Prototypical network (PN) is a simple yet effective few shot learning strategy. It is a metric-based meta-learning technique where classification is performed by computing Euclidean distances to prototypical representations of each class. Conventional PN attributes equal importance to all samples and generates prototypes by simply averaging the support sample embeddings belonging to each class. In this work, we propose a novel version of PN that attributes weights to support samples corresponding to their influence on the support sample distribution. Influence weights of samples are calculated based on maximum mean discrepancy (MMD) between the mean embeddings of sample distributions including and excluding the sample. Further, the influence factor of a sample is measured using MMD based on the shift in the distribution in the absence of that sample.
In this paper, we study the problem of one-shot skeleton-based action recognition, which poses unique challenges in learning transferable representation from base classes to novel classes, particularly for fine-grained actions. Existing meta-learning frameworks typically rely on the body-level representations in spatial dimension, which limits the generalisation to capture subtle visual differences in the fine-grained label space. To overcome the above limitation, we propose a part-aware prototypical representation for one-shot skeleton-based action recognition. Our method captures skeleton motion patterns at two distinctive spatial levels, one for global contexts among all body joints, referred to as body level, and the other attends to local spatial regions of body parts, referred to as the part level. We also devise a class-agnostic attention mechanism to highlight important parts for each action class. Specifically, we develop a part-aware prototypical graph network consisting of three modules: a cascaded embedding module for our dual-level modelling, an attention-based part fusion module to fuse parts and generate part-aware prototypes, and a matching module to perform classification with the part-aware representations. We demonstrate the effectiveness of our method on two public skeleton-based action recognition datasets: NTU RGB+D 120 and NW-UCLA.
We study the meta-learning for support (i.e. the set of non-zero entries) recovery in high-dimensional Principal Component Analysis. We reduce the sufficient sample complexity in a novel task with the information that is learned from auxiliary tasks. We assume each task to be a different random Principal Component (PC) matrix with a possibly different support and that the support union of the PC matrices is small. We then pool the data from all the tasks to execute an improper estimation of a single PC matrix by maximising the $l_1$-regularised predictive covariance to establish that with high probability the true support union can be recovered provided a sufficient number of tasks $m$ and a sufficient number of samples $ O\left(\frac{\log(p)}{m}\right)$ for each task, for $p$-dimensional vectors. Then, for a novel task, we prove that the maximisation of the $l_1$-regularised predictive covariance with the additional constraint that the support is a subset of the estimated support union could reduce the sufficient sample complexity of successful support recovery to $O(\log |J|)$, where $J$ is the support union recovered from the auxiliary tasks. Typically, $|J|$ would be much less than $p$ for sparse matrices. Finally, we demonstrate the validity of our experiments through numerical simulations.
We investigate the inverse problem for Partial Differential Equations (PDEs) in scenarios where the parameters of the given PDE dynamics may exhibit changepoints at random time. We employ Physics-Informed Neural Networks (PINNs) - universal approximators capable of estimating the solution of any physical law described by a system of PDEs, which serves as a regularization during neural network training, restricting the space of admissible solutions and enhancing function approximation accuracy. We demonstrate that when the system exhibits sudden changes in the PDE dynamics, this regularization is either insufficient to accurately estimate the true dynamics, or it may result in model miscalibration and failure. Consequently, we propose a PINNs extension using a Total-Variation penalty, which allows to accommodate multiple changepoints in the PDE dynamics and significantly improves function approximation. These changepoints can occur at random locations over time and are estimated concurrently with the solutions. Additionally, we introduce an online learning method for re-weighting loss function terms dynamically. Through empirical analysis using examples of various equations with parameter changes, we showcase the advantages of our proposed model. In the absence of changepoints, the model reverts to the original PINNs model. However, when changepoints are present, our approach yields superior parameter estimation, improved model fitting, and reduced training error compared to the original PINNs model.
