Nevin L. Zhang

Weiyan Xie, Xiao-Hui Li, Caleb Chen Cao et al.

Despite the popularity of Vision Transformers (ViTs) and eXplainable AI (XAI), only a few explanation methods have been designed specially for ViTs thus far. They mostly use attention weights of the [CLS] token on patch embeddings and often produce unsatisfactory saliency maps. This paper proposes a novel method for explaining ViTs called ViT-CX. It is based on patch embeddings, rather than attentions paid to them, and their causal impacts on the model output. Other characteristics of ViTs such as causal overdetermination are also considered in the design of ViT-CX. The empirical results show that ViT-CX produces more meaningful saliency maps and does a better job revealing all important evidence for the predictions than previous methods. The explanation generated by ViT-CX also shows significantly better faithfulness to the model. The codes and appendix are available at https://github.com/vaynexie/CausalX-ViT.

Weiyan Xie, Xiao-Hui Li, Zhi Lin et al.

The need to explain the output of a deep neural network classifier is now widely recognized. While previous methods typically explain a single class in the output, we advocate explaining the whole output, which is a probability distribution over multiple classes. A whole-output explanation can help a human user gain an overall understanding of model behaviour instead of only one aspect of it. It can also provide a natural framework where one can examine the evidence used to discriminate between competing classes, and thereby obtain contrastive explanations. In this paper, we propose a contrastive whole-output explanation (CWOX) method for image classification, and evaluate it using quantitative metrics and through human subject studies. The source code of CWOX is available at https://github.com/vaynexie/CWOX.

Nevin L. Zhang, Weiyan Xie, Zhi Lin et al.

Some examples are easier for humans to classify than others. The same should be true for deep neural networks (DNNs). We use the term example perplexity to refer to the level of difficulty of classifying an example. In this paper, we propose a method to measure the perplexity of an example and investigate what factors contribute to high example perplexity. The related codes and resources are available at https://github.com/vaynexie/Example-Perplexity.

Yingxiu Zhao, Bowen Yu, Binyuan Hui et al.

Training large language models (LLMs) with open-domain instruction data has yielded remarkable success in aligning to end tasks and human preferences. Extensive research has highlighted the importance of the quality and diversity of instruction data. However, the impact of data complexity, as a crucial metric, remains relatively unexplored from three aspects: (1)where the sustainability of performance improvements with increasing complexity is uncertain; (2)whether the improvement brought by complexity merely comes from introducing more training tokens; and (3)where the potential benefits of incorporating instructions from easy to difficult are not yet fully understood. In this paper, we propose Tree-Instruct to systematically enhance the instruction complexity in a controllable manner. By adding a specified number of nodes to instructions' semantic trees, this approach not only yields new instruction data from the modified tree but also allows us to control the difficulty level of modified instructions. Our preliminary experiments reveal the following insights: (1)Increasing complexity consistently leads to sustained performance improvements of LLMs. (2)Under the same token budget, a few complex instructions outperform diverse yet simple instructions. (3)Curriculum instruction tuning might not yield the anticipated results; focusing on increasing complexity appears to be the key.

Yingxiu Zhao, Zhiliang Tian, Huaxiu Yao et al.

Building models of natural language processing (NLP) is challenging in low-resource scenarios where only limited data are available. Optimization-based meta-learning algorithms achieve promising results in low-resource scenarios by adapting a well-generalized model initialization to handle new tasks. Nonetheless, these approaches suffer from the memorization overfitting issue, where the model tends to memorize the meta-training tasks while ignoring support sets when adapting to new tasks. To address this issue, we propose a memory imitation meta-learning (MemIML) method that enhances the model's reliance on support sets for task adaptation. Specifically, we introduce a task-specific memory module to store support set information and construct an imitation module to force query sets to imitate the behaviors of some representative support-set samples stored in the memory. A theoretical analysis is provided to prove the effectiveness of our method, and empirical results also demonstrate that our method outperforms competitive baselines on both text classification and generation tasks.

Xingzhi Zhou, Nevin L. Zhang

A deep clustering model conceptually consists of a feature extractor that maps data points to a latent space, and a clustering head that groups data points into clusters in the latent space. Although the two components used to be trained jointly in an end-to-end fashion, recent works have proved it beneficial to train them separately in two stages. In the first stage, the feature extractor is trained via self-supervised learning, which enables the preservation of the cluster structures among the data points. To preserve the cluster structures even better, we propose to replace the first stage with another model that is pretrained on a much larger dataset via self-supervised learning. The method is simple and might suffer from domain shift. Nonetheless, we have empirically shown that it can achieve superior clustering performance. When a vision transformer (ViT) architecture is used for feature extraction, our method has achieved clustering accuracy 94.0%, 55.6% and 97.9% on CIFAR-10, CIFAR-100 and STL-10 respectively. The corresponding previous state-of-the-art results are 84.3%, 47.7% and 80.8%. Our code will be available online with the publication of the paper.

Yingxiu Zhao, Yinhe Zheng, Zhiliang Tian et al.

