Molecular property prediction plays a fundamental role in drug discovery to identify candidate molecules with target properties. However, molecular property prediction is essentially a few-shot problem which makes it hard to use regular machine learning models. In this paper, we propose Property-Aware Relation networks (PAR) to handle this problem. In comparison to existing works, we leverage the fact that both relevant substructures and relationships among molecules change across different molecular properties. We first introduce a property-aware embedding function to transform the generic molecular embeddings to substructure-aware space relevant to the target property. Further, we design an adaptive relation graph learning module to jointly estimate molecular relation graph and refine molecular embeddings w.r.t. the target property, such that the limited labels can be effectively propagated among similar molecules. We adopt a meta-learning strategy where the parameters are selectively updated within tasks in order to model generic and property-aware knowledge separately. Extensive experiments on benchmark molecular property prediction datasets show that PAR consistently outperforms existing methods and can obtain property-aware molecular embeddings and model molecular relation graph properly.
Meta-learning provides a popular and effective family of methods for data-efficient learning of new tasks. However, several important issues in meta-learning have proven hard to study thus far. For example, performance degrades in real-world settings where meta-learners must learn from a wide and potentially multi-modal distribution of training tasks; and when distribution shift exists between meta-train and meta-test task distributions. These issues are typically hard to study since the shape of task distributions, and shift between them are not straightforward to measure or control in standard benchmarks. We propose the channel coding problem as a benchmark for meta-learning. Channel coding is an important practical application where task distributions naturally arise, and fast adaptation to new tasks is practically valuable. We use our MetaCC benchmark to study several aspects of meta-learning, including the impact of task distribution breadth and shift, which can be controlled in the coding problem. Going forward, MetaCC provides a tool for the community to study the capabilities and limitations of meta-learning, and to drive research on practically robust and effective meta-learners.
The changing preferences of users towards items trigger the emergence of session-based recommender systems (SBRSs), which aim to model the dynamic preferences of users for next-item recommendations. However, most of the existing studies on SBRSs are based on long sessions only for recommendations, ignoring short sessions, though short sessions, in fact, account for a large proportion in most of the real-world datasets. As a result, the applicability of existing SBRSs solutions is greatly reduced. In a short session, quite limited contextual information is available, making the next-item recommendation very challenging. To this end, in this paper, inspired by the success of few-shot learning (FSL) in effectively learning a model with limited instances, we formulate the next-item recommendation as an FSL problem. Accordingly, following the basic idea of a representative approach for FSL, i.e., meta-learning, we devise an effective SBRS called INter-SEssion collaborative Recommender netTwork (INSERT) for next-item recommendations in short sessions. With the carefully devised local module and global module, INSERT is able to learn an optimal preference representation of the current user in a given short session. In particular, in the global module, a similar session retrieval network (SSRN) is designed to find out the sessions similar to the current short session from the historical sessions of both the current user and other users, respectively. The obtained similar sessions are then utilized to complement and optimize the preference representation learned from the current short session by the local module for more accurate next-item recommendations in this short session. Extensive experiments conducted on two real-world datasets demonstrate the superiority of our proposed INSERT over the state-of-the-art SBRSs when making next-item recommendations in short sessions.
Few-shot learning is a relatively new technique that specializes in problems where we have little amounts of data. The goal of these methods is to classify categories that have not been seen before with just a handful of samples. Recent approaches, such as metric learning, adopt the meta-learning strategy in which we have episodic tasks conformed by support (training) data and query (test) data. Metric learning methods have demonstrated that simple models can achieve good performance by learning a similarity function to compare the support and the query data. However, the feature space learned by a given metric learning approach may not exploit the information given by a specific few-shot task. In this work, we explore the use of dimension reduction techniques as a way to find task-significant features helping to make better predictions. We measure the performance of the reduced features by assigning a score based on the intra-class and inter-class distance, and selecting a feature reduction method in which instances of different classes are far away and instances of the same class are close. This module helps to improve the accuracy performance by allowing the similarity function, given by the metric learning method, to have more discriminative features for the classification. Our method outperforms the metric learning baselines in the miniImageNet dataset by around 2% in accuracy performance.
