Hong Liu

Hong Liu, Sang Michael Xie, Zhiyuan Li et al. · stanford

Language modeling on large-scale datasets leads to impressive performance gains on various downstream language tasks. The validation pre-training loss (or perplexity in autoregressive language modeling) is often used as the evaluation metric when developing language models since the pre-training loss tends to be well-correlated with downstream performance (which is itself difficult to evaluate comprehensively). Contrary to this conventional wisdom, this paper shows that 1) pre-training loss cannot fully explain downstream performance and 2) flatness of the model is well-correlated with downstream performance where pre-training loss is not. On simplified datasets, we identify three ways to produce models with the same (statistically optimal) pre-training loss but different downstream performance: continue pre-training after convergence, increasing the model size, and changing the training algorithm. These experiments demonstrate the existence of implicit bias of pre-training algorithms/optimizers -- among models with the same minimal pre-training loss, they implicitly prefer more transferable ones. Toward understanding this implicit bias, we prove that SGD with standard mini-batch noise implicitly prefers flatter minima in language models, and empirically observe a strong correlation between flatness and downstream performance among models with the same minimal pre-training loss. We also prove in a synthetic language setting that among the models with the minimal pre-training loss, the flattest model transfers to downstream tasks.

Ke Feng, Dahai Liu, Yongxin Liu et al.

The current National Airspace System (NAS) is reaching capacity due to increased air traffic, and is based on outdated pre-tactical planning. This study proposes a more dynamic airspace configuration (DAC) approach that could increase throughput and accommodate fluctuating traffic, ideal for emergencies. The proposed approach constructs the airspace as a constraints-embedded graph, compresses its dimensions, and applies a spectral clustering-enabled adaptive algorithm to generate collaborative airport groups and evenly distribute workloads among them. Under various traffic conditions, our experiments demonstrate a 50\% reduction in workload imbalances. This research could ultimately form the basis for a recommendation system for optimized airspace configuration. Code available at https://github.com/KeFenge2022/GraphDAC.git

Hong Liu

In this paper, we delve into the essential components of adversarial training which is a pioneering defense technique against adversarial attacks. We indicate that some factors such as the loss function, learning rate scheduler, and data augmentation, which are independent of the model architecture, will influence adversarial robustness and generalization. When these factors are controlled for, we introduce a simple baseline approach, termed SimpleAT, that performs competitively with recent methods and mitigates robust overfitting. We conduct extensive experiments on CIFAR-10/100 and Tiny-ImageNet, which validate the robustness of SimpleAT against state-of-the-art adversarial attackers such as AutoAttack. Our results also demonstrate that SimpleAT exhibits good performance in the presence of various image corruptions, such as those found in the CIFAR-10-C. In addition, we empirically show that SimpleAT is capable of reducing the variance in model predictions, which is considered the primary contributor to robust overfitting. Our results also reveal the connections between SimpleAT and many advanced state-of-the-art adversarial defense methods.

Qifang Zhao, Weidong Ren, Tianyu Li et al.

We introduceGraphGPT, a novel self-supervised generative pre-trained model for graph learning based on the Graph Eulerian Transformer (GET). First, we propose GET, which combines a standard transformer encoder or decoder architecture with an innovative graph-to-sequence transformation method. This method converts graphs or sampled subgraphs into sequences of tokens representing nodes, edges, and attributes in a reversible manner using Eulerian paths. We pre-train GET using either of the two self-supervised tasks: next-token prediction (NTP) and scheduled masked-token prediction (SMTP). The pre-trained model is then fine-tuned for downstream tasks such as graph-, edge-, and node-level prediction. Despite its simplicity, GraphGPT achieves performance comparable to or surpassing state-of-the-art methods on multiple large-scale Open Graph Benchmark (OGB) datasets. It demonstrates exceptional results on the molecular property prediction dataset PCQM4Mv2 and the protein-protein interaction dataset ogbl-ppa. Notably, generative pre-training enables scaling GraphGPT to 2 billion parameters while maintaining performance gains - a breakthrough that overcomes the scalability limitations of traditional Graph Neural Networks (GNNs) and prior graph transformers (GTs). To advance research in graph foundation models and facilitate scientific discovery in chemistry, materials science, and related fields, we will release the source code (https://github.com/alibaba/graph-gpt) and pre-trained checkpoints.

Yuhao Yi, Ronghui You, Hong Liu et al.

Byzantine machine learning has garnered considerable attention in light of the unpredictable faults that can occur in large-scale distributed learning systems. The key to secure resilience against Byzantine machines in distributed learning is resilient aggregation mechanisms. Although abundant resilient aggregation rules have been proposed, they are designed in ad-hoc manners, imposing extra barriers on comparing, analyzing, and improving the rules across performance criteria. This paper studies near-optimal aggregation rules using clustering in the presence of outliers. Our outlier-robust clustering approach utilizes geometric properties of the update vectors provided by workers. Our analysis show that constant approximations to the 1-center and 1-mean clustering problems with outliers provide near-optimal resilient aggregators for metric-based criteria, which have been proven to be crucial in the homogeneous and heterogeneous cases respectively. In addition, we discuss two contradicting types of attacks under which no single aggregation rule is guaranteed to improve upon the naive average. Based on the discussion, we propose a two-phase resilient aggregation framework. We run experiments for image classification using a non-convex loss function. The proposed algorithms outperform previously known aggregation rules by a large margin with both homogeneous and heterogeneous data distributions among non-faulty workers. Code and appendix are available at https://github.com/jerry907/AAAI24-RASHB.

