United States

Large-scale cross-lingual language models (LM), such as mBERT, Unicoder and XLM, have achieved great success in cross-lingual representation learning. However, when applied to zero-shot cross-lingual transfer tasks, most existing methods use only single-language input for LM finetuning, without leveraging the intrinsic cross-lingual alignment between different languages that proves essential for multilingual tasks. In this paper, we propose FILTER, an enhanced fusion method that takes cross-lingual data as input for XLM finetuning. Specifically, FILTER first encodes text input in the source language and its translation in the target language independently in the shallow layers, then performs cross-language fusion to extract multilingual knowledge in the intermediate layers, and finally performs further language-specific encoding. During inference, the model makes predictions based on the text input in the target language and its translation in the source language. For simple tasks such as classification, translated text in the target language shares the same label as the source language. However, this shared label becomes less accurate or even unavailable for more complex tasks such as question answering, NER and POS tagging. To tackle this issue, we further propose an additional KL-divergence self-teaching loss for model training, based on auto-generated soft pseudo-labels for translated text in the target language.  Extensive experiments demonstrate that FILTER achieves new state of the art on two challenging multilingual multi-task benchmarks, XTREME and XGLUE.

AAAI 2021

FILTER: An Enhanced Fusion Method for Cross-Lingual Language Understanding

technical paper

The purpose of the AAAI conference is to promote research in artificial intelligence (AI) and scientific exchange among AI researchers, practitioners, scientists, and engineers in affiliated disciplines. 

Most supervised learning models are trained for full automation. However, their predictions are sometimes worse than those by human experts on some specific instances. Motivated by this empirical observation, our goal is to design classifiers that are optimized to operate under different automation levels. More specifically, we focus on convex margin-based classifiers and first show that the problem is NP-hard. Then, we further show that, for support vector machines, the corresponding objective function can be expressed as the difference of two functions f = g - c, where g is monotone, non-negative and gamma-weakly submodular, and c is non-negative and modular. This representation allows a recently introduced deterministic greedy algorithm, as well as a more efficient randomized variant of the algorithm, to enjoy approximation guarantees at solving the problem. Experiments on synthetic and real-world data from several applications in medical diagnosis illustrate our theoretical findings and demonstrate that, under human assistance, supervised learning models trained to operate under different automation levels can outperform those trained for full automation as well as humans operating alone.

Classification under Human Assistance

We consider the Multi-Armed Bandit (MAB) problem, where an agent sequentially chooses actions and observes rewards for the actions it took. While the majority of algorithms try to minimize the regret, i.e., the cumulative difference between the reward of the best action and the agent's action, this criterion might lead to undesirable results. For example, in large problems, or when the interaction with the environment is brief, finding an optimal arm is infeasible, and regret-minimizing algorithms tend to over-explore. To overcome this issue, algorithms for such settings should instead focus on playing near-optimal arms. To this end, we suggest a new, more lenient, regret criterion that ignores suboptimality gaps smaller than some Îµ. We then present a variant of the Thompson Sampling (TS) algorithm, called Îµ-TS, and prove its asymptotic optimality in terms of the lenient regret. Importantly, we show that when the mean of the optimal arm is high enough, the lenient regret of Îµ-TS is bounded by a constant. Finally, we show that Îµ-TS can be applied to improve the performance when the agent knows a lower bound of the suboptimality gaps.

Lenient Regret for Multi-Armed Bandits

Tensor decompositions have found many applications in signal processing, data mining, machine learning, etc. 
In particular, the block term decomposition (BTD), which is a generalization of CP decomposition and Tucker decomposition/HOSVD, 
has been successfully used for the compression and acceleration of neural networks. 
However, computing BTD is NP-hard, and optimization based methods usually suffer from slow convergence or even fail to converge, which limits the applications of BTD. 
This paper considers a â€œblindâ€ block term decomposition (BBTD) of high order tensors, in which the block diagonal structure of the core tensor is unknown. 
Our contributions include: 
1) We establish the necessary and sufficient conditions for the existence of BTD, characterize the condition when a BTD solves the BBTD problem, and show that the BBTD is unique under a â€œlow rankâ€ assumption. 
2) We propose an algebraic method to compute the BBTD. 
This method transforms the problem of determining the block diagonal structure of the core tensor into a clustering problem of complex numbers, in polynomial time. 
And once the clustering problem is solved, the BBTD can be obtained via computing several matrix decompositions.
Numerical results show that our method is able to compute the BBTD, even in the presence of noise to some extent, 
whereas optimization based methods (e.g., MINF and NLS in TENSORLAB) may fail to converge.