Exploration in reinforcement learning is a challenging problem: in the worst case, the agent must search for high-reward states that could be hidden anywhere in the state space. Can we define a more tractable class of RL problems, where the agent is provided with examples of successful outcomes? In this problem setting, the reward function can be obtained automatically by training a classifier to categorize states as successful or not. If trained properly, such a classifier can provide a well-shaped objective landscape that both promotes progress toward good states and provides a calibrated exploration bonus. In this work, we show that an uncertainty aware classifier can solve challenging reinforcement learning problems by both encouraging exploration and provided directed guidance towards positive outcomes. We propose a novel mechanism for obtaining these calibrated, uncertainty-aware classifiers based on an amortized technique for computing the normalized maximum likelihood (NML) distribution. To make this tractable, we propose a novel method for computing the NML distribution by using meta-learning. We show that the resulting algorithm has a number of intriguing connections to both count-based exploration methods and prior algorithms for learning reward functions, while also providing more effective guidance towards the goal. We demonstrate that our algorithm solves a number of challenging navigation and robotic manipulation tasks which prove difficult or impossible for prior methods.
Few-shot NLP research is highly active, yet conducted in disjoint research threads with evaluation suites that lack challenging-yet-realistic testing setups and fail to employ careful experimental design. Consequently, the community does not know which techniques perform best or even if they outperform simple baselines. In response, we formulate the FLEX Principles, a set of requirements and best practices for unified, rigorous, valid, and cost-sensitive few-shot NLP evaluation. These principles include Sample Size Design, a novel approach to benchmark design that optimizes statistical accuracy and precision while keeping evaluation costs manageable. Following the principles, we release the FLEX benchmark, which includes four few-shot transfer settings, zero-shot evaluation, and a public leaderboard that covers diverse NLP tasks. In addition, we present UniFew, a prompt-based model for few-shot learning that unifies pretraining and finetuning prompt formats, eschewing complex machinery of recent prompt-based approaches in adapting downstream task formats to language model pretraining objectives. We demonstrate that despite simplicity, UniFew achieves results competitive with both popular meta-learning and prompt-based approaches.
This paper proposes an extension of principal component analysis for Gaussian process (GP) posteriors, denoted by GP-PCA. Since GP-PCA estimates a low-dimensional space of GP posteriors, it can be used for meta-learning, which is a framework for improving the performance of target tasks by estimating a structure of a set of tasks. The issue is how to define a structure of a set of GPs with an infinite-dimensional parameter, such as coordinate system and a divergence. In this study, we reduce the infiniteness of GP to the finite-dimensional case under the information geometrical framework by considering a space of GP posteriors that have the same prior. In addition, we propose an approximation method of GP-PCA based on variational inference and demonstrate the effectiveness of GP-PCA as meta-learning through experiments.
A fundamental challenge for sequential recommenders is to capture the sequential patterns of users toward modeling how users transit among items. In many practical scenarios, however, there are a great number of cold-start users with only minimal logged interactions. As a result, existing sequential recommendation models will lose their predictive power due to the difficulties in learning sequential patterns over users with only limited interactions. In this work, we aim to improve sequential recommendation for cold-start users with a novel framework named MetaTL, which learns to model the transition patterns of users through meta-learning. Specifically, the proposed MetaTL: (i) formulates sequential recommendation for cold-start users as a few-shot learning problem; (ii) extracts the dynamic transition patterns among users with a translation-based architecture; and (iii) adopts meta transitional learning to enable fast learning for cold-start users with only limited interactions, leading to accurate inference of sequential interactions.
