small number of random variables when the timescale difference is unknown; 3) only apply to pairs of variables; or 4) be unable to find a solution given statistical noise in the data. This research addresses these challenges. Our approach combines constraint programming with both theoretical insights into the problem structure and prior information about admissible causal interactions to achieve multiple orders of magnitude in speed-up. The resulting system maintains theoretical guarantees while scaling to significantly larger sets of random variables (>100) without knowledge of timescale differences. This method is also robust to edge misidentification and can use parametric connection strengths, while optionally finding the optimal solution among many possible ones.
We study the link between Phelps-Aigner-Cain-type statistical discrimination and familiar notions of statistical informativeness. Our central insight is that Blackwell's Theorem, suitably relabeled, characterizes statistical discrimination in terms of statistical informativeness. This delivers one-half of Chambers and Echenique's (2021) characterization of statistical discrimination as a corollary, and suggests a different interpretation of it: that discrimination is inevitable. In addition, Blackwell's Theorem delivers a number of finer-grained insights into the nature of statistical discrimination. We argue that the discrimination-informativeness link is quite general, illustrating with an informativeness characterization of a different type of discrimination.
In mathematics information is a number that measures uncertainty (entropy) based on a probabilistic distribution, often of an obscure origin. In real life language information is a datum, a statement, more precisely, a formula. But such a formula should be justified by a proof. I try to formalize this perception of information. The measure of informativeness of a proof is based on the set of proofs related to the formulas under consideration. This set of possible proofs (`a knowledge base') defines a probabilistic measure, and entropic weight is defined using this measure. The paper is mainly conceptual, it is not clear where and how this approach can be applied.
Abstractive multi document summarization has evolved as a task through the basic sequence to sequence approaches to transformer and graph based techniques. Each of these approaches has primarily focused on the issues of multi document information synthesis and attention based approaches to extract salient information. A challenge that arises with multi document summarization which is not prevalent in single document summarization is the need to effectively summarize multiple documents that might have conflicting polarity, sentiment or subjective information about a given topic. In this paper we propose ACM, attribute conditioned multi document summarization,a model that incorporates attribute conditioning modules in order to decouple conflicting information by conditioning for a certain attribute in the output summary. This approach shows strong gains in ROUGE score over baseline multi document summarization approaches and shows gains in fluency, informativeness and reduction in repetitiveness as shown through a human annotation analysis study.
Open-domain multi-turn conversations normally face the challenges of how to enrich and expand the content of the conversation. Recently, many approaches based on external knowledge are proposed to generate rich semantic and information conversation. Two types of knowledge have been studied for knowledge-aware open-domain dialogue generation: structured triples from knowledge graphs and unstructured texts from documents. To take both advantages of abundant unstructured latent knowledge in the documents and the information expansion capabilities of the structured knowledge graph, this paper presents a new dialogue generation model, Dynamic Multi-form Knowledge Fusion based Open-domain Chatt-ing Machine (DMKCM).In particular, DMKCM applies an indexed text (a virtual Knowledge Base) to locate relevant documents as 1st hop and then expands the content of the dialogue and its 1st hop using a commonsense knowledge graph to get apposite triples as 2nd hop. To merge these two forms of knowledge into the dialogue effectively, we design a dynamic virtual knowledge selector and a controller that help to enrich and expand knowledge space. Moreover, DMKCM adopts a novel dynamic knowledge memory module that effectively uses historical reasoning knowledge to generate better responses. Experimental results indicate the effectiveness of our method in terms of dialogue coherence and informativeness.
Knowledge-aware methods have boosted a range of natural language processing applications over the last decades. With the gathered momentum, knowledge recently has been pumped into enormous attention in document summarization, one of natural language processing applications. Previous works reported that knowledge-embedded document summarizers excel at generating superior digests, especially in terms of informativeness, coherence, and fact consistency. This paper pursues to present the first systematic survey for the state-of-the-art methodologies that embed knowledge into document summarizers. Particularly, we propose novel taxonomies to recapitulate knowledge and knowledge embeddings under the document summarization view. We further explore how embeddings are generated in embedding learning architectures of document summarization models, especially of deep learning models. At last, we discuss the challenges of this topic and future directions.