Lifelong learning (LL) is vital for advanced task-oriented dialogue (ToD) systems. To address the catastrophic forgetting issue of LL, generative replay methods are widely employed to consolidate past knowledge with generated pseudo samples. However, most existing generative replay methods use only a single task-specific token to control their models. This scheme is usually not strong enough to constrain the generative model due to insufficient information involved. In this paper, we propose a novel method, prompt conditioned VAE for lifelong learning (PCLL), to enhance generative replay by incorporating tasks' statistics. PCLL captures task-specific distributions with a conditional variational autoencoder, conditioned on natural language prompts to guide the pseudo-sample generation. Moreover, it leverages a distillation process to further consolidate past knowledge by alleviating the noise in pseudo samples. Experiments on natural language understanding tasks of ToD systems demonstrate that PCLL significantly outperforms competitive baselines in building LL models.

Dongkyu Lee, Ka Chun Cheung, Nevin L. Zhang

Overconfidence has been shown to impair generalization and calibration of a neural network. Previous studies remedy this issue by adding a regularization term to a loss function, preventing a model from making a peaked distribution. Label smoothing smoothes target labels with a pre-defined prior label distribution; as a result, a model is learned to maximize the likelihood of predicting the soft label. Nonetheless, the amount of smoothing is the same in all samples and remains fixed in training. In other words, label smoothing does not reflect the change in probability distribution mapped by a model over the course of training. To address this issue, we propose a regularization scheme that brings dynamic nature into the smoothing parameter by taking model probability distribution into account, thereby varying the parameter per instance. A model in training self-regulates the extent of smoothing on the fly during forward propagation. Furthermore, inspired by recent work in bridging label smoothing and knowledge distillation, our work utilizes self-knowledge as a prior label distribution in softening target labels, and presents theoretical support for the regularization effect by knowledge distillation and the dynamic smoothing parameter. Our regularizer is validated comprehensively, and the result illustrates marked improvements in model generalization and calibration, enhancing robustness and trustworthiness of a model.

Dongkyu Lee, Zhiliang Tian, Yingxiu Zhao et al.

In knowledge distillation, a student model is trained with supervisions from both knowledge from a teacher and observations drawn from a training data distribution. Knowledge of a teacher is considered a subject that holds inter-class relations which send a meaningful supervision to a student; hence, much effort has been put to find such knowledge to be distilled. In this paper, we explore a question that has been given little attention: "when to distill such knowledge." The question is answered in our work with the concept of model calibration; we view a teacher model not only as a source of knowledge but also as a gauge to detect miscalibration of a student. This simple and yet novel view leads to a hard gate knowledge distillation scheme that switches between learning from a teacher model and training data. We verify the gating mechanism in the context of natural language generation at both the token-level and the sentence-level. Empirical comparisons with strong baselines show that hard gate knowledge distillation not only improves model generalization, but also significantly lowers model calibration error.

Kaican Li, Lewei Yao, Jiannan Wu et al.

The ability for AI agents to "think with images" requires a sophisticated blend of reasoning and perception. However, current open multimodal agents still largely fall short on the reasoning aspect crucial for real-world tasks like analyzing documents with dense charts/diagrams and navigating maps. To address this gap, we introduce O3-Bench, a new benchmark designed to evaluate multimodal reasoning with interleaved attention to visual details. O3-Bench features challenging problems that require agents to piece together subtle visual information from distinct image areas through multi-step reasoning. The problems are highly challenging even for frontier systems like OpenAI o3, which only obtains 40.8% accuracy on O3-Bench. To make progress, we propose InSight-o3, a multi-agent framework consisting of a visual reasoning agent (vReasoner) and a visual search agent (vSearcher) for which we introduce the task of generalized visual search -- locating relational, fuzzy, or conceptual regions described in free-form language, beyond just simple objects or figures in natural images. We then present a multimodal LLM purpose-trained for this task via reinforcement learning. As a plug-and-play agent, our vSearcher empowers frontier multimodal models (as vReasoners), significantly improving their performance on a wide range of benchmarks. This marks a concrete step towards powerful o3-like open systems. Our code and dataset can be found at https://github.com/m-Just/InSight-o3 .

Nevin L. Zhang, Kaican Li, Han Gao et al.

Domain generalization (DG) is about learning models that generalize well to new domains that are related to, but different from, the training domain(s). It is a fundamental problem in machine learning and has attracted much attention in recent years. A large number of approaches have been proposed. Different approaches are motivated from different perspectives, making it difficult to gain an overall understanding of the area. In this paper, we propose a causal framework for domain generalization and present an understanding of common DG approaches in the framework. Our work sheds new lights on the following questions: (1) What are the key ideas behind each DG method? (2) Why is it expected to improve generalization to new domains theoretically? (3) How are different DG methods related to each other and what are relative advantages and limitations? By providing a unified perspective on DG, we hope to help researchers better understand the underlying principles and develop more effective approaches for this critical problem in machine learning.

Yingxiu Zhao, Yinhe Zheng, Bowen Yu et al.

Lifelong learning aims to accumulate knowledge and alleviate catastrophic forgetting when learning tasks sequentially. However, existing lifelong language learning methods only focus on the supervised learning setting. Unlabeled data, which can be easily accessed in real-world scenarios, are underexplored. In this paper, we explore a novel setting, semi-supervised lifelong language learning (SSLL), where a model learns sequentially arriving language tasks with both labeled and unlabeled data. We propose an unlabeled data enhanced lifelong learner to explore SSLL. Specially, we dedicate task-specific modules to alleviate catastrophic forgetting and design two modules to exploit unlabeled data: (1) a virtual supervision enhanced task solver is constructed on a teacher-student framework to mine the underlying knowledge from unlabeled data; and (2) a backward augmented learner is built to encourage knowledge transfer from newly arrived unlabeled data to previous tasks. Experimental results on various language tasks demonstrate our model's effectiveness and superiority over competitive baselines under the new setting SSLL.

