Ziyi Yang

Yue Liang, Jiatong Du, Ziyi Yang et al.

Scene graphs provide structured abstractions for scene understanding, yet they often overfit to spurious correlations, severely hindering out-of-distribution generalization. To address this limitation, we propose CURVE, a causality-inspired framework that integrates variational uncertainty modeling with uncertainty-guided structural regularization to suppress high-variance, environment-specific relations. Specifically, we apply prototype-conditioned debiasing to disentangle invariant interaction dynamics from environment-dependent variations, promoting a sparse and domain-stable topology. Empirically, we evaluate CURVE in zero-shot transfer and low-data sim-to-real adaptation, verifying its ability to learn domain-stable sparse topologies and provide reliable uncertainty estimates to support risk prediction under distribution shifts.

Ziyi Yang, Yinfei Yang, Daniel Cer et al.

Language agnostic and semantic-language information isolation is an emerging research direction for multilingual representations models. We explore this problem from a novel angle of geometric algebra and semantic space. A simple but highly effective method "Language Information Removal (LIR)" factors out language identity information from semantic related components in multilingual representations pre-trained on multi-monolingual data. A post-training and model-agnostic method, LIR only uses simple linear operations, e.g. matrix factorization and orthogonal projection. LIR reveals that for weak-alignment multilingual systems, the principal components of semantic spaces primarily encodes language identity information. We first evaluate the LIR on a cross-lingual question answer retrieval task (LAReQA), which requires the strong alignment for the multilingual embedding space. Experiment shows that LIR is highly effectively on this task, yielding almost 100% relative improvement in MAP for weak-alignment models. We then evaluate the LIR on Amazon Reviews and XEVAL dataset, with the observation that removing language information is able to improve the cross-lingual transfer performance.

Qi Zhang, Yilin Chen, Ziyi Yang et al.

In this paper, we study the problem of large-strain consolidation in poromechanics with deep neural networks (DNN). Given different material properties and different loading conditions, the goal is to predict pore pressure and settlement. We propose a novel method "multi-constitutive neural network" (MCNN) such that one model can solve several different constitutive laws. We introduce a one-hot encoding vector as an additional input vector, which is used to label the constitutive law we wish to solve. Then we build a DNN which takes $(\hat{X}, \hat{t})$ as input along with a constitutive law label and outputs the corresponding solution. It is the first time, to our knowledge, that we can evaluate multi-constitutive laws through only one training process while still obtaining good accuracies. We found that MCNN trained to solve multiple PDEs outperforms individual neural network solvers trained with PDE in some cases.

Ziyi Yang, Iman Soltani Bozchalooi, Eric Darve

We study the problem of semi-supervised anomaly detection with domain adaptation. Given a set of normal data from a source domain and a limited amount of normal examples from a target domain, the goal is to have a well-performing anomaly detector in the target domain. We propose the Invariant Representation Anomaly Detection (IRAD) to solve this problem where we first learn to extract a domain-invariant representation. The extraction is achieved by an across-domain encoder trained together with source-specific encoders and generators by adversarial learning. An anomaly detector is then trained using the learnt representations. We evaluate IRAD extensively on digits images datasets (MNIST, USPS and SVHN) and object recognition datasets (Office-Home). Experimental results show that IRAD outperforms baseline models by a wide margin across different datasets. We derive a theoretical lower bound for the joint error that explains the performance decay from overtraining and also an upper bound for the generalization error.

Vin Sachidananda, Ziyi Yang, Chenguang Zhu

Due to widespread interest in machine translation and transfer learning, there are numerous algorithms for mapping multiple embeddings to a shared representation space. Recently, these algorithms have been studied in the setting of bilingual dictionary induction where one seeks to align the embeddings of a source and a target language such that translated word pairs lie close to one another in a common representation space. In this paper, we propose a method, Filtered Inner Product Projection (FIPP), for mapping embeddings to a common representation space and evaluate FIPP in the context of bilingual dictionary induction. As semantic shifts are pervasive across languages and domains, FIPP first identifies the common geometric structure in both embeddings and then, only on the common structure, aligns the Gram matrices of these embeddings. Unlike previous approaches, FIPP is applicable even when the source and target embeddings are of differing dimensionalities. We show that our approach outperforms existing methods on the MUSE dataset for various language pairs. Furthermore, FIPP provides computational benefits both in ease of implementation and scalability.