Both humans and machines learn the meaning of unknown words through contextual information in a sentence, but not all contexts are equally helpful for learning. We introduce an effective method for capturing the level of contextual informativeness with respect to a given target word. Our study makes three main contributions. First, we develop models for estimating contextual informativeness, focusing on the instructional aspect of sentences. Our attention-based approach using pre-trained embeddings demonstrates state-of-the-art performance on our single-context dataset and an existing multi-sentence context dataset. Second, we show how our model identifies key contextual elements in a sentence that are likely to contribute most to a reader's understanding of the target word. Third, we examine how our contextual informativeness model, originally developed for vocabulary learning applications for students, can be used for developing better training curricula for word embedding models in batch learning and few-shot machine learning settings. We believe our results open new possibilities for applications that support language learning for both human and machine learners.
As an important and challenging problem in vision-language tasks, referring expression comprehension (REC) generally requires a large amount of multi-grained information of visual and linguistic modalities to realize accurate reasoning. In addition, due to the diversity of visual scenes and the variation of linguistic expressions, some hard examples have much more abundant multi-grained information than others. How to aggregate multi-grained information from different modalities and extract abundant knowledge from hard examples is crucial in the REC task. To address aforementioned challenges, in this paper, we propose a Self-paced Multi-grained Cross-modal Interaction Modeling framework, which improves the language-to-vision localization ability through innovations in network structure and learning mechanism. Concretely, we design a transformer-based multi-grained cross-modal attention, which effectively utilizes the inherent multi-grained information in visual and linguistic encoders. Furthermore, considering the large variance of samples, we propose a self-paced sample informativeness learning to adaptively enhance the network learning for samples containing abundant multi-grained information. The proposed framework significantly outperforms state-of-the-art methods on widely used datasets, such as RefCOCO, RefCOCO+, RefCOCOg, and ReferItGame datasets, demonstrating the effectiveness of our method.
Recent organizations have started to adopt AI-based decision support tools to optimize human resource development practices, while facing various challenges of using AIs in highly contextual and sensitive domains. We present our case study that aims to help professional assessors make decisions in human assessment, in which they conduct interviews with assessees and evaluate their suitability for certain job roles. Our workshop with two industrial assessors elucidated troubles they face (i.e., maintaining stable and non-subjective observation of assessees' behaviors) and derived requirements of AI systems (i.e., extracting their nonverbal cues from interview videos in an interpretable manner). In response, we employed an unsupervised anomaly detection algorithm using multimodal behavioral features such as facial keypoints, body and head pose, and gaze. The algorithm extracts outlier scenes from the video based on behavioral features as well as informing which feature contributes to the outlierness. We first evaluated how the assessors would perceive the extracted cues and discovered that the algorithm is useful in suggesting scenes to which assessors would pay attention, thanks to its interpretability. Then, we developed an interface prototype incorporating the algorithm and had six assessors use it for their actual assessment. Their comments revealed the effectiveness of introducing unsupervised anomaly detection to enhance their feeling of confidence and objectivity of the assessment along with potential use scenarios of such AI-based systems in human assessment. Our approach, which builds on top of the idea of separating observation and interpretation in human-AI collaboration, will facilitate human decision making in highly contextual domains, such as human assessment, while keeping their trust in the system.