This paper is both an introduction and an invitation. It is an introduction to CARLE, a Life-like cellular automata simulator and reinforcement learning environment. It is also an invitation to Carle's Game, a challenge in open-ended machine exploration and creativity. Inducing machine agents to excel at creating interesting patterns across multiple cellular automata universes is a substantial challenge, and approaching this challenge is likely to require contributions from the fields of artificial life, AI, machine learning, and complexity, at multiple levels of interest. Carle's Game is based on machine agent interaction with CARLE, a Cellular Automata Reinforcement Learning Environment. CARLE is flexible, capable of simulating any of the 262,144 different rules defining Life-like cellular automaton universes. CARLE is also fast and can simulate automata universes at a rate of tens of thousands of steps per second through a combination of vectorization and GPU acceleration. Finally, CARLE is simple. Compared to high-fidelity physics simulators and video games designed for human players, CARLE's two-dimensional grid world offers a discrete, deterministic, and atomic universal playground, despite its complexity. In combination with CARLE, Carle's Game offers an initial set of agent policies, learning and meta-learning algorithms, and reward wrappers that can be tailored to encourage exploration or specific tasks.
We propose a meta-learning technique for offline discovery of physics-informed neural network (PINN) loss functions. We extend earlier works on meta-learning, and develop a gradient-based meta-learning algorithm for addressing diverse task distributions based on parametrized partial differential equations (PDEs) that are solved with PINNs. Furthermore, based on new theory we identify two desirable properties of meta-learned losses in PINN problems, which we enforce by proposing a new regularization method or using a specific parametrization of the loss function. In the computational examples, the meta-learned losses are employed at test time for addressing regression and PDE task distributions. Our results indicate that significant performance improvement can be achieved by using a shared-among-tasks offline-learned loss function even for out-of-distribution meta-testing. In this case, we solve for test tasks that do not belong to the task distribution used in meta-training, and we also employ PINN architectures that are different from the PINN architecture used in meta-training. To better understand the capabilities and limitations of the proposed method, we consider various parametrizations of the loss function and describe different algorithm design options and how they may affect meta-learning performance.
Shortage of annotated data is one of the greatest bottlenecks in biomedical image analysis. Meta learning studies how learning systems can increase in efficiency through experience and could thus evolve as an important concept to overcome data sparsity. However, the core capability of meta learning-based approaches is the identification of similar previous tasks given a new task - a challenge largely unexplored in the biomedical imaging domain. In this paper, we address the problem of quantifying task similarity with a concept that we refer to as task fingerprinting. The concept involves converting a given task, represented by imaging data and corresponding labels, to a fixed-length vector representation. In fingerprint space, different tasks can be directly compared irrespective of their data set sizes, types of labels or specific resolutions. An initial feasibility study in the field of surgical data science (SDS) with 26 classification tasks from various medical and non-medical domains suggests that task fingerprinting could be leveraged for both (1) selecting appropriate data sets for pretraining and (2) selecting appropriate architectures for a new task. Task fingerprinting could thus become an important tool for meta learning in SDS and other fields of biomedical image analysis.
Animals have remarkable abilities to adapt locomotion to different terrains and tasks. However, robots trained by means of reinforcement learning are typically able to solve only a single task and a transferred policy is usually inferior to that trained from scratch. In this work, we demonstrate that meta-reinforcement learning can be used to successfully train a robot capable to solve a wide range of locomotion tasks. The performance of the meta-trained robot is similar to that of a robot that is trained on a single task.
The label bias and selection bias are acknowledged as two reasons in data that will hinder the fairness of machine-learning outcomes. The label bias occurs when the labeling decision is disturbed by sensitive features, while the selection bias occurs when subjective bias exists during the data sampling. Even worse, models trained on such data can inherit or even intensify the discrimination. Most algorithmic fairness approaches perform an empirical risk minimization with predefined fairness constraints, which tends to trade-off accuracy for fairness. However, such methods would achieve the desired fairness level with the sacrifice of the benefits (receive positive outcomes) for individuals affected by the bias. Therefore, we propose a Bias-TolerantFAirRegularizedLoss (B-FARL), which tries to regain the benefits using data affected by label bias and selection bias. B-FARL takes the biased data as input, calls a model that approximates the one trained with fair but latent data, and thus prevents discrimination without constraints required. In addition, we show the effective components by decomposing B-FARL, and we utilize the meta-learning framework for the B-FARL optimization. The experimental results on real-world datasets show that our method is empirically effective in improving fairness towards the direction of true but latent labels.