Xingzhi Zhou, Xin Dong, Chunhao Li et al.

Traditional Chinese medicine (TCM) has relied on specific combinations of herbs in prescriptions to treat various symptoms and signs for thousands of years. Predicting TCM prescriptions poses a fascinating technical challenge with significant practical implications. However, this task faces limitations due to the scarcity of high-quality clinical datasets and the complex relationship between symptoms and herbs. To address these issues, we introduce \textit{DigestDS}, a novel dataset comprising practical medical records from experienced experts in digestive system diseases. We also propose a method, TCM-FTP (TCM Fine-Tuning Pre-trained), to leverage pre-trained large language models (LLMs) via supervised fine-tuning on \textit{DigestDS}. Additionally, we enhance computational efficiency using a low-rank adaptation technique. Moreover, TCM-FTP incorporates data augmentation by permuting herbs within prescriptions, exploiting their order-agnostic nature. Impressively, TCM-FTP achieves an F1-score of 0.8031, significantly outperforming previous methods. Furthermore, it demonstrates remarkable accuracy in dosage prediction, achieving a normalized mean square error of 0.0604. In contrast, LLMs without fine-tuning exhibit poor performance. Although LLMs have demonstrated wide-ranging capabilities, our work underscores the necessity of fine-tuning for TCM prescription prediction and presents an effective way to accomplish this.

Kaican Li, Yifan Zhang, Lanqing Hong et al.

Out-of-distribution (OOD) generalization is a complicated problem due to the idiosyncrasies of possible distribution shifts between training and test domains. Most benchmarks employ diverse datasets to address this issue; however, the degree of the distribution shift between the training domains and the test domains of each dataset remains largely fixed. This may lead to biased conclusions that either underestimate or overestimate the actual OOD performance of a model. Our study delves into a more nuanced evaluation setting that covers a broad range of shift degrees. We show that the robustness of models can be quite brittle and inconsistent under different degrees of distribution shifts, and therefore one should be more cautious when drawing conclusions from evaluations under a limited range of degrees. In addition, we observe that large-scale pre-trained models, such as CLIP, are sensitive to even minute distribution shifts of novel downstream tasks. This indicates that while pre-trained representations may help improve downstream in-distribution performance, they could have minimal or even adverse effects on generalization in certain OOD scenarios of the downstream task if not used properly. In light of these findings, we encourage future research to conduct evaluations across a broader range of shift degrees whenever possible.

Kaican Li, Weiyan Xie, Yongxiang Huang et al.

Fine-tuning foundation models often compromises their robustness to distribution shifts. To remedy this, most robust fine-tuning methods aim to preserve the pre-trained features. However, not all pre-trained features are robust and those methods are largely indifferent to which ones to preserve. We propose dual risk minimization (DRM), which combines empirical risk minimization with worst-case risk minimization, to better preserve the core features of downstream tasks. In particular, we utilize core-feature descriptions generated by LLMs to induce core-based zero-shot predictions which then serve as proxies to estimate the worst-case risk. DRM balances two crucial aspects of model robustness: expected performance and worst-case performance, establishing a new state of the art on various real-world benchmarks. DRM significantly improves the out-of-distribution performance of CLIP ViT-L/14@336 on ImageNet (75.9 to 77.1), WILDS-iWildCam (47.1 to 51.8), and WILDS-FMoW (50.7 to 53.1); opening up new avenues for robust fine-tuning. Our code is available at https://github.com/vaynexie/DRM .

Han Gao, Kaican Li, Weiyan Xie et al.

Domain generalization (DG) is about training models that generalize well under domain shift. Previous research on DG has been conducted mostly in single-source or multi-source settings. In this paper, we consider a third, lesser-known setting where a training domain is endowed with a collection of pairs of examples that share the same semantic information. Such semantic sharing (SS) pairs can be created via data augmentation and then utilized for consistency regularization (CR). We present a theory showing CR is conducive to DG and propose a novel CR method called Logit Attribution Matching (LAM). We conduct experiments on five DG benchmarks and four pretrained models with SS pairs created by both generic and targeted data augmentation methods. LAM outperforms representative single/multi-source DG methods and various CR methods that leverage SS pairs. The code and data of this project are available at https://github.com/Gaohan123/LAM

Rui Min, Zeyu Qin, Nevin L. Zhang et al.

Backdoor attacks pose a significant threat to Deep Neural Networks (DNNs) as they allow attackers to manipulate model predictions with backdoor triggers. To address these security vulnerabilities, various backdoor purification methods have been proposed to purify compromised models. Typically, these purified models exhibit low Attack Success Rates (ASR), rendering them resistant to backdoored inputs. However, Does achieving a low ASR through current safety purification methods truly eliminate learned backdoor features from the pretraining phase? In this paper, we provide an affirmative answer to this question by thoroughly investigating the Post-Purification Robustness of current backdoor purification methods. We find that current safety purification methods are vulnerable to the rapid re-learning of backdoor behavior, even when further fine-tuning of purified models is performed using a very small number of poisoned samples. Based on this, we further propose the practical Query-based Reactivation Attack (QRA) which could effectively reactivate the backdoor by merely querying purified models. We find the failure to achieve satisfactory post-purification robustness stems from the insufficient deviation of purified models from the backdoored model along the backdoor-connected path. To improve the post-purification robustness, we propose a straightforward tuning defense, Path-Aware Minimization (PAM), which promotes deviation along backdoor-connected paths with extra model updates. Extensive experiments demonstrate that PAM significantly improves post-purification robustness while maintaining a good clean accuracy and low ASR. Our work provides a new perspective on understanding the effectiveness of backdoor safety tuning and highlights the importance of faithfully assessing the model's safety.