How to sample high quality negative instances from unlabeled data, i.e., negative sampling, is important for training implicit collaborative filtering and contrastive learning models. Although previous studies have proposed some approaches to sample informative instances, few has been done to discriminating false negative from true negative for unbiased negative sampling. On the basis of our order relation analysis of negatives' scores, we first derive the class conditional density of true negatives and that of false negatives. We next design a Bayesian classifier for negative classification, from which we define a model-agnostic posterior probability estimate of an instance being true negative as a quantitative negative signal measure. We also propose a Bayesian optimal sampling rule to sample high-quality negatives. The proposed Bayesian Negative Sampling (BNS) algorithm has a linear time complexity. Experimental studies validate the superiority of BNS over the peers in terms of better sampling quality and better recommendation performance.
Spurious correlations are a threat to the trustworthiness of natural language processing systems, motivating research into methods for identifying and eliminating them. However, addressing the problem of spurious correlations requires more clarity on what they are and how they arise in language data. Gardner et al (2021) argue that the compositional nature of language implies that \emph{all} correlations between labels and individual "input features" are spurious. This paper analyzes this proposal in the context of a toy example, demonstrating three distinct conditions that can give rise to feature-label correlations in a simple PCFG. Linking the toy example to a structured causal model shows that (1) feature-label correlations can arise even when the label is invariant to interventions on the feature, and (2) feature-label correlations may be absent even when the label is sensitive to interventions on the feature. Because input features will be individually correlated with labels in all but very rare circumstances, domain knowledge must be applied to identify spurious correlations that pose genuine robustness threats.
Probing has become an important tool for analyzing representations in Natural Language Processing (NLP). For graphical NLP tasks such as dependency parsing, linear probes are currently limited to extracting undirected or unlabeled parse trees which do not capture the full task. This work introduces DepProbe, a linear probe which can extract labeled and directed dependency parse trees from embeddings while using fewer parameters and compute than prior methods. Leveraging its full task coverage and lightweight parametrization, we investigate its predictive power for selecting the best transfer language for training a full biaffine attention parser. Across 13 languages, our proposed method identifies the best source treebank 94% of the time, outperforming competitive baselines and prior work. Finally, we analyze the informativeness of task-specific subspaces in contextual embeddings as well as which benefits a full parser's non-linear parametrization provides.
A limitation of current neural dialog models is that they tend to suffer from a lack of specificity and informativeness in generated responses, primarily due to dependence on training data that covers a limited variety of scenarios and conveys limited knowledge. One way to alleviate this issue is to extract relevant knowledge from external sources at decoding time and incorporate it into the dialog response. In this paper, we propose a post-hoc knowledge-injection technique where we first retrieve a diverse set of relevant knowledge snippets conditioned on both the dialog history and an initial response from an existing dialog model. We construct multiple candidate responses, individually injecting each retrieved snippet into the initial response using a gradient-based decoding method, and then select the final response with an unsupervised ranking step. Our experiments in goal-oriented and knowledge-grounded dialog settings demonstrate that human annotators judge the outputs from the proposed method to be more engaging and informative compared to responses from prior dialog systems. We further show that knowledge-augmentation promotes success in achieving conversational goals in both experimental settings.
Learning representations of images that are invariant to sensitive or unwanted attributes is important for many tasks including bias removal and cross domain retrieval. Here, our objective is to learn representations that are invariant to the domain (sensitive attribute) for which labels are provided, while being informative over all other image attributes, which are unlabeled. We present a new approach, proposing a new domain-wise contrastive objective for ensuring invariant representations. This objective crucially restricts negative image pairs to be drawn from the same domain, which enforces domain invariance whereas the standard contrastive objective does not. This domain-wise objective is insufficient on its own as it suffers from shortcut solutions resulting in feature suppression. We overcome this issue by a combination of a reconstruction constraint, image augmentations and initialization with pre-trained weights. Our analysis shows that the choice of augmentations is important, and that a misguided choice of augmentations can harm the invariance and informativeness objectives. In an extensive evaluation, our method convincingly outperforms the state-of-the-art in terms of representation invariance, representation informativeness, and training speed. Furthermore, we find that in some cases our method can achieve excellent results even without the reconstruction constraint, leading to a much faster and resource efficient training.