Sequence models are a critical component of modern NLP systems, but their predictions are difficult to explain. We consider model explanations though rationales, subsets of context that can explain individual model predictions. We find sequential rationales by solving a combinatorial optimization: the best rationale is the smallest subset of input tokens that would predict the same output as the full sequence. Enumerating all subsets is intractable, so we propose an efficient greedy algorithm to approximate this objective. The algorithm, which is called greedy rationalization, applies to any model. For this approach to be effective, the model should form compatible conditional distributions when making predictions on incomplete subsets of the context. This condition can be enforced with a short fine-tuning step. We study greedy rationalization on language modeling and machine translation. Compared to existing baselines, greedy rationalization is best at optimizing the combinatorial objective and provides the most faithful rationales. On a new dataset of annotated sequential rationales, greedy rationales are most similar to human rationales.
相關內容
Constrained tensor and matrix factorization models allow to extract interpretable patterns from multiway data. Therefore identifiability properties and efficient algorithms for constrained low-rank approximations are nowadays important research topics. This work deals with columns of factor matrices of a low-rank approximation being sparse in a known and possibly overcomplete basis, a model coined as Dictionary-based Low-Rank Approximation (DLRA). While earlier contributions focused on finding factor columns inside a dictionary of candidate columns, i.e. one-sparse approximations, this work is the first to tackle DLRA with sparsity larger than one. I propose to focus on the sparse-coding subproblem coined Mixed Sparse-Coding (MSC) that emerges when solving DLRA with an alternating optimization strategy. Several algorithms based on sparse-coding heuristics (greedy methods, convex relaxations) are provided to solve MSC. The performance of these heuristics is evaluated on simulated data. Then, I show how to adapt an efficient MSC solver based on the LASSO to compute Dictionary-based Matrix Factorization and Canonical Polyadic Decomposition in the context of hyperspectral image processing and chemometrics. These experiments suggest that DLRA extends the modeling capabilities of low-rank approximations, helps reducing estimation variance and enhances the identifiability and interpretability of estimated factors.
In this paper, we aim at providing an effective Pairwise Learning Neural Link Prediction (PLNLP) framework. The framework treats link prediction as a pairwise learning to rank problem and consists of four main components, i.e., neighborhood encoder, link predictor, negative sampler and objective function. The framework is flexible that any generic graph neural convolution or link prediction specific neural architecture could be employed as neighborhood encoder. For link predictor, we design different scoring functions, which could be selected based on different types of graphs. In negative sampler, we provide several sampling strategies, which are problem specific. As for objective function, we propose to use an effective ranking loss, which approximately maximizes the standard ranking metric AUC. We evaluate the proposed PLNLP framework on 4 link property prediction datasets of Open Graph Benchmark, including ogbl-ddi, ogbl-collab, ogbl-ppa and ogbl-ciation2. PLNLP achieves top 1 performance on ogbl-ddi and ogbl-collab, and top 2 performance on ogbl-ciation2 only with basic neural architecture. The performance demonstrates the effectiveness of PLNLP.
Prediction over tabular data is an essential task in many data science applications such as recommender systems, online advertising, medical treatment, etc. Tabular data is structured into rows and columns, with each row as a data sample and each column as a feature attribute. Both the columns and rows of the tabular data carry useful patterns that could improve the model prediction performance. However, most existing models focus on the cross-column patterns yet overlook the cross-row patterns as they deal with single samples independently. In this work, we propose a general learning framework named Retrieval & Interaction Machine (RIM) that fully exploits both cross-row and cross-column patterns among tabular data. Specifically, RIM first leverages search engine techniques to efficiently retrieve useful rows of the table to assist the label prediction of the target row, then uses feature interaction networks to capture the cross-column patterns among the target row and the retrieved rows so as to make the final label prediction. We conduct extensive experiments on 11 datasets of three important tasks, i.e., CTR prediction (classification), top-n recommendation (ranking) and rating prediction (regression). Experimental results show that RIM achieves significant improvements over the state-of-the-art and various baselines, demonstrating the superiority and efficacy of RIM.
When we humans look at a video of human-object interaction, we can not only infer what is happening but we can even extract actionable information and imitate those interactions. On the other hand, current recognition or geometric approaches lack the physicality of action representation. In this paper, we take a step towards a more physical understanding of actions. We address the problem of inferring contact points and the physical forces from videos of humans interacting with objects. One of the main challenges in tackling this problem is obtaining ground-truth labels for forces. We sidestep this problem by instead using a physics simulator for supervision. Specifically, we use a simulator to predict effects and enforce that estimated forces must lead to the same effect as depicted in the video. Our quantitative and qualitative results show that (a) we can predict meaningful forces from videos whose effects lead to accurate imitation of the motions observed, (b) by jointly optimizing for contact point and force prediction, we can improve the performance on both tasks in comparison to independent training, and (c) we can learn a representation from this model that generalizes to novel objects using few shot examples.
