This article focuses on large language models (LLMs) fine-tuning in the scarce data regime (also known as “few-shot learning setting”). We propose a method to increase the generalization capabilities of LLMs based on neural network subspaces. This optimization method, recently introduced in computer vision, aims to improve model generalization by identifying wider local optima through the joint optimization of an entire simplex of models in parameter space. Although this property would be highly beneficial in the context of training large language models in the “few-shot learning” setting, its adaptation to massive, pretrained transformers poses some challenges. First, their considerable number of parameters make it difficult to train several model jointly, and second, their deterministic parameter initialisation schemes make them unfit to the subspace method as originaly proposed. We show in this paper that its application to “Parameter Efficient Fine-Tuning” (PEFT) methods, however, is relatively natural, and we propose to apply it to prefix-tuning, by learning entire simplexes of continous prefixes. We test our method on a variant of the GLUE benchmark adapted to the few-shot learning setting, and show that both our contributions (learning prefix simplexes, and non-deterministic validation metric inference) jointly lead to a gain in average performances compared to state of the art methods.
论文检索
输入标题、作者或关键词，从 2,893 篇学术成果中精准定位
In many real-world scenarios (e.g., academic networks, social platforms), different types of entities are not only associated with texts but also connected by various relationships, which can be abstracted as Text-Attributed Heterogeneous Graphs (TAHGs). Current pretraining tasks for Language Models (LMs) primarily focus on separately learning the textual information of each entity and overlook the crucial aspect of capturing topological connections among entities in TAHGs. In this paper, we present a new pretraining framework for LMs that explicitly considers the topological and heterogeneous information in TAHGs. Firstly, we define a context graph as neighborhoods of a target node within specific orders and propose a topology-aware pretraining task to predict nodes involved in the context graph by jointly optimizing an LM and an auxiliary heterogeneous graph neural network. Secondly, based on the observation that some nodes are text-rich while others have little text, we devise a text augmentation strategy to enrich textless nodes with their neighbors’ texts for handling the imbalance issue. We conduct link prediction and node classification tasks on three datasets from various domains. Experimental results demonstrate the superiority of our approach over existing methods and the rationality of each design. Our code is available at https://github.com/Hope-Rita/THLM.
Uncertainty-aware Parameter-Efficient Self-training for Semi-supervised Language Understanding
PDF ↗The recent success of large pre-trained language models (PLMs) heavily hinges on massive labeled data, which typically produces inferior performance in low-resource scenarios. To remedy this dilemma, we study self-training as one of the predominant semi-supervised learning (SSL) approaches, which utilizes large-scale unlabeled data to generate synthetic examples. However, too many noisy labels will hurt the model performance, and the self-training procedure requires multiple training iterations making it more expensive if all the model parameters of the PLM are updated. This paper presents UPET, a novel Uncertainty-aware Parameter-Efficient self-Training framework to effectively and efficiently address the labeled data scarcity issue. Specifically, we incorporate Monte Carlo (MC) dropout in Bayesian neural network (BNN) to perform uncertainty estimation for the teacher model and then judiciously select reliable pseudo-labeled examples based on confidence and certainty. During the student training, we introduce multiple parameter-efficient learning (PEL) paradigms that allow optimizes only a small percentage of parameters. We also propose a novel Easy-Hard Contrastive Tuning to enhance the robustness and generalization. Extensive experiments over multiple downstream tasks demonstrate that UPET achieves a substantial improvement in terms of performance and efficiency. Our codes and data are released at https: //github.com/wjn1996/UPET.
GRENADE: Graph-Centric Language Model for Self-Supervised Representation Learning on Text-Attributed Graphs
PDF ↗Self-supervised representation learning on text-attributed graphs, which aims to create expressive and generalizable representations for various downstream tasks, has received increasing research attention lately. However, existing methods either struggle to capture the full extent of structural context information or rely on task-specific training labels, which largely hampers their effectiveness and generalizability in practice. To solve the problem of self-supervised representation learning on text-attributed graphs, we develop a novel Graph-Centric Language model – GRENADE. Specifically, GRENADE harnesses the synergy of both pre-trained language model and graph neural network by optimizing with two specialized self-supervised learning algorithms: graph-centric contrastive learning and graph-centric knowledge alignment. The proposed graph-centric self-supervised learning algorithms effectively help GRENADE to capture informative textual semantics as well as structural context information on text-attributed graphs. Through extensive experiments, GRENADE shows its superiority over state-of-the-art methods.