Model-Agnostic Meta-Learning (MAML) is one of the most successful meta-learning techniques for few-shot learning. It uses gradient descent to learn commonalities between various tasks, enabling the model to learn the meta-initialization of its own parameters to quickly adapt to new tasks using a small amount of labeled training data. A key challenge to few-shot learning is task uncertainty. Although a strong prior can be obtained from meta-learning with a large number of tasks, a precision model of the new task cannot be guaranteed because the volume of the training dataset is normally too small. In this study, first,in the process of choosing initialization parameters, the new method is proposed for task-specific learner adaptively learn to select initialization parameters that minimize the loss of new tasks. Then, we propose two improved methods for the meta-loss part: Method 1 generates weights by comparing meta-loss differences to improve the accuracy when there are few classes, and Method 2 introduces the homoscedastic uncertainty of each task to weigh multiple losses based on the original gradient descent,as a way to enhance the generalization ability to novel classes while ensuring accuracy improvement. Compared with previous gradient-based meta-learning methods, our model achieves better performance in regression tasks and few-shot classification and improves the robustness of the model to the learning rate and query sets in the meta-test set.
Meta-learning is an effective method to handle imbalanced and noisy-label learning, but it depends on a validation set containing randomly selected, manually labelled and balanced distributed samples. The random selection and manual labelling and balancing of this validation set is not only sub-optimal for meta-learning, but it also scales poorly with the number of classes. Hence, recent meta-learning papers have proposed ad-hoc heuristics to automatically build and label this validation set, but these heuristics are still sub-optimal for meta-learning. In this paper, we analyse the meta-learning algorithm and propose new criteria to characterise the utility of the validation set, based on: 1) the informativeness of the validation set; 2) the class distribution balance of the set; and 3) the correctness of the labels of the set. Furthermore, we propose a new imbalanced noisy-label meta-learning (INOLML) algorithm that automatically builds a validation set by maximising its utility using the criteria above. Our method shows significant improvements over previous meta-learning approaches and sets the new state-of-the-art on several benchmarks.
Few-shot learning (FSL) is an emergent paradigm of learning that attempts to learn to reason with low sample complexity to mimic the way humans learn, generalise and extrapolate from only a few seen examples. While FSL attempts to mimic these human characteristics, fundamentally, the task of FSL as conventionally formulated using meta-learning with episodic-based training does not in actuality align with how humans acquire and reason with knowledge. FSL with episodic training, while only requires $K$ instances of each test class, still requires a large number of labelled training instances from disjoint classes. In this paper, we introduce the novel task of constrained few-shot learning (CFSL), a special case of FSL where $M$, the number of instances of each training class is constrained such that $M \leq K$ thus applying a similar restriction during FSL training and test. We propose a method for CFSL leveraging Cat2Vec using a novel categorical contrastive loss inspired by cognitive theories such as fuzzy trace theory and prototype theory.
We regularly consider answering counterfactual questions in practice, such as "Would people with diabetes take a turn for the better had they choose another medication?". Observational studies are growing in significance in answering such questions due to their widespread accumulation and comparatively easier acquisition than Randomized Control Trials (RCTs). Recently, some works have introduced representation learning and domain adaptation into counterfactual inference. However, most current works focus on the setting of binary treatments. None of them considers that different treatments' sample sizes are imbalanced, especially data examples in some treatment groups are relatively limited due to inherent user preference. In this paper, we design a new algorithmic framework for counterfactual inference, which brings an idea from Meta-learning for Estimating Individual Treatment Effects (MetaITE) to fill the above research gaps, especially considering multiple imbalanced treatments. Specifically, we regard data episodes among treatment groups in counterfactual inference as meta-learning tasks. We train a meta-learner from a set of source treatment groups with sufficient samples and update the model by gradient descent with limited samples in target treatment. Moreover, we introduce two complementary losses. One is the supervised loss on multiple source treatments. The other loss which aligns latent distributions among various treatment groups is proposed to reduce the discrepancy. We perform experiments on two real-world datasets to evaluate inference accuracy and generalization ability. Experimental results demonstrate that the model MetaITE matches/outperforms state-of-the-art methods.