A Blind Block Term Decomposition of High Order Tensors

Pretrained language models have recently advanced a wide range of natural language processing tasks. Nowadays, the application of pretrained language models to IR tasks has also achieved impressive results. Typical methods either directly apply a pretrained model to improve the re-ranking stage, or use it to conduct passage expansion and term weighting for first-stage retrieval. We observe that the passage ranking and passage expansion tasks share certain inherent relations, and can benefit from each other. Therefore, in this paper, we propose a general pretraining framework to enhance both tasks with Unified Encoder-Decoder networks (UED). The overall ranking framework consists of two parts in a cascade manner: (1) passage expansion with a pretraining-based query generation method; (2) re-ranking of passage candidates from a traditional retrieval method with a pretrained transformer encoder. Both the two parts are based on the same pretrained UED model, where we jointly train the passage ranking and query generation tasks for further improving the full ranking pipeline. An extensive set of experiments have been conducted on two large-scale passage retrieval datasets to demonstrate the state-of-the-art results of the proposed framework in both the first-stage retrieval and the final re-ranking. In addition, we successfully deploy the framework to our online production system, which can stably serve industrial applications with a request volume of up to 100 QPS in less than 300ms.

A Unified Pretraining Framework for Passage Ranking and Expansion

As a scene graph compactly summarizes the high-level content of an image in a structured and symbolic manner, the similarity between scene graphs of two images reflects the relevance of their contents. Based on this idea, we propose a novel approach for image-to-image retrieval using scene graph similarity measured by graph neural networks. In our approach, graph neural networks are trained to predict the proxy image relevance measure, computed from human-annotated captions using a pre-trained sentence similarity model. We collect and publish the dataset for image relevance measured by human annotators to evaluate retrieval algorithms. The collected dataset shows that our method agrees well with the human perception of image similarity than other competitive baselines.

Image-to-Image Retrieval by Learning Similarity between Scene Graphs

Blackwell approachability is a framework for reasoning about repeated games with vector-valued payoffs. We introduce predictive Blackwell approachability, where an estimate of the next payoff vector is given, and the decision maker tries to achieve better performance based on the accuracy of that estimator. In order to derive algorithms that achieve predictive Blackwell approachability, we start by showing a powerful connection between four well-known algorithms. Follow-the-regularized-leader (FTRL) and online mirror descent (OMD) are the most prevalent regret minimizers in online convex optimization. In spite of this prevalence, the regret matching (RM) and regret matching+ (RM+) algorithms have been preferred in the practice of solving large-scale games (as the local regret minimizers within the counterfactual regret minimization framework). We show that RM and RM+ are the algorithms that result from running FTRL and OMD, respectively, to select the halfspace to force at all times in the underlying Blackwell approachability game. By applying the predictive variants of FTRL or OMD to this connection, we obtain predictive Blackwell approachability algorithms, as well as predictive variants of RM and RM+. In experiments across 18 common zero-sum extensive-form benchmark games, we show that predictive RM+ coupled with counterfactual regret minimization converges vastly faster than the fastest prior algorithms (CFR+, DCFR, LCFR) across all games but two of the poker games, sometimes by two or more orders of magnitude.

Faster Game Solving via Predictive Blackwell Approachability: Connecting Regret Matching and Mirror Descent

Neural network visualization techniques mark image locations by their relevancy to the network's classification. Existing methods are effective in highlighting the regions that affect the resulting classification the most. However, as we show, these methods are limited in their ability to identify the support for alternative classifications, an effect we name the saliency bias hypothesis. In this work, we integrate two lines of research: gradient-based methods and attribution-based methods, and develop an algorithm that provides per-class explainability.
The algorithm back-projects the per pixel local influence, in a manner that is guided by the local attributions, while correcting for salient features that would otherwise bias the explanation.
In an extensive battery of experiments, we demonstrate the ability of our methods to class-specific visualization, and not just the predicted label. Remarkably, the method obtains state of the art results in benchmarks that are commonly applied to gradient-based methods as well as in those that are employed mostly for evaluating attribution methods. Using a new unsupervised procedure, our method is also successful in demonstrating that self-supervised methods learn semantic information. Our code is available at: https://github.com/shirgur/AGFVisualization.