Hong Liu, Zhaobiao Lv, Zhijian Ou et al.

Energy-based language models (ELMs) parameterize an unnormalized distribution for natural sentences and are radically different from popular autoregressive language models (ALMs). As an important application, ELMs have been successfully used as a means for calculating sentence scores in speech recognition, but they all use less-modern CNN or LSTM networks. The recent progress in Transformer networks and large pretrained models such as BERT and GPT2 opens new possibility to further advancing ELMs. In this paper, we explore different architectures of energy functions and different training methods to investigate the capabilities of ELMs in rescoring for speech recognition, all using large pretrained models as backbones.

Bei Yang, Jie Gu, Ke Liu et al.

User Modeling plays an essential role in industry. In this field, task-agnostic approaches, which generate general-purpose representation applicable to diverse downstream user cognition tasks, is a promising direction being more valuable and economical than task-specific representation learning. With the rapid development of Internet service platforms, user behaviors have been accumulated continuously. However, existing general-purpose user representation researches have little ability for full-life cycle modeling on extremely long behavior sequences since user registration. In this study, we propose a novel framework called full- Life cycle User Representation Model (LURM) to tackle this challenge. Specifically, LURM consists of two cascaded sub-models: (I) Bag-of-Interests (BoI) encodes user behaviors in any time period into a sparse vector with super-high dimension (e.g., 10^5); (II) Self-supervised Multi-anchor Encoder Network (SMEN) maps sequences of BoI features to multiple low-dimensional user representations. Specially, SMEN achieves almost lossless dimensionality reduction, benefiting from a novel multi-anchor module which can learn different aspects of user interests. Experiments on several benchmark datasets show that our approach outperforms state-of-the-art general-purpose representation methods.

Qinghui Sun, Jie Gu, Bei Yang et al.

User representation is essential for providing high-quality commercial services in industry. Universal user representation has received many interests recently, with which we can be free from the cumbersome work of training a specific model for each downstream application. In this paper, we attempt to improve universal user representation from two points of views. First, a contrastive self-supervised learning paradigm is presented to guide the representation model training. It provides a unified framework that allows for long-term or short-term interest representation learning in a data-driven manner. Moreover, a novel multi-interest extraction module is presented. The module introduces an interest dictionary to capture principal interests of the given user, and then generate his/her interest-oriented representations via behavior aggregation. Experimental results demonstrate the effectiveness and applicability of the learned user representations.

Yiming Hei, Renyu Yang, Hao Peng et al.

Android is undergoing unprecedented malicious threats daily, but the existing methods for malware detection often fail to cope with evolving camouflage in malware. To address this issue, we present HAWK, a new malware detection framework for evolutionary Android applications. We model Android entities and behavioural relationships as a heterogeneous information network (HIN), exploiting its rich semantic metastructures for specifying implicit higher-order relationships. An incremental learning model is created to handle the applications that manifest dynamically, without the need for re-constructing the whole HIN and the subsequent embedding model. The model can pinpoint rapidly the proximity between a new application and existing in-sample applications and aggregate their numerical embeddings under various semantics. Our experiments examine more than 80,860 malicious and 100,375 benign applications developed over a period of seven years, showing that HAWK achieves the highest detection accuracy against baselines and takes only 3.5ms on average to detect an out-of-sample application, with the accelerated training time of 50x faster than the existing approach.

Da Sun Handason Tam, Wing Cheong Lau, Bin Hu et al.

Rapid and massive adoption of mobile/ online payment services has brought new challenges to the service providers as well as regulators in safeguarding the proper uses such services/ systems. In this paper, we leverage recent advances in deep-neural-network-based graph representation learning to detect abnormal/ suspicious financial transactions in real-world e-payment networks. In particular, we propose an end-to-end Graph Convolution Network (GCN)-based algorithm to learn the embeddings of the nodes and edges of a large-scale time-evolving graph. In the context of e-payment transaction graphs, the resultant node and edge embeddings can effectively characterize the user-background as well as the financial transaction patterns of individual account holders. As such, we can use the graph embedding results to drive downstream graph mining tasks such as node-classification to identify illicit accounts within the payment networks. Our algorithm outperforms state-of-the-art schemes including GraphSAGE, Gradient Boosting Decision Tree and Random Forest to deliver considerably higher accuracy (94.62% and 86.98% respectively) in classifying user accounts within 2 practical e-payment transaction datasets. It also achieves outstanding accuracy (97.43%) for another biomedical entity identification task while using only edge-related information.