Dynamic ensemble selection systems work by estimating the level of competence of each classifier from a pool of classifiers. Only the most competent ones are selected to classify a given test sample. This is achieved by defining a criterion to measure the level of competence of a base classifier, such as, its accuracy in local regions of the feature space around the query instance. However, using only one criterion about the behavior of a base classifier is not sufficient to accurately estimate its level of competence. In this paper, we present a novel dynamic ensemble selection framework using meta-learning. We propose five distinct sets of meta-features, each one corresponding to a different criterion to measure the level of competence of a classifier for the classification of input samples. The meta-features are extracted from the training data and used to train a meta-classifier to predict whether or not a base classifier is competent enough to classify an input instance. During the generalization phase, the meta-features are extracted from the query instance and passed down as input to the meta-classifier. The meta-classifier estimates, whether a base classifier is competent enough to be added to the ensemble. Experiments are conducted over several small sample size classification problems, i.e., problems with a high degree of uncertainty due to the lack of training data. Experimental results show the proposed meta-learning framework greatly improves classification accuracy when compared against current state-of-the-art dynamic ensemble selection techniques.
Resource-efficient convolution neural networks enable not only the intelligence on edge devices but also opportunities in system-level optimization such as scheduling. In this work, we aim to improve the performance of resource-constrained filter pruning by merging two sub-problems commonly considered, i.e., (i) how many filters to prune for each layer and (ii) which filters to prune given a per-layer pruning budget, into a global filter ranking problem. Our framework entails a novel algorithm, dubbed layer-compensated pruning, where meta-learning is involved to determine better solutions. We show empirically that the proposed algorithm is superior to prior art in both effectiveness and efficiency. Specifically, we reduce the accuracy gap between the pruned and original networks from 0.9% to 0.7% with 8x reduction in time needed for meta-learning, i.e., from 1 hour down to 7 minutes. To this end, we demonstrate the effectiveness of our algorithm using ResNet and MobileNetV2 networks under CIFAR-10, ImageNet, and Bird-200 datasets.
We present a meta-learning approach for adaptive text-to-speech (TTS) with few data. During training, we learn a multi-speaker model using a shared conditional WaveNet core and independent learned embeddings for each speaker. The aim of training is not to produce a neural network with fixed weights, which is then deployed as a TTS system. Instead, the aim is to produce a network that requires few data at deployment time to rapidly adapt to new speakers. We introduce and benchmark three strategies: (i) learning the speaker embedding while keeping the WaveNet core fixed, (ii) fine-tuning the entire architecture with stochastic gradient descent, and (iii) predicting the speaker embedding with a trained neural network encoder. The experiments show that these approaches are successful at adapting the multi-speaker neural network to new speakers, obtaining state-of-the-art results in both sample naturalness and voice similarity with merely a few minutes of audio data from new speakers.
Transfer-learning and meta-learning are two effective methods to apply knowledge learned from large data sources to new tasks. In few-class, few-shot target task settings (i.e. when there are only a few classes and training examples available in the target task), meta-learning approaches that optimize for future task learning have outperformed the typical transfer approach of initializing model weights from a pre-trained starting point. But as we experimentally show, meta-learning algorithms that work well in the few-class setting do not generalize well in many-shot and many-class cases. In this paper, we propose a joint training approach that combines both transfer-learning and meta-learning. Benefiting from the advantages of each, our method obtains improved generalization performance on unseen target tasks in both few- and many-class and few- and many-shot scenarios.
Multi-object manipulation problems in continuous state and action spaces can be solved by planners that search over sampled values for the continuous parameters of operators. The efficiency of these planners depends critically on the effectiveness of the samplers used, but effective sampling in turn depends on details of the robot, environment, and task. Our strategy is to learn functions called "specializers" that generate values for continuous operator parameters, given a state description and values for the discrete parameters. Rather than trying to learn a single specializer for each operator from large amounts of data on a single task, we take a modular meta-learning approach. We train on multiple tasks and learn a variety of specializers that, on a new task, can be quickly adapted using relatively little data -- thus, our system "learns quickly to plan quickly" using these specializers. We validate our approach experimentally in simulated 3D pick-and-place tasks with continuous state and action spaces. Visit http://tinyurl.com/chitnis-icra-19 for a supplementary video.
