Search

The remarkable success of modern machine learning models on large datasets often demands extensive training time and resource consumption. To save cost, a prevalent research line, known as online batch selection, explores selecting informative subsets during the training process. Although recent efforts achieve advancements by measuring the impact of each sample on generalization, their reliance on additional reference models inherently limits their practical applications, when there are no such ideal models available. On the other hand, the vanilla reference-model-free methods involve independently scoring and selecting data in a sample-wise manner, which sacrifices the diversity and induces the redundancy. To tackle this dilemma, we propose Diversified Batch Selection (DivBS), which is reference-model-free and can efficiently select diverse and representative samples. Specifically, we define a novel selection objective that measures the group-wise orthogonalized representativeness to combat the redundancy issue of previous sample-wise criteria, and provide a principled selection-efficient realization. Extensive experiments across various tasks demonstrate the significant superiority of DivBS in the performance-speedup trade-off. The code is publicly available., Comment: ICML 2024

Diffusion models have demonstrated great potential in generating high-quality content for images, natural language, protein domains, etc. However, how to perform user-preferred targeted generation via diffusion models with only black-box target scores of users remains challenging. To address this issue, we first formulate the fine-tuning of the targeted reserve-time stochastic differential equation (SDE) associated with a pre-trained diffusion model as a sequential black-box optimization problem. Furthermore, we propose a novel covariance-adaptive sequential optimization algorithm to optimize cumulative black-box scores under unknown transition dynamics. Theoretically, we prove a $O(\frac{d^2}{\sqrt{T}})$ convergence rate for cumulative convex functions without smooth and strongly convex assumptions. Empirically, experiments on both numerical test problems and target-guided 3D-molecule generation tasks show the superior performance of our method in achieving better target scores.

Variance reduction techniques are designed to decrease the sampling variance, thereby accelerating convergence rates of first-order (FO) and zeroth-order (ZO) optimization methods. However, in composite optimization problems, ZO methods encounter an additional variance called the coordinate-wise variance, which stems from the random gradient estimation. To reduce this variance, prior works require estimating all partial derivatives, essentially approximating FO information. This approach demands O(d) function evaluations (d is the dimension size), which incurs substantial computational costs and is prohibitive in high-dimensional scenarios. This paper proposes the Zeroth-order Proximal Double Variance Reduction (ZPDVR) method, which utilizes the averaging trick to reduce both sampling and coordinate-wise variances. Compared to prior methods, ZPDVR relies solely on random gradient estimates, calls the stochastic zeroth-order oracle (SZO) in expectation $\mathcal{O}(1)$ times per iteration, and achieves the optimal $\mathcal{O}(d(n + \kappa)\log (\frac{1}{\epsilon}))$ SZO query complexity in the strongly convex and smooth setting, where $\kappa$ represents the condition number and $\epsilon$ is the desired accuracy. Empirical results validate ZPDVR's linear convergence and demonstrate its superior performance over other related methods.

State-of-the-art model for zero-shot cross-lingual spoken language understanding performs cross-lingual unsupervised contrastive learning to achieve the label-agnostic semantic alignment between each utterance and its code-switched data. However, it ignores the precious intent/slot labels, whose label information is promising to help capture the label-aware semantics structure and then leverage supervised contrastive learning to improve both source and target languages' semantics. In this paper, we propose Hybrid and Cooperative Contrastive Learning to address this problem. Apart from cross-lingual unsupervised contrastive learning, we design a holistic approach that exploits source language supervised contrastive learning, cross-lingual supervised contrastive learning and multilingual supervised contrastive learning to perform label-aware semantics alignments in a comprehensive manner. Each kind of supervised contrastive learning mechanism includes both single-task and joint-task scenarios. In our model, one contrastive learning mechanism's input is enhanced by others. Thus the total four contrastive learning mechanisms are cooperative to learn more consistent and discriminative representations in the virtuous cycle during the training process. Experiments show that our model obtains consistent improvements over 9 languages, achieving new state-of-the-art performance., Comment: Accepted by IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI). arXiv admin note: text overlap with arXiv:2312.03716