Ziyi Yang, Iman Soltani Bozchalooi, Eric Darve

In this paper, we investigate algorithms for anomaly detection. Previous anomaly detection methods focus on modeling the distribution of non-anomalous data provided during training. However, this does not necessarily ensure the correct detection of anomalous data. We propose a new Regularized Cycle Consistent Generative Adversarial Network (RCGAN) in which deep neural networks are adversarially trained to better recognize anomalous samples. This approach is based on leveraging a penalty distribution with a new definition of the loss function and novel use of discriminator networks. It is based on a solid mathematical foundation, and proofs show that our approach has stronger guarantees for detecting anomalous examples compared to the current state-of-the-art. Experimental results on both real-world and synthetic data show that our model leads to significant and consistent improvements on previous anomaly detection benchmarks. Notably, RCGAN improves on the state-of-the-art on the KDDCUP, Arrhythmia, Thyroid, Musk and CIFAR10 datasets.

Ziyi Yang, Chenguang Zhu, Robert Gmyr et al.

Text summarization aims to extract essential information from a piece of text and transform the text into a concise version. Existing unsupervised abstractive summarization models leverage recurrent neural networks framework while the recently proposed transformer exhibits much more capability. Moreover, most of previous summarization models ignore abundant unlabeled corpora resources available for pretraining. In order to address these issues, we propose TED, a transformer-based unsupervised abstractive summarization system with pretraining on large-scale data. We first leverage the lead bias in news articles to pretrain the model on millions of unlabeled corpora. Next, we finetune TED on target domains through theme modeling and a denoising autoencoder to enhance the quality of generated summaries. Notably, TED outperforms all unsupervised abstractive baselines on NYT, CNN/DM and English Gigaword datasets with various document styles. Further analysis shows that the summaries generated by TED are highly abstractive, and each component in the objective function of TED is highly effective.

Chenguang Zhu, Ziyi Yang, Robert Gmyr et al.

A typical journalistic convention in news articles is to deliver the most salient information in the beginning, also known as the lead bias. While this phenomenon can be exploited in generating a summary, it has a detrimental effect on teaching a model to discriminate and extract important information in general. We propose that this lead bias can be leveraged in our favor in a simple and effective way to pre-train abstractive news summarization models on large-scale unlabeled news corpora: predicting the leading sentences using the rest of an article. We collect a massive news corpus and conduct data cleaning and filtering via statistical analysis. We then apply self-supervised pre-training on this dataset to existing generation models BART and T5 for domain adaptation. Via extensive experiments on six benchmark datasets, we show that this approach can dramatically improve the summarization quality and achieve state-of-the-art results for zero-shot news summarization without any fine-tuning. For example, in the DUC2003 dataset, the ROUGE-1 score of BART increases 13.7% after the lead-bias pre-training. We deploy the model in Microsoft News and provide public APIs as well as a demo website for multi-lingual news summarization.

Ziyi Yang, Chenguang Zhu, Vin Sachidananda et al.

Due to the ubiquitous use of embeddings as input representations for a wide range of natural language tasks, imputation of embeddings for rare and unseen words is a critical problem in language processing. Embedding imputation involves learning representations for rare or unseen words during the training of an embedding model, often in a post-hoc manner. In this paper, we propose an approach for embedding imputation which uses grounded information in the form of a knowledge graph. This is in contrast to existing approaches which typically make use of vector space properties or subword information. We propose an online method to construct a graph from grounded information and design an algorithm to map from the resulting graphical structure to the space of the pre-trained embeddings. Finally, we evaluate our approach on a range of rare and unseen word tasks across various domains and show that our model can learn better representations. For example, on the Card-660 task our method improves Pearson's and Spearman's correlation coefficients upon the state-of-the-art by 11% and 17.8% respectively using GloVe embeddings.

Ziyi Yang, Chenguang Zhu, Weizhu Chen

We propose a simple and robust non-parameterized approach for building sentence representations. Inspired by the Gram-Schmidt Process in geometric theory, we build an orthogonal basis of the subspace spanned by a word and its surrounding context in a sentence. We model the semantic meaning of a word in a sentence based on two aspects. One is its relatedness to the word vector subspace already spanned by its contextual words. The other is the word's novel semantic meaning which shall be introduced as a new basis vector perpendicular to this existing subspace. Following this motivation, we develop an innovative method based on orthogonal basis to combine pre-trained word embeddings into sentence representations. This approach requires zero parameters, along with efficient inference performance. We evaluate our approach on 11 downstream NLP tasks. Our model shows superior performance compared with non-parameterized alternatives and it is competitive to other approaches relying on either large amounts of labelled data or prolonged training time.