Data imputation is an effective way to handle missing data, which is common in practical applications. In this study, we propose and test a novel data imputation process that achieve two important goals: (1) preserve the row-wise similarities among observations and column-wise contextual relationships among features in the feature matrix, and (2) tailor the imputation process to specific downstream label prediction task. The proposed imputation process uses Transformer network and graph structure learning to iteratively refine the contextual relationships among features and similarities among observations. Moreover, it uses a meta-learning framework to select features that are influential to the downstream prediction task of interest. We conduct experiments on real-world large data sets, and show that the proposed imputation process consistently improves imputation and label prediction performance over a variety of benchmark methods.
Visual (re)localization addresses the problem of estimating the 6-DoF (Degree of Freedom) camera pose of a query image captured in a known scene, which is a key building block of many computer vision and robotics applications. Recent advances in structure-based localization solve this problem by memorizing the mapping from image pixels to scene coordinates with neural networks to build 2D-3D correspondences for camera pose optimization. However, such memorization requires training by amounts of posed images in each scene, which is heavy and inefficient. On the contrary, few-shot images are usually sufficient to cover the main regions of a scene for a human operator to perform visual localization. In this paper, we propose a scene region classification approach to achieve fast and effective scene memorization with few-shot images. Our insight is leveraging a) pre-learned feature extractor, b) scene region classifier, and c) meta-learning strategy to accelerate training while mitigating overfitting. We evaluate our method on both indoor and outdoor benchmarks. The experiments validate the effectiveness of our method in the few-shot setting, and the training time is significantly reduced to only a few minutes. Code available at: \url{https://github.com/siyandong/SRC}
Graph Neural Networks (GNNs) have been successfully applied to many real-world static graphs. However, the success of static graphs has not fully translated to dynamic graphs due to the limitations in model design, evaluation settings, and training strategies. Concretely, existing dynamic GNNs do not incorporate state-of-the-art designs from static GNNs, which limits their performance. Current evaluation settings for dynamic GNNs do not fully reflect the evolving nature of dynamic graphs. Finally, commonly used training methods for dynamic GNNs are not scalable. Here we propose ROLAND, an effective graph representation learning framework for real-world dynamic graphs. At its core, the ROLAND framework can help researchers easily repurpose any static GNN to dynamic graphs. Our insight is to view the node embeddings at different GNN layers as hierarchical node states and then recurrently update them over time. We then introduce a live-update evaluation setting for dynamic graphs that mimics real-world use cases, where GNNs are making predictions and being updated on a rolling basis. Finally, we propose a scalable and efficient training approach for dynamic GNNs via incremental training and meta-learning. We conduct experiments over eight different dynamic graph datasets on future link prediction tasks. Models built using the ROLAND framework achieve on average 62.7% relative mean reciprocal rank (MRR) improvement over state-of-the-art baselines under the standard evaluation settings on three datasets. We find state-of-the-art baselines experience out-of-memory errors for larger datasets, while ROLAND can easily scale to dynamic graphs with 56 million edges. After re-implementing these baselines using the ROLAND training strategy, ROLAND models still achieve on average 15.5% relative MRR improvement over the baselines.
Few-shot object detection (FSOD) aims to classify and detect few images of novel categories. Existing meta-learning methods insufficiently exploit features between support and query images owing to structural limitations. We propose a hierarchical attention network with sequentially large receptive fields to fully exploit the query and support images. In addition, meta-learning does not distinguish the categories well because it determines whether the support and query images match. In other words, metric-based learning for classification is ineffective because it does not work directly. Thus, we propose a contrastive learning method called meta-contrastive learning, which directly helps achieve the purpose of the meta-learning strategy. Finally, we establish a new state-of-the-art network, by realizing significant margins. Our method brings 2.3, 1.0, 1.3, 3.4 and 2.4% AP improvements for 1-30 shots object detection on COCO dataset. Our code is available at: https://github.com/infinity7428/hANMCL
The cold-start problem is a long-standing challenge in recommender systems due to the lack of user-item interactions, which significantly hurts the recommendation effect over new users and items. Recently, meta-learning based methods attempt to learn globally shared prior knowledge across all users, which can be rapidly adapted to new users and items with very few interactions. Though with significant performance improvement, the globally shared parameter may lead to local optimum. Besides, they are oblivious to the inherent information and feature interactions existing in the new users and items, which are critical in cold-start scenarios. In this paper, we propose a Task aligned Meta-learning based Augmented Graph (TMAG) to address cold-start recommendation. Specifically, a fine-grained task aligned constructor is proposed to cluster similar users and divide tasks for meta-learning, enabling consistent optimization direction. Besides, an augmented graph neural network with two graph enhanced approaches is designed to alleviate data sparsity and capture the high-order user-item interactions. We validate our approach on three real-world datasets in various cold-start scenarios, showing the superiority of TMAG over state-of-the-art methods for cold-start recommendation.