Research on neural networks has focused on understanding a single model trained on a single dataset. However, relatively little is known about the relationships between different models, particularly those trained or tested on different datasets. We address this by studying how the weight space and the underlying loss landscape of different models are interconnected. Specifically, we demonstrate that finetuned models that were optimized for high performance, reside in well-defined regions in weight space, and vice versa – that any model that resides anywhere in those regions also exhibits high performance. Notably, we show that language models that have been finetuned on the same dataset form a tight cluster in the weight space, while models finetuned on different datasets from the same underlying task form a looser cluster. Moreover, traversing around the region between the models leads to new models that perform comparably or even better than models obtained via finetuning, even on tasks that the original models were not finetuned on. Our findings provide insight into the relationships between models, demonstrating that a model positioned between two similar models can acquire the knowledge of both. We leverage this and design a method for selecting a better model for efficient finetuning. Specifically, we show that starting from the center of the region is as effective, if not more, than using the pretrained model in 11 out of 12 datasets, resulting in an average accuracy improvement of 3.06.
Toeplitz Neural Networks (TNNs) have exhibited outstanding performance in various sequence modeling tasks. They outperform commonly used Transformer-based models while benefiting from log-linear space-time complexities. On the other hand, State Space Models (SSMs) achieve lower performance than TNNs in language modeling but offer the advantage of constant inference complexity. In this paper, we aim to combine the strengths of TNNs and SSMs by converting TNNs to SSMs during inference, thereby enabling TNNs to achieve the same constant inference complexities as SSMs. To accomplish this, we formulate the conversion process as an optimization problem and provide a closed-form solution. We demonstrate how to transform the target equation into a Vandermonde linear system problem, which can be efficiently solved using the Discrete Fourier Transform (DFT). Notably, our method requires no training and maintains numerical stability. It can be also applied to any LongConv-based model. To assess its effectiveness, we conduct extensive experiments on language modeling tasks across various settings. Additionally, we compare our method to other gradient-descent solutions, highlighting the superior numerical stability of our approach. The source code is available at https://github.com/OpenNLPLab/ETSC-Exact-Toeplitz-to-SSM-Conversion.
Modeling multi-party conversations (MPCs) with graph neural networks has been proven effective at capturing complicated and graphical information flows. However, existing methods rely heavily on the necessary addressee labels and can only be applied to an ideal setting where each utterance must be tagged with an “@” or other equivalent addressee label. To study the scarcity of addressee labels which is a common issue in MPCs, we propose MADNet that maximizes addressee deduction expectation in heterogeneous graph neural networks for MPC generation. Given an MPC with a few addressee labels missing, existing methods fail to build a consecutively connected conversation graph, but only a few separate conversation fragments instead. To ensure message passing between these conversation fragments, four additional types of latent edges are designed to complete a fully-connected graph. Besides, to optimize the edge-type-dependent message passing for those utterances without addressee labels, an Expectation-Maximization-based method that iteratively generates silver addressee labels (E step), and optimizes the quality of generated responses (M step), is designed. Experimental results on two Ubuntu IRC channel benchmarks show that MADNet outperforms various baseline models on the task of MPC generation, especially under the more common and challenging setting where part of addressee labels are missing.
ATFormer: A Learned Performance Model with Transfer Learning Across Devices for Deep Learning Tensor Programs
PDF ↗The training and inference efficiency of ever-larger deep neural networks highly rely on the performance of tensor operators on specific hardware platforms. Therefore, a compilation-based optimization flow with automatic tensor generation and parameter tuning is necessary for efficient model deployment. While compilation-based methods with performance models can provide dynamic and suitable code optimization, they suffer from a large design space exploration with rough measurement accuracy and poor transferability among different hardware platforms. This paper presents ATFormer, a simple yet efficient design with attention-inspired modules to accurately predict the performance of optimized operators by capturing global and long-range dependencies within a complete scheduling space. Compared with state-of-the-arts, ATFormer can predict the optimal implementation of tensor operators to reduce inference time with minimal effort on modern DNN benchmarks. Furthermore, ATFormer with pre-trained parameters can quickly adapt to different workloads and hardware via transfer learning.
Multilingual neural machine translation (MNMT) offers the convenience of translating between multiple languages with a single model. However, MNMT often suffers from performance degradation in high-resource languages compared to bilingual counterparts. This degradation is commonly attributed to parameter interference, which occurs when parameters are fully shared across all language pairs. In this work, to tackle this issue we propose a gradient-based gradual pruning technique for MNMT. Our approach aims to identify an optimal sub-network for each language pair within the multilingual model by leveraging gradient-based information as pruning criterion and gradually increasing the pruning ratio as schedule. Our approach allows for partial parameter sharing across language pairs to alleviate interference, and each pair preserves its unique parameters to capture language-specific information. Comprehensive experiments on IWSLT and WMT datasets show that our approach yields a notable performance gain on both datasets.