Classic supervised learning makes the closed-world assumption, meaning that classes seen in testing must have been seen in training. However, in the dynamic world, new or unseen class examples may appear constantly. A model working in such an environment must be able to reject unseen classes (not seen or used in training). If enough data is collected for the unseen classes, the system should incrementally learn to accept/classify them. This learning paradigm is called open-world learning (OWL). Existing OWL methods all need some form of re-training to accept or include the new classes in the overall model. In this paper, we propose a meta-learning approach to the problem. Its key novelty is that it only needs to train a meta-classifier, which can then continually accept new classes when they have enough labeled data for the meta-classifier to use, and also detect/reject future unseen classes. No re-training of the meta-classifier or a new overall classifier covering all old and new classes is needed. In testing, the method only uses the examples of the seen classes (including the newly added classes) on-the-fly for classification and rejection. Experimental results demonstrate the effectiveness of the new approach.
The design of neural network architectures is an important component for achieving state-of-the-art performance with machine learning systems across a broad array of tasks. Much work has endeavored to design and build architectures automatically through clever construction of a search space paired with simple learning algorithms. Recent progress has demonstrated that such meta-learning methods may exceed scalable human-invented architectures on image classification tasks. An open question is the degree to which such methods may generalize to new domains. In this work we explore the construction of meta-learning techniques for dense image prediction focused on the tasks of scene parsing, person-part segmentation, and semantic image segmentation. Constructing viable search spaces in this domain is challenging because of the multi-scale representation of visual information and the necessity to operate on high resolution imagery. Based on a survey of techniques in dense image prediction, we construct a recursive search space and demonstrate that even with efficient random search, we can identify architectures that outperform human-invented architectures and achieve state-of-the-art performance on three dense prediction tasks including 82.7\% on Cityscapes (street scene parsing), 71.3\% on PASCAL-Person-Part (person-part segmentation), and 87.9\% on PASCAL VOC 2012 (semantic image segmentation). Additionally, the resulting architecture is more computationally efficient, requiring half the parameters and half the computational cost as previous state of the art systems.
While recent progress has spawned very powerful machine learning systems, those agents remain extremely specialized and fail to transfer the knowledge they gain to similar yet unseen tasks. In this paper, we study a simple reinforcement learning problem and focus on learning policies that encode the proper invariances for generalization to different settings. We evaluate three potential methods for policy generalization: data augmentation, meta-learning and adversarial training. We find our data augmentation method to be effective, and study the potential of meta-learning and adversarial learning as alternative task-agnostic approaches.
Effective feature representation is key to the predictive performance of any algorithm. This paper introduces a meta-procedure, called Non-Euclidean Upgrading (NEU), which learns feature maps that are expressive enough to embed the universal approximation property (UAP) into most model classes while only outputting feature maps that preserve any model class's UAP. We show that NEU can learn any feature map with these two properties if that feature map is asymptotically deformable into the identity. We also find that the feature-representations learned by NEU are always submanifolds of the feature space. NEU's properties are derived from a new deep neural model that is universal amongst all orientation-preserving homeomorphisms on the input space. We derive qualitative and quantitative approximation guarantees for this architecture. We quantify the number of parameters required for this new architecture to memorize any set of input-output pairs while simultaneously fixing every point of the input space lying outside some compact set, and we quantify the size of this set as a function of our model's depth. Moreover, we show that no deep feed-forward network with commonly used activation function has all these properties. NEU's performance is evaluated against competing machine learning methods on various regression and dimension reduction tasks both with financial and simulated data.
Meta-learning is increasingly used to support the recommendation of machine learning algorithms and their configurations. Such recommendations are made based on meta-data, consisting of performance evaluations of algorithms on prior datasets, as well as characterizations of these datasets. These characterizations, also called meta-features, describe properties of the data which are predictive for the performance of machine learning algorithms trained on them. Unfortunately, despite being used in a large number of studies, meta-features are not uniformly described, organized and computed, making many empirical studies irreproducible and hard to compare. This paper aims to deal with this by systematizing and standardizing data characterization measures for classification datasets used in meta-learning. Moreover, it presents MFE, a new tool for extracting meta-features from datasets and identifying more subtle reproducibility issues in the literature, proposing guidelines for data characterization that strengthen reproducible empirical research in meta-learning.