Visualization of Supervised and Self-Supervised Neural Networks via Attribution Guided Factorization

The estimation and inference of human predictive uncertainty have great potential to improve the sampling efficiency and prediction reliability of human-in-the-loop systems for smart healthcare, smart education, and human-computer interactions. Predictive uncertainty in humans is highly interpretable, but its measurement is poorly accessible. Contrarily, the predictive uncertainty of machine learning models, albeit with poor interpretability, is relatively easily accessible. Here, we demonstrate that the poor accessibility of human uncertainty can be resolved by exploiting simple and universally accessible deterministic neural networks. We propose a new model for human uncertainty inference, called proxy ensemble network (PEN). Simulations with a few benchmark datasets demonstrated that the model can efficiently learn human uncertainty from a small amount of data. To show its applicability in real-world problems, we performed behavioral experiments, in which 64 physicians classified medical images and reported their level of confidence. We showed that the PEN could predict both the uncertainty range and diagnoses given by subjects with high accuracy. Our results demonstrate the ability of machine learning in guiding human decision making; it can also help humans in learning more efficiently and accurately. To the best of our knowledge, this is the first study that explored the possibility of accessing human uncertainty via the lens of deterministic neural networks.

Human Uncertainty Inference via Deterministic Ensemble Neural Networks

Efficiently computing the weighted Jaccard similarity has become an active research topic in machine learning and theory. For sparse data, the standard technique is based on the consistent weighed sampling (CWS). For dense data, however, methods based on rejection sampling (RS) can be much more efficient. Nevertheless, existing RS methods are still slow for practical purposes. In this paper, we propose to improve RS by  a strategy, which we call  efficient rejection sampling (ERS), based on ``early stopping + densification''.  We analyze the statistical property of  ERS and provide experimental results to compare ERS with RS and other algorithms for hashing weighted Jaccard. The results demonstrate that ERS significantly improves the existing methods for estimating the weighted Jaccard similarity in relatively dense data.

Rejection Sampling for Weighted Jaccard Similarity Revisited

Asynchronous distributed algorithms are a popular way to reduce synchronization costs in large-scale optimization, and in particular for neural network training. However, for nonsmooth and nonconvex objectives, few convergence guarantees exist beyond cases where closed-form proximal operator solutions are available. 
As training most popular deep neural networks corresponds to optimizing nonsmooth and nonconvex objectives, there is a pressing need for such convergence guarantees. In this paper, we analyze for the first time the convergence of stochastic asynchronous optimization for this general class of objectives. In particular, we focus on stochastic subgradient methods allowing for block variable partitioning, where the shared model is asynchronously updated by concurrent processes. To this end, we use a probabilistic model which captures key features of real asynchronous scheduling between concurrent processes. Under this model, we establish convergence with probability one to an invariant set for stochastic subgradient methods with momentum. 
From a practical perspective, one issue with the family of algorithms that we consider is that they are not efficiently supported by machine learning frameworks, which mostly focus on distributed data-parallel strategies. 
To address this, we propose a new implementation strategy for  shared-memory based training of deep neural networks for a partitioned but shared model in single- and multi-GPU settings.  
Based on this implementation, we achieve on average1.2x speed-up in comparison to state-of-the-art training methods for popular image classification tasks, without compromising accuracy.

FILTER: An Enhanced Fusion Method for Cross-Lingual Language Understanding

Downloads

Next from AAAI 2021

Classification under Human Assistance

Stay up to date with the latest Underline news!

PRESENTATIONS

CONFERENCES

COMPANY

RESOURCES

FILTER: An Enhanced Fusion Method for Cross-Lingual Language Understanding

.css-70qvj9{display:-webkit-box;display:-webkit-flex;display:-ms-flexbox;display:flex;-webkit-align-items:center;-webkit-box-align:center;-ms-flex-align:center;align-items:center;}Downloads

Next from AAAI 2021

Classification under Human Assistance

Stay up to date with the latest Underline news!

PRESENTATIONS

CONFERENCES

COMPANY

RESOURCES

Downloads