YipTin Po, Jianming Wang, Yutao Miao et al.

Driving fatigue is a major contributor to traffic accidents and poses a serious threat to road safety. Electroencephalography (EEG) provides a direct measurement of neural activity, yet EEG-based fatigue recognition is hindered by strong non-stationarity and asymmetric neural dynamics. To address these challenges, we propose DeltaGateNet, a novel framework that explicitly captures Bidirectional temporal dynamics for EEG-based driving fatigue recognition. Our key idea is to introduce a Bidirectional Delta module that decomposes first-order temporal differences into positive and negative components, enabling explicit modeling of asymmetric neural activation and suppression patterns. Furthermore, we design a Gated Temporal Convolution module to capture long-term temporal dependencies for each EEG channel using depthwise temporal convolutions and residual learning, preserving channel-wise specificity while enhancing temporal representation robustness. Extensive experiments conducted under both intra-subject and inter-subject evaluation settings on the public SEED-VIG and SADT driving fatigue datasets demonstrate that DeltaGateNet consistently outperforms existing methods. On SEED-VIG, DeltaGateNet achieves an intra-subject accuracy of 81.89% and an inter-subject accuracy of 55.55%. On the balanced SADT 2022 dataset, it attains intra-subject and inter-subject accuracies of 96.81% and 83.21%, respectively, while on the unbalanced SADT 2952 dataset, it achieves 96.84% intra-subject and 84.49% inter-subject accuracy. These results indicate that explicitly modeling Bidirectional temporal dynamics yields robust and generalizable performance under varying subject and class-distribution conditions.

Weiyan Xie, Han Gao, Didan Deng et al.

Recent advances in text-to-image (T2I) models have enabled training-free regional image editing by leveraging the generative priors of foundation models. However, existing methods struggle to balance text adherence in edited regions, context fidelity in unedited areas, and seamless integration of edits. We introduce CannyEdit, a novel training-free framework that addresses this trilemma through two key innovations. First, Selective Canny Control applies structural guidance from a Canny ControlNet only to the unedited regions, preserving the original image's details while allowing for precise, text-driven changes in the specified editable area. Second, Dual-Prompt Guidance utilizes both a local prompt for the specific edit and a global prompt for overall scene coherence. Through this synergistic approach, these components enable controllable local editing for object addition, replacement, and removal, achieving a superior trade-off among text adherence, context fidelity, and editing seamlessness compared to current region-based methods. Beyond this, CannyEdit offers exceptional flexibility: it operates effectively with rough masks or even single-point hints in addition tasks. Furthermore, the framework can seamlessly integrate with vision-language models in a training-free manner for complex instruction-based editing that requires planning and reasoning. Our extensive evaluations demonstrate CannyEdit's strong performance against leading instruction-based editors in complex object addition scenarios.

Jialin Yu, Yuxiang Zhou, Yulan He et al.

Adapting to latent-confounded shifts remains a core challenge in modern AI. These shifts are propagated via latent variables that induce spurious, non-transportable correlations between inputs and labels. One practical failure mode arises when fine-tuning pre-trained foundation models on confounded data (e.g., where certain text tokens or image backgrounds spuriously correlate with the label), leaving models vulnerable at deployment. We frame causal fine-tuning as an identification problem and pose an explicit causal model that decomposes inputs into low-level spurious features and high-level causal representations. Under this family of models, we formalize the assumptions required for identification. Using pre-trained language models as a case study, we show how identifying and adjusting these components during causal fine-tuning enables automatic adaptation to latent-confounded shifts at test time. Experiments on semi-synthetic benchmarks derived from real-world problems demonstrate that our method outperforms black-box domain generalization baselines, illustrating the benefits of explicitly modeling causal structure.

Xingzhi Zhou, Zhiliang Tian, Ka Chun Cheung et al.

Test-time domain adaptation effectively adjusts the source domain model to accommodate unseen domain shifts in a target domain during inference. However, the model performance can be significantly impaired by continuous distribution changes in the target domain and non-independent and identically distributed (non-i.i.d.) test samples often encountered in practical scenarios. While existing memory bank methodologies use memory to store samples and mitigate non-i.i.d. effects, they do not inherently prevent potential model degradation. To address this issue, we propose a resilient practical test-time adaptation (ResiTTA) method focused on parameter resilience and data quality. Specifically, we develop a resilient batch normalization with estimation on normalization statistics and soft alignments to mitigate overfitting and model degradation. We use an entropy-driven memory bank that accounts for timeliness, the persistence of over-confident samples, and sample uncertainty for high-quality data in adaptation. Our framework periodically adapts the source domain model using a teacher-student model through a self-training loss on the memory samples, incorporating soft alignment losses on batch normalization. We empirically validate ResiTTA across various benchmark datasets, demonstrating state-of-the-art performance.