Recent research has seen numerous supervised learning-based methods for 3D shape segmentation and remarkable performance has been achieved on various benchmark datasets. These supervised methods require a large amount of annotated data to train deep neural networks to ensure the generalization ability on the unseen test set. In this paper, we introduce a meta-learning-based method for few-shot 3D shape segmentation where only a few labeled samples are provided for the unseen classes. To achieve this, we treat the shape segmentation as a point labeling problem in the metric space. Specifically, we first design a meta-metric learner to transform input shapes into embedding space and our model learns to learn a proper metric space for each object class based on point embeddings. Then, for each class, we design a metric learner to extract part-specific prototype representations from a few support shapes and our model performs per-point segmentation over the query shapes by matching each point to its nearest prototype in the learned metric space. A metric-based loss function is used to dynamically modify distances between point embeddings thus maximizes in-part similarity while minimizing inter-part similarity. A dual segmentation branch is adopted to make full use of the support information and implicitly encourages consistency between the support and query prototypes. We demonstrate the superior performance of our proposed on the ShapeNet part dataset under the few-shot scenario, compared with well-established baseline and state-of-the-art semi-supervised methods.
Meta learning has attracted much attention recently in machine learning community. Contrary to conventional machine learning aiming to learn inherent prediction rules to predict labels for new query data, meta learning aims to learn the learning methodology for machine learning from observed tasks, so as to generalize to new query tasks by leveraging the meta-learned learning methodology. In this study, we interpret such learning methodology as learning an explicit hyper-parameter prediction function shared by all training tasks. Specifically, this function is represented as a parameterized function called meta-learner, mapping from a training/test task to its suitable hyper-parameter setting, extracted from a pre-specified function set called meta learning machine. Such setting guarantees that the meta-learned learning methodology is able to flexibly fit diverse query tasks, instead of only obtaining fixed hyper-parameters by many current meta learning methods, with less adaptability to query task's variations. Such understanding of meta learning also makes it easily succeed from traditional learning theory for analyzing its generalization bounds with general losses/tasks/models. The theory naturally leads to some feasible controlling strategies for ameliorating the quality of the extracted meta-learner, verified to be able to finely ameliorate its generalization capability in some typical meta learning applications, including few-shot regression, few-shot classification and domain generalization.
Meta-learning aims to learn a model that can handle multiple tasks generated from an unknown but shared distribution. However, typical meta-learning algorithms have assumed the tasks to be similar such that a single meta-learner is sufficient to aggregate the variations in all aspects. In addition, there has been less consideration on uncertainty when limited information is given as context. In this paper, we devise a novel meta-learning framework, called Meta-learning Amidst Heterogeneity and Ambiguity (MAHA), that outperforms previous works in terms of prediction based on its ability on task identification. By extensively conducting several experiments in regression and classification, we demonstrate the validity of our model, which turns out to be robust to both task heterogeneity and ambiguity.
Thompson sampling and other Bayesian sequential decision-making algorithms are among the most popular approaches to tackle explore/exploit trade-offs in (contextual) bandits. The choice of prior in these algorithms offers flexibility to encode domain knowledge but can also lead to poor performance when misspecified. In this paper, we demonstrate that performance degrades gracefully with misspecification. We prove that the expected reward accrued by Thompson sampling (TS) with a misspecified prior differs by at most $\tilde{\mathcal{O}}(H^2 \epsilon)$ from TS with a well specified prior, where $\epsilon$ is the total-variation distance between priors and $H$ is the learning horizon. Our bound does not require the prior to have any parametric form. For priors with bounded support, our bound is independent of the cardinality or structure of the action space, and we show that it is tight up to universal constants in the worst case. Building on our sensitivity analysis, we establish generic PAC guarantees for algorithms in the recently studied Bayesian meta-learning setting and derive corollaries for various families of priors. Our results generalize along two axes: (1) they apply to a broader family of Bayesian decision-making algorithms, including a Monte-Carlo implementation of the knowledge gradient algorithm (KG), and (2) they apply to Bayesian POMDPs, the most general Bayesian decision-making setting, encompassing contextual bandits as a special case. Through numerical simulations, we illustrate how prior misspecification and the deployment of one-step look-ahead (as in KG) can impact the convergence of meta-learning in multi-armed and contextual bandits with structured and correlated priors.