Social Aspects/Accountability, Transparency and Interpretability
We present **VeriX** (**Veri**fied e**X**plainability), a system for producing *optimal robust explanations* and generating *counterfactuals* along decision boundaries of machine learning models. We build such explanations and counterfactuals iteratively using constraint solving techniques and a heuristic based on feature-level sensitivity ranking. We evaluate our method on image recognition benchmarks and a real-world scenario of autonomous aircraft taxiing.
Social Aspects/Trustworthy Machine Learning
The prosperity of deep neural networks (DNNs) is largely benefited from open-source datasets, based on which users can evaluate and improve their methods. In this paper, we revisit backdoor-based dataset ownership verification (DOV), which is currently the only feasible approach to protect the copyright of open-source datasets. We reveal that these methods are fundamentally harmful given that they could introduce malicious misclassification behaviors to watermarked DNNs by the adversaries. In this paper, we design DOV from another perspective by making watermarked models (trained on the protected dataset) correctly classify some `hard' samples that will be misclassified by the benign model. Our method is inspired by the generalization property of DNNs, where we find a \emph{hardly-generalized domain} for the original dataset (as its \emph{domain watermark}). It can be easily learned with the protected dataset containing modified samples. Specifically, we formulate the domain generation as a bi-level optimization and propose to optimize a set of visually-indistinguishable clean-label modified data with similar effects to domain-watermarked samples from the hardly-generalized domain to ensure watermark stealthiness. We also design a hypothesis-test-guided ownership verification via our domain watermark and provide the theoretical analyses of our method. Extensive experiments on three benchmark datasets are conducted, which verify the effectiveness of our method and its resistance to potential adaptive methods.
Optimization
Weight decay is a simple yet powerful regularization technique that has been very widely used in training of deep neural networks (DNNs). While weight decay has attracted much attention, previous studies fail to discover some overlooked pitfalls on large gradient norms resulted by weight decay. In this paper, we discover that, weight decay can unfortunately lead to large gradient norms at the final phase (or the terminated solution) of training, which often indicates bad convergence and poor generalization. To mitigate the gradient-norm-centered pitfalls, we present the first practical scheduler for weight decay, called the Scheduled Weight Decay (SWD) method that can dynamically adjust the weight decay strength according to the gradient norm and significantly penalize large gradient norms during training. Our experiments also support that SWD indeed mitigates large gradient norms and often significantly outperforms the conventional constant weight decay strategy for Adaptive Moment Estimation (Adam).
We study robust adversarial training of two-layer neural networks as a bi-level optimization problem. In particular, for the inner loop that implements the adversarial attack during training using projected gradient descent (PGD), we propose maximizing a \emph{lower bound} on the $0/1$-loss by reflecting a surrogate loss about the origin. This allows us to give a convergence guarantee for the inner-loop PGD attack. Furthermore, assuming the data is linearly separable, we provide precise iteration complexity results for end-to-end adversarial training, which holds for any width and initialization. We provide empirical evidence to support our theoretical results.
Theory/Probabilistic Methods
Most work in neural networks focuses on estimating the conditional mean of a continuous response variable given a set of covariates. In this article, we consider estimating the conditional distribution function using neural networks for both censored and uncensored data. The algorithm is built upon the data structure particularly constructed for the Cox regression with time-dependent covariates. Without imposing any model assumptions, we consider a loss function that is based on the full likelihood where the conditional hazard function is the only unknown nonparametric parameter, for which unconstrained optimization methods can be applied. Through simulation studies, we show that the proposed method possesses desirable performance, whereas the partial likelihood method and the traditional neural networks with $L_2$ loss yields biased estimates when model assumptions are violated. We further illustrate the proposed method with several real-world data sets.
Deep Learning/Graph Neural Networks
We systematically investigate graph transformations that enable standard message passing to simulate state-of-the-art graph neural networks (GNNs) without loss of expressivity. Using these, many state-of-the-art GNNs can be implemented with message passing operations from standard libraries, eliminating many sources of implementation issues and allowing for better code optimization. We distinguish between weak and strong simulation: weak simulation achieves the same expressivity only after several message passing steps while strong simulation achieves this after every message passing step. Our contribution leads to a direct way to translate common operations of non-standard GNNs to graph transformations that allow for strong or weak simulation. Our empirical evaluation shows competitive predictive performance of message passing on transformed graphs for various molecular benchmark datasets, in several cases surpassing the original GNNs.