Stance detection is the view towards a specific target by a given context (\textit{e.g.} tweets, commercial reviews). Target-related knowledge is often needed to assist stance detection models in understanding the target well and making detection correctly. However, prevailing works for knowledge-infused stance detection predominantly incorporate target knowledge from a singular source that lacks knowledge verification in limited domain knowledge. The low-resource training data further increases the challenge for the data-driven large models in this task. To address those challenges, we propose a collaborative knowledge infusion approach for low-resource stance detection tasks, employing a combination of aligned knowledge enhancement and efficient parameter learning techniques. Specifically, our stance detection approach leverages target background knowledge collaboratively from different knowledge sources with the help of knowledge alignment. Additionally, we also introduce the parameter-efficient collaborative adaptor with a staged optimization algorithm, which collaboratively addresses the challenges associated with low-resource stance detection tasks from both network structure and learning perspectives. To assess the effectiveness of our method, we conduct extensive experiments on three public stance detection datasets, including low-resource and cross-target settings. The results demonstrate significant performance improvements compared to the existing stance detection approaches., Comment: 13 pages, 3 figures, Big Data Mining and Analysis

Fine-tuning large language models (LLMs) with classic first-order optimizers entails prohibitive GPU memory due to the backpropagation process. Recent works have turned to zeroth-order optimizers for fine-tuning, which save substantial memory by using two forward passes. However, these optimizers are plagued by the heterogeneity of parameter curvatures across different dimensions. In this work, we propose HiZOO, a diagonal Hessian informed zeroth-order optimizer which is the first work to leverage the diagonal Hessian to enhance zeroth-order optimizer for fine-tuning LLMs. What's more, HiZOO avoids the expensive memory cost and only increases one forward pass per step. Extensive experiments on various models (350M~66B parameters) indicate that HiZOO improves model convergence, significantly reducing training steps and effectively enhancing model accuracy. Moreover, we visualize the optimization trajectories of HiZOO on test functions, illustrating its effectiveness in handling heterogeneous curvatures. Lastly, we provide theoretical proofs of convergence for HiZOO. Code is publicly available at https://anonymous.4open.science/r/HiZOO27F8.

In this study, we present a transductive inference approach on that reward information propagation graph, which enables the effective estimation of rewards for unlabelled data in offline reinforcement learning. Reward inference is the key to learning effective policies in practical scenarios, while direct environmental interactions are either too costly or unethical and the reward functions are rarely accessible, such as in healthcare and robotics. Our research focuses on developing a reward inference method based on the contextual properties of information propagation on graphs that capitalizes on a constrained number of human reward annotations to infer rewards for unlabelled data. We leverage both the available data and limited reward annotations to construct a reward propagation graph, wherein the edge weights incorporate various influential factors pertaining to the rewards. Subsequently, we employ the constructed graph for transductive reward inference, thereby estimating rewards for unlabelled data. Furthermore, we establish the existence of a fixed point during several iterations of the transductive inference process and demonstrate its at least convergence to a local optimum. Empirical evaluations on locomotion and robotic manipulation tasks validate the effectiveness of our approach. The application of our inferred rewards improves the performance in offline reinforcement learning tasks.

Long-term time series forecasting (LTSF) represents a critical frontier in time series analysis, characterized by extensive input sequences, as opposed to the shorter spans typical of traditional approaches. While longer sequences inherently offer richer information for enhanced predictive precision, prevailing studies often respond by escalating model complexity. These intricate models can inflate into millions of parameters, resulting in prohibitive parameter scales. Our study demonstrates, through both analytical and empirical evidence, that decomposition is key to containing excessive model inflation while achieving uniformly superior and robust results across various datasets. Remarkably, by tailoring decomposition to the intrinsic dynamics of time series data, our proposed model outperforms existing benchmarks, using over 99 \% fewer parameters than the majority of competing methods. Through this work, we aim to unleash the power of a restricted set of parameters by capitalizing on domain characteristics--a timely reminder that in the realm of LTSF, bigger is not invariably better.