RNN models have achieved the state-of-the-art performance in a wide range of text mining tasks. However, these models are often regarded as black-boxes and are criticized due to the lack of interpretability. In this paper, we enhance the interpretability of RNNs by providing interpretable rationales for RNN predictions. Nevertheless, interpreting RNNs is a challenging problem. Firstly, unlike existing methods that rely on local approximation, we aim to provide rationales that are more faithful to the decision making process of RNN models. Secondly, a flexible interpretation method should be able to assign contribution scores to text segments of varying lengths, instead of only to individual words. To tackle these challenges, we propose a novel attribution method, called REAT, to provide interpretations to RNN predictions. REAT decomposes the final prediction of a RNN into additive contribution of each word in the input text. This additive decomposition enables REAT to further obtain phrase-level attribution scores. In addition, REAT is generally applicable to various RNN architectures, including GRU, LSTM and their bidirectional versions. Experimental results demonstrate the faithfulness and interpretability of the proposed attribution method. Comprehensive analysis shows that our attribution method could unveil the useful linguistic knowledge captured by RNNs. Some analysis further demonstrates our method could be utilized as a debugging tool to examine the vulnerability and failure reasons of RNNs, which may lead to several promising future directions to promote generalization ability of RNNs.
We present a new approach for pretraining a bi-directional transformer model that provides significant performance gains across a variety of language understanding problems. Our model solves a cloze-style word reconstruction task, where each word is ablated and must be predicted given the rest of the text. Experiments demonstrate large performance gains on GLUE and new state of the art results on NER as well as constituency parsing benchmarks, consistent with the concurrently introduced BERT model. We also present a detailed analysis of a number of factors that contribute to effective pretraining, including data domain and size, model capacity, and variations on the cloze objective.
Knowledge graph embedding aims to learn distributed representations for entities and relations, and is proven to be effective in many applications. Crossover interactions --- bi-directional effects between entities and relations --- help select related information when predicting a new triple, but haven't been formally discussed before. In this paper, we propose CrossE, a novel knowledge graph embedding which explicitly simulates crossover interactions. It not only learns one general embedding for each entity and relation as most previous methods do, but also generates multiple triple specific embeddings for both of them, named interaction embeddings. We evaluate embeddings on typical link prediction tasks and find that CrossE achieves state-of-the-art results on complex and more challenging datasets. Furthermore, we evaluate embeddings from a new perspective --- giving explanations for predicted triples, which is important for real applications. In this work, an explanation for a triple is regarded as a reliable closed-path between the head and the tail entity. Compared to other baselines, we show experimentally that CrossE, benefiting from interaction embeddings, is more capable of generating reliable explanations to support its predictions.
There has been much recent work on training neural attention models at the sequence-level using either reinforcement learning-style methods or by optimizing the beam. In this paper, we survey a range of classical objective functions that have been widely used to train linear models for structured prediction and apply them to neural sequence to sequence models. Our experiments show that these losses can perform surprisingly well by slightly outperforming beam search optimization in a like for like setup. We also report new state of the art results on both IWSLT'14 German-English translation as well as Gigaword abstractive summarization. On the larger WMT'14 English-French translation task, sequence-level training achieves 41.5 BLEU which is on par with the state of the art.
Traditional methods for link prediction can be categorized into three main types: graph structure feature-based, latent feature-based, and explicit feature-based. Graph structure feature methods leverage some handcrafted node proximity scores, e.g., common neighbors, to estimate the likelihood of links. Latent feature methods rely on factorizing networks' matrix representations to learn an embedding for each node. Explicit feature methods train a machine learning model on two nodes' explicit attributes. Each of the three types of methods has its unique merits. In this paper, we propose SEAL (learning from Subgraphs, Embeddings, and Attributes for Link prediction), a new framework for link prediction which combines the power of all the three types into a single graph neural network (GNN). GNN is a new type of neural network which directly accepts graphs as input and outputs their labels. In SEAL, the input to the GNN is a local subgraph around each target link. We prove theoretically that our local subgraphs also reserve a great deal of high-order graph structure features related to link existence. Another key feature is that our GNN can naturally incorporate latent features and explicit features. It is achieved by concatenating node embeddings (latent features) and node attributes (explicit features) in the node information matrix for each subgraph, thus combining the three types of features to enhance GNN learning. Through extensive experiments, SEAL shows unprecedentedly strong performance against a wide range of baseline methods, including various link prediction heuristics and network embedding methods.
Networks provide a powerful formalism for modeling complex systems, by representing the underlying set of pairwise interactions. But much of the structure within these systems involves interactions that take place among more than two nodes at once; for example, communication within a group rather than person-to-person, collaboration among a team rather than a pair of co-authors, or biological interaction between a set of molecules rather than just two. We refer to these type of simultaneous interactions on sets of more than two nodes as higher-order interactions; they are ubiquitous, but the empirical study of them has lacked a general framework for evaluating higher-order models. Here we introduce such a framework, based on link prediction, a fundamental problem in network analysis. The traditional link prediction problem seeks to predict the appearance of new links in a network, and here we adapt it to predict which (larger) sets of elements will have future interactions. We study the temporal evolution of 19 datasets from a variety of domains, and use our higher-order formulation of link prediction to assess the types of structural features that are most predictive of new multi-way interactions. Among our results, we find that different domains vary considerably in their distribution of higher-order structural parameters, and that the higher-order link prediction problem exhibits some fundamental differences from traditional pairwise link prediction, with a greater role for local rather than long-range information in predicting the appearance of new interactions.
小貼士
登錄享
注冊地址： 北京市海淀區羊坊店路18號2幢3層301-191