Yingxiu Zhao, Bowen Yu, Haiyang Yu et al.

The goal of document-grounded dialogue (DocGD) is to generate a response by grounding the evidence in a supporting document in accordance with the dialogue context. This process involves four variables that are causally connected. Recently, task-specific pre-training has greatly boosted performances on many downstream tasks. Existing DocGD methods, however, continue to rely on general pre-trained language models without a specifically tailored pre-training approach that explicitly captures the causal relationships. To tackle this issue, we are the first to present a causally-complete dataset construction strategy for building million-level DocGD pre-training corpora. To better capture causality, we further propose a causally-perturbed pre-training strategy, which introduces causal perturbations on the variables and optimizes the overall causal effect. Experiments on three benchmark datasets demonstrate that our causal pre-training achieves considerable and consistent improvements under fully-supervised, low-resource, few-shot, and zero-shot settings.

Dongkyu Lee, Zhiliang Tian, Lanqing Xue et al.

Text style transfer aims to alter the style (e.g., sentiment) of a sentence while preserving its content. A common approach is to map a given sentence to content representation that is free of style, and the content representation is fed to a decoder with a target style. Previous methods in filtering style completely remove tokens with style at the token level, which incurs the loss of content information. In this paper, we propose to enhance content preservation by implicitly removing the style information of each token with reverse attention, and thereby retain the content. Furthermore, we fuse content information when building the target style representation, making it dynamic with respect to the content. Our method creates not only style-independent content representation, but also content-dependent style representation in transferring style. Empirical results show that our method outperforms the state-of-the-art baselines by a large margin in terms of content preservation. In addition, it is also competitive in terms of style transfer accuracy and fluency.

Lanqing Xue, Kaitao Song, Duocai Wu et al.

Rap generation, which aims to produce lyrics and corresponding singing beats, needs to model both rhymes and rhythms. Previous works for rap generation focused on rhyming lyrics but ignored rhythmic beats, which are important for rap performance. In this paper, we develop DeepRapper, a Transformer-based rap generation system that can model both rhymes and rhythms. Since there is no available rap dataset with rhythmic beats, we develop a data mining pipeline to collect a large-scale rap dataset, which includes a large number of rap songs with aligned lyrics and rhythmic beats. Second, we design a Transformer-based autoregressive language model which carefully models rhymes and rhythms. Specifically, we generate lyrics in the reverse order with rhyme representation and constraint for rhyme enhancement and insert a beat symbol into lyrics for rhythm/beat modeling. To our knowledge, DeepRapper is the first system to generate rap with both rhymes and rhythms. Both objective and subjective evaluations demonstrate that DeepRapper generates creative and high-quality raps with rhymes and rhythms. Code will be released on GitHub.

Zhiliang Tian, Wei Bi, Zihan Zhang et al.

Personalized conversation models (PCMs) generate responses according to speaker preferences. Existing personalized conversation tasks typically require models to extract speaker preferences from user descriptions or their conversation histories, which are scarce for newcomers and inactive users. In this paper, we propose a few-shot personalized conversation task with an auxiliary social network. The task requires models to generate personalized responses for a speaker given a few conversations from the speaker and a social network. Existing methods are mainly designed to incorporate descriptions or conversation histories. Those methods can hardly model speakers with so few conversations or connections between speakers. To better cater for newcomers with few resources, we propose a personalized conversation model (PCM) that learns to adapt to new speakers as well as enabling new speakers to learn from resource-rich speakers. Particularly, based on a meta-learning based PCM, we propose a task aggregator (TA) to collect other speakers' information from the social network. The TA provides prior knowledge of the new speaker in its meta-learning. Experimental results show our methods outperform all baselines in appropriateness, diversity, and consistency with speakers.

Leonard K. M. Poon, Nevin L. Zhang, Haoran Xie et al.

Topic modeling has been one of the most active research areas in machine learning in recent years. Hierarchical latent tree analysis (HLTA) has been recently proposed for hierarchical topic modeling and has shown superior performance over state-of-the-art methods. However, the models used in HLTA have a tree structure and cannot represent the different meanings of multiword expressions sharing the same word appropriately. Therefore, we propose a method for extracting and selecting collocations as a preprocessing step for HLTA. The selected collocations are replaced with single tokens in the bag-of-words model before running HLTA. Our empirical evaluation shows that the proposed method led to better performance of HLTA on three of the four data sets tested.

Zhiliang Tian, Wei Bi, Dongkyu Lee et al.

Neural conversation models are known to generate appropriate but non-informative responses in general. A scenario where informativeness can be significantly enhanced is Conversing by Reading (CbR), where conversations take place with respect to a given external document. In previous work, the external document is utilized by (1) creating a context-aware document memory that integrates information from the document and the conversational context, and then (2) generating responses referring to the memory. In this paper, we propose to create the document memory with some anticipated responses in mind. This is achieved using a teacher-student framework. The teacher is given the external document, the context, and the ground-truth response, and learns how to build a response-aware document memory from three sources of information. The student learns to construct a response-anticipated document memory from the first two sources, and the teacher's insight on memory creation. Empirical results show that our model outperforms the previous state-of-the-art for the CbR task.