The automated machine learning (AutoML) process can require searching through complex configuration spaces of not only machine learning (ML) components and their hyperparameters but also ways of composing them together, i.e. forming ML pipelines. Optimisation efficiency and the model accuracy attainable for a fixed time budget suffer if this pipeline configuration space is excessively large. A key research question is whether it is both possible and practical to preemptively avoid costly evaluations of poorly performing ML pipelines by leveraging their historical performance for various ML tasks, i.e. meta-knowledge. The previous experience comes in the form of classifier/regressor accuracy rankings derived from either (1) a substantial but non-exhaustive number of pipeline evaluations made during historical AutoML runs, i.e. 'opportunistic' meta-knowledge, or (2) comprehensive cross-validated evaluations of classifiers/regressors with default hyperparameters, i.e. 'systematic' meta-knowledge. Numerous experiments with the AutoWeka4MCPS package suggest that (1) opportunistic/systematic meta-knowledge can improve ML outcomes, typically in line with how relevant that meta-knowledge is, and (2) configuration-space culling is optimal when it is neither too conservative nor too radical. However, the utility and impact of meta-knowledge depend critically on numerous facets of its generation and exploitation, warranting extensive analysis; these are often overlooked/underappreciated within AutoML and meta-learning literature. In particular, we observe strong sensitivity to the `challenge' of a dataset, i.e. whether specificity in choosing a predictor leads to significantly better performance. Ultimately, identifying `difficult' datasets, thus defined, is crucial to both generating informative meta-knowledge bases and understanding optimal search-space reduction strategies.
In a K-best detector for multiple-input-multiple-output(MIMO) systems, the value of K needs to be sufficiently large to achieve near-maximum-likelihood (ML) performance. By treating K as a variable that can be adjusted according to a fitting function of some learnable coefficients, an intelligent MIMO detection network based on deep neural networks (DNN) is proposed to reduce complexity of the detection algorithm with little performance degradation. In particular, the proposed intelligent detection algorithm uses meta learning to learn the coefficients of the fitting function for K to circumvent the problem of learning K directly. The idea of network fusion is used to combine the learning results of the meta learning component networks. Simulation results show that the proposed scheme achieves near-ML detection performance while its complexity is close to that of linear detectors. Besides, it also exhibits strong ability of fast training.
In recent years, unmanned aerial vehicle (UAV) imaging is a suitable solution for real-time monitoring different vehicles on the urban scale. Real-time vehicle detection with the use of uncertainty estimation in deep meta-learning for the portable platforms (e.g., UAV) potentially improves video understanding in real-world applications with a small training dataset, while many vehicle monitoring approaches appear to understand single-time detection with a big training dataset. The purpose of real-time vehicle detection from oblique UAV images is to locate the vehicle on the time series UAV images by using semantic segmentation. Real-time vehicle detection is more difficult due to the variety of depth and scale vehicles in oblique view UAV images. Motivated by these facts, in this manuscript, we consider the problem of real-time vehicle detection for oblique UAV images based on a small training dataset and deep meta-learning. The proposed architecture, called SA-Net.v2, is a developed method based on the SA-CNN for real-time vehicle detection by reformulating the squeeze-and-attention mechanism. The SA-Net.v2 is composed of two components, including the squeeze-and-attention function that extracts the high-level feature based on a small training dataset, and the gated CNN. For the real-time vehicle detection scenario, we test our model on the UAVid dataset. UAVid is a time series oblique UAV images dataset consisting of 30 video sequences. We examine the proposed method's applicability for stand real-time vehicle detection in urban environments using time series UAV images. The experiments show that the SA-Net.v2 achieves promising performance in time series oblique UAV images.
