ICLR 2024 Papers

Transformers have demonstrated effectiveness in in-context solving data-fitting problems from various (latent) models, as reported by Garg et al. (2022). However, the absence of an inherent iterative structure in the transformer architecture presents a challenge in emulating the iterative algorithms, which are commonly employed in traditional machine learning methods. To address this, we propose the utilization of looped transformer architecture and its associated training methodology, with the aim of incorporating iterative characteristics into the transformer architectures. Experimental results suggest that the looped transformer achieves performance comparable to the standard transformer in solving various data-fitting problems, while utilizing less than 10% of the parameter count.

An effective technique for obtaining high-quality representations is adding a projection head on top of the encoder during training, then discarding it and using the pre-projection representations. Despite its proven practical effectiveness, the reason behind the success of this technique is poorly understood. The pre-projection representations are not directly optimized by the loss function, raising the question: what makes them better? In this work, we provide a rigorous theoretical answer to this question. We start by examining linear models trained with self-supervised contrastive loss. We reveal that the implicit bias of training algorithms leads to layer-wise progressive feature weighting, where features become increasingly unequal as we go deeper into the layers. Consequently, lower layers tend to have more normalized and less specialized representations. We theoretically characterize scenarios where such representations are more beneficial, highlighting the intricate interplay between data augmentation and input features. Additionally, we demonstrate that introducing non-linearity into the network allows lower layers to learn features that are completely absent in higher layers. Finally, we show how this mechanism improves the robustness in supervised contrastive learning and supervised learning. We empirically validate our results through various experiments on CIFAR-10/100, UrbanCars and shifted versions of ImageNet. We also introduce a potential alternative to projection head, which offers a more interpretable and controllable design.

While backpropagation (BP) has achieved widespread success in deep learning, it faces two prominent challenges: computational inefficiency and biological implausibility. In response to these challenges, local supervision, encompassing Local Learning (LL) and Forward Learning (FL), has emerged as a promising research direction. LL employs module-wise BP to achieve competitive results yet relies on module-wise auxiliary networks, which increase memory and parameter demands. Conversely, FL updates layer weights without BP and auxiliary networks but falls short of BP’s performance. This paper proposes a simple yet effective objective within a contrastive learning framework for local supervision without auxiliary networks. Given the insight that the existing contrastive learning framework for local supervision is susceptible to task-irrelevant information without auxiliary networks, we present DICTIONARY CONTRASTIVE LEARNING (DCL) that optimizes the similarity between local features and label embeddings. Our method using static label embeddings yields substantial performance improvements in the FL scenario, outperforming state-of-the-art FL approaches. Moreover, our method using adaptive label embeddings closely approaches the performance achieved by LL while achieving superior memory and parameter efficiency.

Tabular data have been playing a mostly important role in diverse real-world fields, such as healthcare, engineering, finance, etc. With the recent success of deep learning, many tabular machine learning (ML) methods based on deep networks (e.g., Transformer, ResNet) have achieved competitive performance on tabular benchmarks. However, existing deep tabular ML methods suffer from the representation entanglement and localization, which largely hinders their prediction performance and leads to performance inconsistency on tabular tasks. To overcome these problems, we explore a novel direction of applying prototype learning for tabular ML and propose a prototype-based tabular representation learning framework, PTaRL, for tabular prediction tasks. The core idea of PTaRL is to construct prototype-based projection space (P-Space) and learn the disentangled representation around global data prototypes. Specifically, PTaRL mainly involves two stages: (i) Prototype Generating, that constructs global prototypes as the basis vectors of P-Space for representation, and (ii) Prototype Projecting, that projects the data samples into P-Space and keeps the core global data information via Optimal Transport. Then, to further acquire the disentangled representations, we constrain PTaRL with two strategies: (i) to diversify the coordinates towards global prototypes of different representations within P-Space, we bring up a diversifying constraint for representation calibration; (ii) to avoid prototype entanglement in P-Space, we introduce a matrix orthogonalization constraint to ensure the independence of global prototypes. Finally, we conduct extensive experiments in PTaRL coupled with state-of-the-art deep tabular ML models on various tabular benchmarks and the results have shown our consistent superiority.

Partial-label learning (PLL) allows each training example to be equipped with a set of candidate labels. Existing deep PLL research focuses on a \emph{learning-centric} perspective to design various training strategies for label disambiguation i.e., identifying the concealed true label from the candidate label set, for model training. However, when the size of the candidate label set becomes excessively large, these learning-centric strategies would be unable to find the true label for model training, thereby causing performance degradation. This motivates us to think from a \emph{data-centric} perspective and pioneer a new PLL-related task called candidate label set pruning (CLSP) that aims to filter out certain potential false candidate labels in a training-free manner. To this end, we propose the first CLSP method based on the inconsistency between the representation space and the candidate label space. Specifically, for each candidate label of a training instance, if it is not a candidate label of the instance's nearest neighbors in the representation space, then it has a high probability of being a false label. Based on this intuition, we employ a per-example pruning scheme that filters out a specific proportion of high-probability false candidate labels. Theoretically, we prove an upper bound of the pruning error rate and analyze how the quality of representations affects our proposed method. Empirically, extensive experiments on both benchmark-simulated and real-world PLL datasets validate the great value of CLSP to significantly improve many state-of-the-art deep PLL methods.