Lanqing Xue, Xiaopeng Li, Nevin L. Zhang

Although deep neural networks generally have fixed network structures, the concept of dynamic mechanism has drawn more and more attention in recent years. Attention mechanisms compute input-dependent dynamic attention weights for aggregating a sequence of hidden states. Dynamic network configuration in convolutional neural networks (CNNs) selectively activates only part of the network at a time for different inputs. In this paper, we combine the two dynamic mechanisms for text classification tasks. Traditional attention mechanisms attend to the whole sequence of hidden states for an input sentence, while in most cases not all attention is needed especially for long sequences. We propose a novel method called Gated Attention Network (GA-Net) to dynamically select a subset of elements to attend to using an auxiliary network, and compute attention weights to aggregate the selected elements. It avoids a significant amount of unnecessary computation on unattended elements, and allows the model to pay attention to important parts of the sequence. Experiments in various datasets show that the proposed method achieves better performance compared with all baseline models with global or local attention while requiring less computation and achieving better interpretability. It is also promising to extend the idea to more complex attention-based models, such as transformers and seq-to-seq models.

Farhan Khawar, Nevin L. Zhang

Memory-based collaborative filtering methods like user or item k-nearest neighbors (kNN) are a simple yet effective solution to the recommendation problem. The backbone of these methods is the estimation of the empirical similarity between users/items. In this paper, we analyze the spectral properties of the Pearson and the cosine similarity estimators, and we use tools from random matrix theory to argue that they suffer from noise and eigenvalues spreading. We argue that, unlike the Pearson correlation, the cosine similarity naturally possesses the desirable property of eigenvalue shrinkage for large eigenvalues. However, due to its zero-mean assumption, it overestimates the largest eigenvalues. We quantify this overestimation and present a simple re-scaling and noise cleaning scheme. This results in better performance of the memory-based methods compared to their vanilla counterparts.

Farhan Khawar, Nevin L. Zhang

Implicit feedback is the simplest form of user feedback that can be used for item recommendation. It is easy to collect and domain independent. However, there is a lack of negative examples. Existing works circumvent this problem by making various assumptions regarding the unconsumed items, which fail to hold when the user did not consume an item because she was unaware of it. In this paper, we propose as a novel method for addressing the lack of negative examples in implicit feedback. The motivation is that if there is a large group of users who share the same taste and none of them consumed an item, then it is highly likely that the item is irrelevant to this taste. We use Hierarchical Latent Tree Analysis(HLTA) to identify taste-based user groups and make recommendations for a user based on her memberships in the groups.

Farhan Khawar, Nevin L. Zhang

Matrix factorization is a simple and effective solution to the recommendation problem. It has been extensively employed in the industry and has attracted much attention from the academia. However, it is unclear what the low-dimensional matrices represent. We show that matrix factorization can actually be seen as simultaneously calculating the eigenvectors of the user-user and item-item sample co-occurrence matrices. We then use insights from random matrix theory (RMT) to show that picking the top eigenvectors corresponds to removing sampling noise from user/item co-occurrence matrices. Therefore, the low-dimension matrices represent a reduced noise user and item co-occurrence space. We also analyze the structure of the top eigenvector and show that it corresponds to global effects and removing it results in less popular items being recommended. This increases the diversity of the items recommended without affecting the accuracy.

Farhan Khawar, Nevin L. Zhang

Searching, browsing, and recommendations are common ways in which the "choice overload" faced by users in the online marketplace can be mitigated. In this paper we propose the use of hierarchical item categories, obtained from implicit feedback data, to enable efficient browsing and recommendations. We present a method of creating hierarchical item categories from implicit feedback data only i.e., without any other information on the items like name, genre etc. Categories created in this fashion are based on users' co-consumption of items. Thus, they can be more useful for users in finding interesting and relevant items while they are browsing through the hierarchy. We also show that this item hierarchy can be useful in making category based recommendations, which makes the recommendations more explainable and increases the diversity of the recommendations without compromising much on the accuracy. Item hierarchy can also be useful in the creation of an automatic item taxonomy skeleton by bypassing manual labeling and annotation. This can especially be useful for small vendors. Our data-driven hierarchical categories are based on hierarchical latent tree analysis (HLTA) which has been previously used for text analysis. We present a scaled up learning algorithm \emph{HLTA-Forest} so that HLTA can be applied to implicit feedback data.

Zhourong Chen, Xiaopeng Li, Nevin L. Zhang

Despite the popularity of deep learning, structure learning for deep models remains a relatively under-explored area. In contrast, structure learning has been studied extensively for probabilistic graphical models (PGMs). In particular, an efficient algorithm has been developed for learning a class of tree-structured PGMs called hierarchical latent tree models (HLTMs), where there is a layer of observed variables at the bottom and multiple layers of latent variables on top. In this paper, we propose a simple method for learning the structures of feedforward neural networks (FNNs) based on HLTMs. The idea is to expand the connections in the tree skeletons from HLTMs and to use the resulting structures for FNNs. An important characteristic of FNN structures learned this way is that they are sparse. We present extensive empirical results to show that, compared with standard FNNs tuned-manually, sparse FNNs learned by our method achieve better or comparable classification performance with much fewer parameters. They are also more interpretable.