Generalization in reinforcement learning (RL) is of importance for real deployment of RL algorithms. Various schemes are proposed to address the generalization issues, including transfer learning, multi-task learning and meta learning, as well as the robust and adversarial reinforcement learning. However, there is not a unified formulation of the various schemes, as well as the comprehensive comparisons of methods across different schemes. In this work, we propose a game-theoretic framework for the generalization in reinforcement learning, named GiRL, where an RL agent is trained against an adversary over a set of tasks, where the adversary can manipulate the distributions over tasks within a given threshold. With different configurations, GiRL can reduce the various schemes mentioned above. To solve GiRL, we adapt the widely-used method in game theory, policy space response oracle (PSRO) with the following three important modifications: i) we use model-agnostic meta learning (MAML) as the best-response oracle, ii) we propose a modified projected replicated dynamics, i.e., R-PRD, which ensures the computed meta-strategy of the adversary fall in the threshold, and iii) we also propose a protocol for the few-shot learning of the multiple strategies during testing. Extensive experiments on MuJoCo environments demonstrate that our proposed methods can outperform existing baselines, e.g., MAML.
Reinforcement learning agents must generalize beyond their training experience. Prior work has focused mostly on identical training and evaluation environments. Starting from the recently introduced Crafter benchmark, a 2D open world survival game, we introduce a new set of environments suitable for evaluating some agent's ability to generalize on previously unseen (numbers of) objects and to adapt quickly (meta-learning). In Crafter, the agents are evaluated by the number of unlocked achievements (such as collecting resources) when trained for 1M steps. We show that current agents struggle to generalize, and introduce novel object-centric agents that improve over strong baselines. We also provide critical insights of general interest for future work on Crafter through several experiments. We show that careful hyper-parameter tuning improves the PPO baseline agent by a large margin and that even feedforward agents can unlock almost all achievements by relying on the inventory display. We achieve new state-of-the-art performance on the original Crafter environment. Additionally, when trained beyond 1M steps, our tuned agents can unlock almost all achievements. We show that the recurrent PPO agents improve over feedforward ones, even with the inventory information removed. We introduce CrafterOOD, a set of 15 new environments that evaluate OOD generalization. On CrafterOOD, we show that the current agents fail to generalize, whereas our novel object-centric agents achieve state-of-the-art OOD generalization while also being interpretable. Our code is public.
Transfer learning, multi-task learning, and meta-learning are well-studied topics concerned with the generalization of knowledge across learning tasks and are closely related to general intelligence. But, the formal, general systems differences between them are underexplored in the literature. This lack of systems-level formalism leads to difficulties in coordinating related, inter-disciplinary engineering efforts. This manuscript formalizes transfer learning, multi-task learning, and meta-learning as abstract learning systems, consistent with the formal-minimalist abstract systems theory of Mesarovic and Takahara. Moreover, it uses the presented formalism to relate the three concepts of learning in terms of composition, hierarchy, and structural homomorphism. Findings are readily depicted in terms of input-output systems, highlighting the ease of delineating formal, general systems differences between transfer, multi-task, and meta-learning.
Meta-learning from learning curves is an important yet often neglected research area in the Machine Learning community. We introduce a series of Reinforcement Learning-based meta-learning challenges, in which an agent searches for the best suited algorithm for a given dataset, based on feedback of learning curves from the environment. The first round attracted participants both from academia and industry. This paper analyzes the results of the first round (accepted to the competition program of WCCI 2022), to draw insights into what makes a meta-learner successful at learning from learning curves. With the lessons learned from the first round and the feedback from the participants, we have designed the second round of our challenge with a new protocol and a new meta-dataset. The second round of our challenge is accepted at the AutoML-Conf 2022 and currently ongoing .