The performance of automatic speech recognition systems can be improved by adapting an acoustic model to compensate for the mismatch between training and testing conditions, for example by adapting to unseen speakers. The success of speaker adaptation methods relies on selecting weights that are suitable for adaptation and using good adaptation schedules to update these weights in order not to overfit to the adaptation data. In this paper we investigate a principled way of adapting all the weights of the acoustic model using a meta-learning. We show that the meta-learner can learn to perform supervised and unsupervised speaker adaptation and that it outperforms a strong baseline adapting LHUC parameters when adapting a DNN AM with 1.5M parameters. We also report initial experiments on adapting TDNN AMs, where the meta-learner achieves comparable performance with LHUC.
In this paper, we propose to extend the recently introduced model-agnostic meta-learning algorithm (MAML) for low-resource neural machine translation (NMT). We frame low-resource translation as a meta-learning problem, and we learn to adapt to low-resource languages based on multilingual high-resource language tasks. We use the universal lexical representation~\citep{gu2018universal} to overcome the input-output mismatch across different languages. We evaluate the proposed meta-learning strategy using eighteen European languages (Bg, Cs, Da, De, El, Es, Et, Fr, Hu, It, Lt, Nl, Pl, Pt, Sk, Sl, Sv and Ru) as source tasks and five diverse languages (Ro, Lv, Fi, Tr and Ko) as target tasks. We show that the proposed approach significantly outperforms the multilingual, transfer learning based approach~\citep{zoph2016transfer} and enables us to train a competitive NMT system with only a fraction of training examples. For instance, the proposed approach can achieve as high as 22.04 BLEU on Romanian-English WMT'16 by seeing only 16,000 translated words (~600 parallel sentences).
Deep Learning shows very good performance when trained on large labeled data sets. The problem of training a deep net on a few or one sample per class requires a different learning approach which can generalize to unseen classes using only a few representatives of these classes. This problem has previously been approached by meta-learning. Here we propose a novel meta-learner which shows state-of-the-art performance on common benchmarks for one/few shot classification. Our model features three novel components: First is a feed-forward embedding that takes random class support samples (after a customary CNN embedding) and transfers them to a better class representation in terms of a classification problem. Second is a novel attention mechanism, inspired by competitive learning, which causes class representatives to compete with each other to become a temporary class prototype with respect to the query point. This mechanism allows switching between representatives depending on the position of the query point. Once a prototype is chosen for each class, the predicated label is computed using a simple attention mechanism over prototypes of all considered classes. The third feature is the ability of our meta-learner to incorporate deeper CNN embedding, enabling larger capacity. Finally, to ease the training procedure and reduce overfitting, we averages the top $t$ models (evaluated on the validation) over the optimization trajectory. We show that this approach can be viewed as an approximation to an ensemble, which saves the factor of $t$ in training and test times and the factor of of $t$ in the storage of the final model.
Ensembling word embeddings to improve distributed word representations has shown good success for natural language processing tasks in recent years. These approaches either carry out straightforward mathematical operations over a set of vectors or use unsupervised learning to find a lower-dimensional representation. This work compares meta-embeddings trained for different losses, namely loss functions that account for angular distance between the reconstructed embedding and the target and those that account normalized distances based on the vector length. We argue that meta-embeddings are better to treat the ensemble set equally in unsupervised learning as the respective quality of each embedding is unknown for upstream tasks prior to meta-embedding. We show that normalization methods that account for this such as cosine and KL-divergence objectives outperform meta-embedding trained on standard $\ell_1$ and $\ell_2$ loss on \textit{defacto} word similarity and relatedness datasets and find it outperforms existing meta-learning strategies.
In this paper we present the methods of our submission to the ISIC 2018 challenge for skin lesion diagnosis (Task 3). The dataset consists of 10000 images with seven image-level classes to be distinguished by an automated algorithm. We employ an ensemble of convolutional neural networks for this task. In particular, we fine-tune pretrained state-of-the-art deep learning models such as Densenet, SENet and ResNeXt. We identify heavy class imbalance as a key problem for this challenge and consider multiple balancing approaches such as loss weighting and balanced batch sampling. Another important feature of our pipeline is the use of a vast amount of unscaled crops for evaluation. Last, we consider meta learning approaches for the final predictions. Our team placed second at the challenge while being the best approach using only publicly available data.