In recent research, Tensor Product Representation (TPR) is applied for the systematic generalization task of deep neural networks by learning the compositional structure of data. However, such prior works show limited performance in discovering and representing the symbolic structure from unseen test data because their decomposition to the structural representations was incomplete. In this work, we propose an Attention-based Iterative Decomposition (AID) module designed to enhance the decomposition operations for the structured representations encoded from the sequential input data with TPR. Our AID can be easily adapted to any TPR-based model and provides enhanced systematic decomposition through a competitive attention mechanism between input features and structured representations. In our experiments, AID shows effectiveness by significantly improving the performance of TPR-based prior works on the series of systematic generalization tasks. Moreover, in the quantitative and qualitative evaluations, AID produces more compositional and well-bound structural representations than other works.

Fixed classifiers in neural networks for classification problems have demonstrated cost efficiency and even outperformed learnable classifiers in some popular benchmarks when incorporating orthogonality. Despite these advantages, prior research has yet to investigate the training dynamics of fixed orthogonal classifiers on neural collapse, a recently clarified phenomenon that last-layer features converge to a specific form, called simplex ETF, in training classification models involving the post-zero-error phase. Ensuring this phenomenon is critical for obtaining global optimality in a layer-peeled model, potentially leading to enhanced performance in practice. However, fixed orthogonal classifiers cannot invoke neural collapse due to their geometric limitations. To overcome the limits, we analyze a $\textit{zero-mean neural collapse}$ considering the orthogonality in non-negative Euclidean space. Then, we propose a $\textit{fixed non-negative orthogonal classifier}$ that induces the optimal solution and maximizes the margin of an orthogonal layer-peeled model by satisfying the properties of zero-mean neural collapse. Building on this foundation, we exploit a $\textit{feature dimension separation}$ effect inherent in our classifier for further purposes: (1) enhances softmax masking by mitigating feature interference in continual learning and (2) tackles the limitations of mixup on the hypersphere in imbalanced learning. We conducted comprehensive experiments on various datasets and architectures and demonstrated significant performance improvements.

Self-supervised representation learning (SSRL) has advanced considerably by exploiting the transformation invariance assumption under artificially designed data augmentations. While augmentation-based SSRL algorithms push the boundaries of performance in computer vision and natural language processing, they are often not directly applicable to other data modalities, and can conflict with application-specific data augmentation constraints. This paper presents an SSRL approach that can be applied to any data modality and network architecture because it does not rely on augmentations or masking. Specifically, we show that high-quality data representations can be learned by reconstructing random data projections. We evaluate the proposed approach on a wide range of representation learning tasks that span diverse modalities and real-world applications. We show that it outperforms multiple state-of-the-art SSRL baselines. Due to its wide applicability and strong empirical results, we argue that learning from randomness is a fruitful research direction worthy of attention and further study.

Modern techniques like contrastive learning have been effectively used in many areas, including computer vision, natural language processing, and graph-structured data. Creating positive examples that assist the model in learning robust and discriminative representations is a crucial stage in contrastive learning approaches. Usually, preset human intuition directs the selection of relevant data augmentations. Due to patterns that are easily recognized by humans, this rule of thumb works well in the vision and language domains. However, it is impractical to visually inspect the temporal structures in time series. The diversity of time series augmentations at both the dataset and instance levels makes it difficult to choose meaningful augmentations on the fly. Thus, although prevalent, contrastive learning with data augmentation has been less studied in the time series domain. In this study, we address this gap by analyzing time series data augmentation using information theory and summarizing the most commonly adopted augmentations in a unified format. We then propose a parametric augmentation method, AutoTCL, which can be adaptively employed to support time series representation learning. The proposed approach is encoder-agnostic, allowing it to be seamlessly integrated with different backbone encoders. Experiments on univariate forecasting tasks demonstrate the highly competitive results of our method, with an average 6.5\% reduction in MSE and 4.7\% in MAE over the leading baselines. In classification tasks, AutoTCL achieves a $1.2\%$ increase in average accuracy.