Xiaopeng Li, Zhourong Chen, Nevin L. Zhang

Sparse connectivity is an important factor behind the success of convolutional neural networks and recurrent neural networks. In this paper, we consider the problem of learning sparse connectivity for feedforward neural networks (FNNs). The key idea is that a unit should be connected to a small number of units at the next level below that are strongly correlated. We use Chow-Liu's algorithm to learn a tree-structured probabilistic model for the units at the current level, use the tree to identify subsets of units that are strongly correlated, and introduce a new unit with receptive field over the subsets. The procedure is repeated on the new units to build multiple layers of hidden units. The resulting model is called a TRF-net. Empirical results show that, when compared to dense FNNs, TRF-net achieves better or comparable classification performance with much fewer parameters and sparser structures. They are also more interpretable.

Xiaopeng Li, Zhourong Chen, Leonard K. M. Poon et al.

We investigate a variant of variational autoencoders where there is a superstructure of discrete latent variables on top of the latent features. In general, our superstructure is a tree structure of multiple super latent variables and it is automatically learned from data. When there is only one latent variable in the superstructure, our model reduces to one that assumes the latent features to be generated from a Gaussian mixture model. We call our model the latent tree variational autoencoder (LTVAE). Whereas previous deep learning methods for clustering produce only one partition of data, LTVAE produces multiple partitions of data, each being given by one super latent variable. This is desirable because high dimensional data usually have many different natural facets and can be meaningfully partitioned in multiple ways.

Peixian Chen, Zhourong Chen, Nevin L. Zhang

We propose a novel document generation process based on hierarchical latent tree models (HLTMs) learned from data. An HLTM has a layer of observed word variables at the bottom and multiple layers of latent variables on top. For each document, we first sample values for the latent variables layer by layer via logic sampling, then draw relative frequencies for the words conditioned on the values of the latent variables, and finally generate words for the document using the relative word frequencies. The motivation for the work is to take word counts into consideration with HLTMs. In comparison with LDA-based hierarchical document generation processes, the new process achieves drastically better model fit with much fewer parameters. It also yields more meaningful topics and topic hierarchies. It is the new state-of-the-art for the hierarchical topic detection.

Farhan Khawar, Nevin L. Zhang

Implicit feedback is the simplest form of user feedback that can be used for item recommendation. It is easy to collect and is domain independent. However, there is a lack of negative examples. Previous work tackles this problem by assuming that users are not interested or not as much interested in the unconsumed items. Those assumptions are often severely violated since non-consumption can be due to factors like unawareness or lack of resources. Therefore, non-consumption by a user does not always mean disinterest or irrelevance. In this paper, we propose a novel method called Conformative Filtering (CoF) to address the issue. The motivating observation is that if there is a large group of users who share the same taste and none of them have consumed an item before, then it is likely that the item is not of interest to the group. We perform multidimensional clustering on implicit feedback data using hierarchical latent tree analysis (HLTA) to identify user `tastes' groups and make recommendations for a user based on her memberships in the groups and on the past behavior of the groups. Experiments on two real-world datasets from different domains show that CoF has superior performance compared to several common baselines.

Nevin L. Zhang, Leonard K. M. Poon

Latent tree analysis seeks to model the correlations among a set of random variables using a tree of latent variables. It was proposed as an improvement to latent class analysis --- a method widely used in social sciences and medicine to identify homogeneous subgroups in a population. It provides new and fruitful perspectives on a number of machine learning areas, including cluster analysis, topic detection, and deep probabilistic modeling. This paper gives an overview of the research on latent tree analysis and various ways it is used in practice.

Leonard K. M. Poon, Nevin L. Zhang

Academic researchers often need to face with a large collection of research papers in the literature. This problem may be even worse for postgraduate students who are new to a field and may not know where to start. To address this problem, we have developed an online catalog of research papers where the papers have been automatically categorized by a topic model. The catalog contains 7719 papers from the proceedings of two artificial intelligence conferences from 2000 to 2015. Rather than the commonly used Latent Dirichlet Allocation, we use a recently proposed method called hierarchical latent tree analysis for topic modeling. The resulting topic model contains a hierarchy of topics so that users can browse the topics from the top level to the bottom level. The topic model contains a manageable number of general topics at the top level and allows thousands of fine-grained topics at the bottom level. It also can detect topics that have emerged recently.

Zhourong Chen, Nevin L. Zhang, Dit-Yan Yeung et al.

We are interested in exploring the possibility and benefits of structure learning for deep models. As the first step, this paper investigates the matter for Restricted Boltzmann Machines (RBMs). We conduct the study with Replicated Softmax, a variant of RBMs for unsupervised text analysis. We present a method for learning what we call Sparse Boltzmann Machines, where each hidden unit is connected to a subset of the visible units instead of all of them. Empirical results show that the method yields models with significantly improved model fit and interpretability as compared with RBMs where each hidden unit is connected to all visible units.

Peixian Chen, Nevin L. Zhang, Tengfei Liu et al.

We present a novel method for hierarchical topic detection where topics are obtained by clustering documents in multiple ways. Specifically, we model document collections using a class of graphical models called hierarchical latent tree models (HLTMs). The variables at the bottom level of an HLTM are observed binary variables that represent the presence/absence of words in a document. The variables at other levels are binary latent variables, with those at the lowest latent level representing word co-occurrence patterns and those at higher levels representing co-occurrence of patterns at the level below. Each latent variable gives a soft partition of the documents, and document clusters in the partitions are interpreted as topics. Latent variables at high levels of the hierarchy capture long-range word co-occurrence patterns and hence give thematically more general topics, while those at low levels of the hierarchy capture short-range word co-occurrence patterns and give thematically more specific topics. Unlike LDA-based topic models, HLTMs do not refer to a document generation process and use word variables instead of token variables. They use a tree structure to model the relationships between topics and words, which is conducive to the discovery of meaningful topics and topic hierarchies.