Theory/Learning Theory
Deep Neural Networks and Reinforcement Learning methods have empirically shown great promise in tackling challenging combinatorial problems. In those methods a deep neural network is used as a solution generator which is then trained by gradient-based methods (e.g., policy gradient) to successively obtain better solution distributions.In this work we introduce a novel theoretical framework for analyzing the effectiveness of such methods. We ask whether there exist generative models that (i) are expressive enough to generate approximately optimal solutions; (ii) have a tractable, i.e, polynomial in the size of the input, number of parameters; (iii) their optimization landscape is benign in the sense that it does not contain sub-optimal stationary points. Our main contribution is a positive answer to this question. Our result holds for a broad class of combinatorial problems including Max- and Min-Cut, Max-$k$-CSP, Maximum-Weight-Bipartite-Matching, and the Traveling Salesman Problem. As a byproduct of our analysis we introduce a novel regularization process over vanilla gradient descent and provide theoretical and experimental evidence that it helps address vanishing-gradient issues and escape bad stationary points.
Theory/Learning Theory
The explicit incorporation of task-specific inductive biases through symmetry has emerged as a general design precept in the development of high-performance machine learning models. For example, group equivariant neural networks have demonstrated impressive performance across various domains and applications such as protein and drug design. A prevalent intuition about such models is that the integration of relevant symmetry results in enhanced generalization. Moreover, it is posited that when the data and/or the model exhibits only approximate or partial symmetry, the optimal or best-performing model is one where the model symmetry aligns with the data symmetry. In this paper, we conduct a formal unified investigation of these intuitions. To begin, we present quantitative bounds that demonstrate how models capturing task-specific symmetries lead to improved generalization. Utilizing this quantification, we examine the more general question of dealing with approximate/partial symmetries. We establish, for a given symmetry group, a quantitative comparison between the approximate equivariance of the model and that of the data distribution, precisely connecting model equivariance error and data equivariance error. Our result delineates the conditions under which the model equivariance error is optimal, thereby yielding the best-performing model for the given task and data.
Deep Learning/Graph Neural Networks
Graph Neural Networks (GNNs) have emerged as a powerful tool for semi-supervised node classification tasks. However, recent studies have revealed various biases in GNNs stemming from both node features and graph topology. In this work, we uncover a new bias - label position bias, which indicates that the node closer to the labeled nodes tends to perform better. We introduce a new metric, the Label Proximity Score, to quantify this bias, and find that it is closely related to performance disparities. To address the label position bias, we propose a novel optimization framework for learning a label position unbiased graph structure, which can be applied to existing GNNs. Extensive experiments demonstrate that our proposed method not only outperforms backbone methods but also significantly mitigates the issue of label position bias in GNNs.
Deep Learning/Graph Neural Networks
We reveal a one-class homophily phenomenon, which is one prevalent property we find empirically in real-world graph anomaly detection (GAD) datasets, i.e., normal nodes tend to have strong connection/affinity with each other, while the homophily in abnormal nodes is significantly weaker than normal nodes. However, this anomaly-discriminative property is ignored by existing GAD methods that are typically built using a conventional anomaly detection objective, such as data reconstruction.In this work, we explore this property to introduce a novel unsupervised anomaly scoring measure for GAD -- local node affinity-- that assigns a larger anomaly score to nodes that are less affiliated with their neighbors, with the affinity defined as similarity on node attributes/representations. We further propose Truncated Affinity Maximization (TAM) that learns tailored node representations for our anomaly measure by maximizing the local affinity of nodes to their neighbors. Optimizing on the original graph structure can be biased by non-homophily edges(i.e., edges connecting normal and abnormal nodes). Thus, TAM is instead optimized on truncated graphs where non-homophily edges are removed iteratively to mitigate this bias. The learned representations result in significantly stronger local affinity for normal nodes than abnormal nodes. Extensive empirical results on 10 real-world GAD datasets show that TAM substantially outperforms seven competing models, achieving over 10% increase in AUROC/AUPRC compared to the best contenders on challenging datasets. Our code is available at https://github.com/mala-lab/TAM-master/.
Deep Learning/Graph Neural Networks
Graph convolutional networks (GCN) with one or two hidden layers have been widely used in handling graph data that are prevalent in various disciplines. Many studies showed that the gain of making GCNs deeper is tiny or even negative. This implies that the complexity of graph data is often limited and shallow models are often sufficient to extract expressive features for various tasks such as node classification. Therefore, in this work, we present a framework called graph convolutional kernel machine (GCKM) for graph-based machine learning. GCKMs are built upon kernel functions integrated with graph convolution. An example is the graph convolutional kernel support vector machine (GCKSVM) for node classification, for which we analyze the generalization error bound and discuss the impact of the graph structure. Compared to GCNs, GCKMs require much less effort in architecture design, hyperparameter tuning, and optimization. More importantly, GCKMs are guaranteed to obtain globally optimal solutions and have strong generalization ability and high interpretability. GCKMs are composable, can be extended to large-scale data, and are applicable to various tasks (e.g., node or graph classification, clustering, feature extraction, dimensionality reduction). The numerical results on benchmark datasets show that, besides the aforementioned advantages, GCKMs have at least competitive accuracy compared to GCNs.