Meta learning approaches to few-shot classification are computationally efficient at test time, requiring just a few optimization steps or single forward pass to learn a new task, but they remain highly memory-intensive to train. This limitation arises because a task's entire support set, which can contain up to 1000 images, must be processed before an optimization step can be taken. Harnessing the performance gains offered by large images thus requires either parallelizing the meta-learner across multiple GPUs, which may not be available, or trade-offs between task and image size when memory constraints apply. We improve on both options by proposing LITE, a general and memory efficient episodic training scheme that enables meta-training on large tasks composed of large images on a single GPU. We achieve this by observing that the gradients for a task can be decomposed into a sum of gradients over the task's training images. This enables us to perform a forward pass on a task's entire training set but realize significant memory savings by back-propagating only a random subset of these images which we show is an unbiased approximation of the full gradient. We use LITE to train meta-learners and demonstrate new state-of-the-art accuracy on the real-world ORBIT benchmark and 3 of the 4 parts of the challenging VTAB+MD benchmark relative to leading meta-learners. LITE also enables meta-learners to be competitive with transfer learning approaches but at a fraction of the test-time computational cost, thus serving as a counterpoint to the recent narrative that transfer learning is all you need for few-shot classification.
Despite the success of recent Neural Architecture Search (NAS) methods on various tasks which have shown to output networks that largely outperform human-designed networks, conventional NAS methods have mostly tackled the optimization of searching for the network architecture for a single task (dataset), which does not generalize well across multiple tasks (datasets). Moreover, since such task-specific methods search for a neural architecture from scratch for every given task, they incur a large computational cost, which is problematic when the time and monetary budget are limited. In this paper, we propose an efficient NAS framework that is trained once on a database consisting of datasets and pretrained networks and can rapidly search for a neural architecture for a novel dataset. The proposed MetaD2A (Meta Dataset-to-Architecture) model can stochastically generate graphs (architectures) from a given set (dataset) via a cross-modal latent space learned with amortized meta-learning. Moreover, we also propose a meta-performance predictor to estimate and select the best architecture without direct training on target datasets. The experimental results demonstrate that our model meta-learned on subsets of ImageNet-1K and architectures from NAS-Bench 201 search space successfully generalizes to multiple unseen datasets including CIFAR-10 and CIFAR-100, with an average search time of 33 GPU seconds. Even under MobileNetV3 search space, MetaD2A is 5.5K times faster than NSGANetV2, a transferable NAS method, with comparable performance. We believe that the MetaD2A proposes a new research direction for rapid NAS as well as ways to utilize the knowledge from rich databases of datasets and architectures accumulated over the past years. Code is available at https://github.com/HayeonLee/MetaD2A.
The ratio of two probability densities, called a density-ratio, is a vital quantity in machine learning. In particular, a relative density-ratio, which is a bounded extension of the density-ratio, has received much attention due to its stability and has been used in various applications such as outlier detection and dataset comparison. Existing methods for (relative) density-ratio estimation (DRE) require many instances from both densities. However, sufficient instances are often unavailable in practice. In this paper, we propose a meta-learning method for relative DRE, which estimates the relative density-ratio from a few instances by using knowledge in related datasets. Specifically, given two datasets that consist of a few instances, our model extracts the datasets' information by using neural networks and uses it to obtain instance embeddings appropriate for the relative DRE. We model the relative density-ratio by a linear model on the embedded space, whose global optimum solution can be obtained as a closed-form solution. The closed-form solution enables fast and effective adaptation to a few instances, and its differentiability enables us to train our model such that the expected test error for relative DRE can be explicitly minimized after adapting to a few instances. We empirically demonstrate the effectiveness of the proposed method by using three problems: relative DRE, dataset comparison, and outlier detection.