We present Point2SSM, a novel unsupervised learning approach for constructing correspondence-based statistical shape models (SSMs) directly from raw point clouds. SSM is crucial in clinical research, enabling population-level analysis of morphological variation in bones and organs. Traditional methods of SSM construction have limitations, including the requirement of noise-free surface meshes or binary volumes, reliance on assumptions or templates, and prolonged inference times due to simultaneous optimization of the entire cohort. Point2SSM overcomes these barriers by providing a data-driven solution that infers SSMs directly from raw point clouds, reducing inference burdens and increasing applicability as point clouds are more easily acquired. While deep learning on 3D point clouds has seen success in unsupervised representation learning and shape correspondence, its application to anatomical SSM construction is largely unexplored. We conduct a benchmark of state-of-the-art point cloud deep networks on the SSM task, revealing their limited robustness to clinical challenges such as noisy, sparse, or incomplete input and limited training data. Point2SSM addresses these issues through an attention-based module, providing effective correspondence mappings from learned point features. Our results demonstrate that the proposed method significantly outperforms existing networks in terms of accurate surface sampling and correspondence, better capturing population-level statistics. The source code is provided at https://github.com/jadie1/Point2SSM.

In this research, we focus on the problem of learning monotonic neural networks, as preserving the monotonicity of a model with respect to a subset of inputs is crucial for practical applications across various domains. Although several methods have recently been proposed to address this problem, they have limitations such as not guaranteeing monotonicity in certain cases, requiring additional inference time, lacking scalability with increasing network size and number of monotonic inputs, and manipulating network weights during training. To overcome these limitations, we introduce a simple but novel architecture of the partially connected network which incorporates a 'scalable monotonic hidden layer' comprising three units: the exponentiated unit, ReLU unit, and confluence unit. This allows for the repetitive integration of the scalable monotonic hidden layers without other structural constraints. Consequently, our method offers ease of implementation and rapid training through the conventional error-backpropagation algorithm. We accordingly term this method as Scalable Monotonic Neural Networks (SMNN). Numerical experiments demonstrated that our method achieved comparable prediction accuracy to the state-of-the-art approaches while effectively addressing the aforementioned weaknesses.

Many real-world networks evolve over time, and predicting the evolution of such networks remains a challenging task. Graph Neural Networks (GNNs) have shown empirical success for learning on static graphs, but they lack the ability to effectively learn from nodes and edges with different timestamps. Consequently, the prediction of future properties in temporal graphs remains a relatively under-explored area. In this paper, we aim to bridge this gap by introducing a principled framework, named GraphPulse. The framework combines two important techniques for the analysis of temporal graphs within a Newtonian framework. First, we employ the Mapper method, a key tool in topological data analysis, to extract essential clustering information from graph nodes. Next, we harness the sequential modeling capabilities of Recurrent Neural Networks (RNNs) for temporal reasoning regarding the graph's evolution. Through extensive experimentation, we demonstrate that our model enhances the ROC-AUC metric by 10.2\% in comparison to the top-performing state-of-the-art method across various temporal networks. We provide the implementation of GraphPulse at https://github.com/kiarashamsi/GraphPulse.

Hypergraphs are marked by complex topology, expressing higher-order interactions among multiple nodes with hyperedges, and better capturing the topology is essential for effective representation learning. Recent advances in generative self-supervised learning (SSL) suggest that hypergraph neural networks (HNNs) learned from generative self-supervision have the potential to effectively encode the complex hypergraph topology. Designing a generative SSL strategy for hypergraphs, however, is not straightforward. Questions remain with regard to its generative SSL task, connection to downstream tasks, and empirical properties of learned representations. In light of the promises and challenges, we propose a novel generative SSL strategy for hypergraphs. We first formulate a generative SSL task on hypergraphs, hyperedge filling, and highlight its theoretical connection to node classification. Based on the generative SSL task, we propose a hypergraph SSL method, HYPEBOY. HYPEBOY learns effective general-purpose hypergraph representations, outperforming 15 baseline methods across 11 benchmark datasets. To our knowledge, this is the first study on generative SSL on hypergraphs, and we demonstrate its theoretical and empirical strengths for hypergraph representation learning.

Despite the dominance and effectiveness of scaling, resulting in large networks with hundreds of billions of parameters, the necessity to train overparameterized models remains poorly understood, while training costs grow exponentially. In this paper, we explore parameter-efficient training techniques as an approach to training large neural networks. We introduce a novel method called ReLoRA, which utilizes low-rank updates to train high-rank networks. We apply ReLoRA to training transformer language models with up to 1.3B parameters and demonstrate comparable performance to regular neural network training. ReLoRA saves up to 5.5Gb of RAM per GPU and improves training speed by 9-40% depending on the model size and hardware setup. Our findings show the potential of parameter- efficient techniques for large-scale pre-training. Our code is available on GitHub.