Chen Fu, Nevin L. Zhang, Bao Xin Chen et al.

Objective: To treat patients with vascular mild cognitive impairment (VMCI) using TCM, it is necessary to classify the patients into TCM syndrome types and to apply different treatments to different types. We investigate how to properly carry out the classification using a novel data-driven method known as latent tree analysis. Method: A cross-sectional survey on VMCI was carried out in several regions in northern China from 2008 to 2011, which resulted in a data set that involves 803 patients and 93 symptoms. Latent tree analysis was performed on the data to reveal symptom co-occurrence patterns, and the patients were partitioned into clusters in multiple ways based on the patterns. The patient clusters were matched up with syndrome types, and population statistics of the clusters are used to quantify the syndrome types and to establish classification rules. Results: Eight syndrome types are identified: Qi Deficiency, Qi Stagnation, Blood Deficiency, Blood Stasis, Phlegm-Dampness, Fire-Heat, Yang Deficiency, and Yin Deficiency. The prevalence and symptom occurrence characteristics of each syndrome type are determined. Quantitative classification rules are established for determining whether a patient belongs to each of the syndrome types. Conclusions: A solution for the TCM syndrome classification problem associated with VMCI is established based on the latent tree analysis of unlabeled symptom survey data. The results can be used as a reference in clinic practice to improve the quality of syndrome differentiation and to reduce diagnosis variances across physicians. They can also be used for patient selection in research projects aimed at finding biomarkers for the syndrome types and in randomized control trials aimed at determining the efficacy of TCM treatments of VMCI.

Peixian Chen, Nevin L. Zhang, Leonard K. M. Poon et al.

Hierarchical latent tree analysis (HLTA) is recently proposed as a new method for topic detection. It differs fundamentally from the LDA-based methods in terms of topic definition, topic-document relationship, and learning method. It has been shown to discover significantly more coherent topics and better topic hierarchies. However, HLTA relies on the Expectation-Maximization (EM) algorithm for parameter estimation and hence is not efficient enough to deal with large datasets. In this paper, we propose a method to drastically speed up HLTA using a technique inspired by recent advances in the moments method. Empirical experiments show that our method greatly improves the efficiency of HLTA. It is as efficient as the state-of-the-art LDA-based method for hierarchical topic detection and finds substantially better topics and topic hierarchies.

Nevin L. Zhang, Chen Fu, Teng Fei Liu et al.

Objective: The efficacy of traditional Chinese medicine (TCM) treatments for Western medicine (WM) diseases relies heavily on the proper classification of patients into TCM syndrome types. We develop a data-driven method for solving the classification problem, where syndrome types are identified and quantified based on patterns detected in unlabeled symptom survey data. Method: Latent class analysis (LCA) has been applied in WM research to solve a similar problem, i.e., to identify subtypes of a patient population in the absence of a gold standard. A widely known weakness of LCA is that it makes an unrealistically strong independence assumption. We relax the assumption by first detecting symptom co-occurrence patterns from survey data and use those patterns instead of the symptoms as features for LCA. Results: The result of the investigation is a six-step method: Data collection, symptom co-occurrence pattern discovery, pattern interpretation, syndrome identification, syndrome type identification, and syndrome type classification. A software package called Lantern is developed to support the application of the method. The method is illustrated using a data set on Vascular Mild Cognitive Impairment (VMCI). Conclusions: A data-driven method for TCM syndrome identification and classification is presented. The method can be used to answer the following questions about a Western medicine disease: What TCM syndrome types are there among the patients with the disease? What is the prevalence of each syndrome type? What are the statistical characteristics of each syndrome type in terms of occurrence of symptoms? How can we determine the syndrome type(s) of a patient?

Raphaël Mourad, Christine Sinoquet, Nevin L. Zhang et al.

In data analysis, latent variables play a central role because they help provide powerful insights into a wide variety of phenomena, ranging from biological to human sciences. The latent tree model, a particular type of probabilistic graphical models, deserves attention. Its simple structure - a tree - allows simple and efficient inference, while its latent variables capture complex relationships. In the past decade, the latent tree model has been subject to significant theoretical and methodological developments. In this review, we propose a comprehensive study of this model. First we summarize key ideas underlying the model. Second we explain how it can be efficiently learned from data. Third we illustrate its use within three types of applications: latent structure discovery, multidimensional clustering, and probabilistic inference. Finally, we conclude and give promising directions for future researches in this field.

Yi Wang, Nevin L. Zhang, Tao Chen

We propose a novel method for approximate inference in Bayesian networks (BNs). The idea is to sample data from a BN, learn a latent tree model (LTM) from the data offline, and when online, make inference with the LTM instead of the original BN. Because LTMs are tree-structured, inference takes linear time. In the meantime, they can represent complex relationship among leaf nodes and hence the approximation accuracy is often good. Empirical evidence shows that our method can achieve good approximation accuracy at low online computational cost.