Utilizing massive web-scale datasets has led to unprecedented performance gains in machine learning models, but also imposes outlandish compute requirements for their training. In order to improve training and data efficiency, we here push the limits of pruning large-scale multimodal datasets for training CLIP-style models. Today’s most effective pruning method on ImageNet clusters data samples into separate concepts according to their embedding and prunes away the most proto- typical samples. We scale this approach to LAION and improve it by noting that the pruning rate should be concept-specific and adapted to the complexity of the concept. Using a simple and intuitive complexity measure, we are able to reduce the training cost to a quarter of regular training. More specifically, we are able to outperform the LAION-trained OpenCLIP-ViT-B/32 model on ImageNet zero-shot accuracy by 1.1p.p. while only using 27.7% of the data and training compute. On the DataComp Medium benchmark, we achieve a new state-of-the-art ImageNet zero-shot accuracy and a competitive average zero-shot accuracy on 38 evaluation tasks.

Updating truncated Singular Value Decomposition (SVD) has extensive applications in representation learning. The continuous evolution of massive-scaled data matrices in practical scenarios highlights the importance of aligning SVD-based models with fast-paced updates. Recent methods for updating truncated SVD can be recognized as Rayleigh-Ritz projection procedures where their projection matrices are augmented based on the original singular vectors. However, the updating process in these methods densifies the update matrix and applies the projection to all singular vectors, resulting in inefficiency. This paper presents a novel method for dynamically approximating the truncated SVD of a sparse and temporally evolving matrix. The proposed method takes advantage of sparsity in the orthogonalization process of the augment matrices and employs an extended decomposition to store projections in the column space of singular vectors independently. Numerical experimental results on updating truncated SVD for evolving sparse matrices show an order of magnitude improvement in the efficiency of our proposed method while maintaining precision comparing to previous methods.

Early stopping methods in deep learning face the challenge of balancing the volume of training and validation data, especially in the presence of label noise. Concretely, sparing more data for validation from training data would limit the performance of the learned model, yet insufficient validation data could result in a sub-optimal selection of the desired model. In this paper, we propose a novel early stopping method called Label Wave, which does not require validation data for selecting the desired model in the presence of label noise. It works by tracking the changes in the model's predictions on the training set during the training process, aiming to halt training before the model unduly fits mislabeled data. This method is empirically supported by our observation that minimum fluctuations in predictions typically occur at the training epoch before the model excessively fits mislabeled data. Through extensive experiments, we show both the effectiveness of the Label Wave method across various settings and its capability to enhance the performance of existing methods for learning with noisy labels.

We propose a novel framework for incorporating unlabeled data into semi-supervised classification problems, where scenarios involving the minimization of either i) adversarially robust or ii) non-robust loss functions have been considered. Notably, we allow the unlabeled samples to deviate slightly (in total variation sense) from the in-domain distribution. The core idea behind our framework is to combine Distributionally Robust Optimization (DRO) with self-supervised training. As a result, we also leverage efficient polynomial-time algorithms for the training stage. From a theoretical standpoint, we apply our framework on the classification problem of a mixture of two Gaussians in $\mathbb{R}^d$, where in addition to the $m$ independent and labeled samples from the true distribution, a set of $n$ (usually with $n\gg m$) out of domain and unlabeled samples are gievn as well. Using only the labeled data, it is known that the generalization error can be bounded by $\propto\left(d/m\right)^{1/2}$. However, using our method on both isotropic and non-isotropic Gaussian mixture models, one can derive a new set of analytically explicit and non-asymptotic bounds which show substantial improvement on the generalization error compared ERM. Our results underscore two significant insights: 1) out-of-domain samples, even when unlabeled, can be harnessed to narrow the generalization gap, provided that the true data distribution adheres to a form of the "cluster assumption", and 2) the semi-supervised learning paradigm can be regarded as a special case of our framework when there are no distributional shifts. We validate our claims through experiments conducted on a variety of synthetic and real-world datasets.

The disparity in accuracy between classes in standard training is amplified during adversarial training, a phenomenon termed the robust fairness problem. Existing methodologies aimed to enhance robust fairness by sacrificing the model's performance on easier classes in order to improve its performance on harder ones. However, we observe that under adversarial attacks, the majority of the model's predictions for samples from the worst class are biased towards classes similar to the worst class, rather than towards the easy classes. Through theoretical and empirical analysis, we demonstrate that robust fairness deteriorates as the distance between classes decreases. Motivated by these insights, we introduce the Distance-Aware Fair Adversarial Training (DAFA) methodology, which addresses robust fairness by taking into account the similarities between classes. Specifically, our method assigns distinct adversarial margins and loss weights to each class and adjusts them to encourage a trade-off in robustness among similar classes. Experimental results across various datasets demonstrate that our method not only maintains average robust accuracy but also significantly improves the worst robust accuracy, indicating a marked improvement in robust fairness compared to existing methods.