Fine-grained visual categorization (FGVC) is challenging due in part to the fact that it is often difficult to acquire an enough number of training samples. To employ large models for FGVC without suffering from overfitting, existing methods usually adopt a strategy of pre-training the models using a rich set of auxiliary data, followed by fine-tuning on the target FGVC task. However, the objective of pre-training does not take the target task into account, and consequently such obtained models are suboptimal for fine-tuning. To address this issue, we propose in this paper a new deep FGVC model termed MetaFGNet. Training of MetaFGNet is based on a novel regularized meta-learning objective, which aims to guide the learning of network parameters so that they are optimal for adapting to the target FGVC task. Based on MetaFGNet, we also propose a simple yet effective scheme for selecting more useful samples from the auxiliary data. Experiments on benchmark FGVC datasets show the efficacy of our proposed method.
Compared to humans, machine learning models generally require significantly more training examples and fail to extrapolate from experience to solve previously unseen challenges. To help close this performance gap, we augment single-task neural networks with a meta-recognition model which learns a succinct model code via its autoencoder structure, using just a few informative examples. The model code is then employed by a meta-generative model to construct parameters for the task-specific model. We demonstrate that for previously unseen tasks, without additional training, this Meta-Learning Autoencoder (MeLA) framework can build models that closely match the true underlying models, with loss significantly lower than given by fine-tuned baseline networks, and performance that compares favorably with state-of-the-art meta-learning algorithms. MeLA also adds the ability to identify influential training examples and predict which additional data will be most valuable to acquire to improve model prediction.
Gaussian Process (GP) regression has seen widespread use in robotics due to its generality, simplicity of use, and the utility of Bayesian predictions. The predominant implementation of GP regression is a nonparameteric kernel-based approach, as it enables fitting of arbitrary nonlinear functions. However, this approach suffers from two main drawbacks: (1) it is computationally inefficient, as computation scales poorly with the number of samples; and (2) it can be data inefficient, as encoding prior knowledge that can aid the model through the choice of kernel and associated hyperparameters is often challenging and unintuitive. In this work, we propose ALPaCA, an algorithm for efficient Bayesian regression which addresses these issues. ALPaCA uses a dataset of sample functions to learn a domain-specific, finite-dimensional feature encoding, as well as a prior over the associated weights, such that Bayesian linear regression in this feature space yields accurate online predictions of the posterior predictive density. These features are neural networks, which are trained via a meta-learning (or "learning-to-learn") approach. ALPaCA extracts all prior information directly from the dataset, rather than restricting prior information to the choice of kernel hyperparameters. Furthermore, by operating in the weight space, it substantially reduces sample complexity. We investigate the performance of ALPaCA on two simple regression problems, two simulated robotic systems, and on a lane-change driving task performed by humans. We find our approach outperforms kernel-based GP regression, as well as state of the art meta-learning approaches, thereby providing a promising plug-in tool for many regression tasks in robotics where scalability and data-efficiency are important.
The main contribution of this paper is to introduce and describe a new recommender-systems dataset (RARD II). It is based on data from Mr. DLib, a recommender-system as-a-service in the digital library and reference-management-software domain. As such, RARD II complements datasets from other domains such as books, movies, and music. The dataset encompasses 94m recommendations, delivered in the two years from September 2016 to September 2018. The dataset covers an item-space of 24m unique items. RARD II provides a range of rich recommendation data, beyond conventional ratings. For example, in addition to the usual (implicit) ratings matrices, RARD II includes the original recommendation logs, which provide a unique insight into many aspects of the algorithms that generated the recommendations. The logs enable researchers to conduct various analyses about a real-world recommender system. This includes the evaluation of meta-learning approaches for predicting algorithm performance. In this paper, we summarise the key features of this dataset release, describe how it was generated and discuss some of its unique features. Compared to its predecessor RARD, RARD II contains 64% more recommendations, 187% more features (algorithms, parameters, and statistics), 50% more clicks, 140% more documents, and one additional service partner (JabRef).
This work summarizes our submission for the Task 3: Disease Classification of ISIC 2018 challenge in Skin Lesion Analysis Towards Melanoma Detection. We use a novel deep neural network (DNN) ensemble architecture introduced by us that can effectively classify skin lesions by using data-augmentation and bagging to address paucity of data and prevent over-fitting. The ensemble is composed of two DNN architectures: Inception-v4 and Inception-Resnet-v2. The DNN architectures are combined in to an ensemble by using a $1\times1$ convolution for fusion in a meta-learning layer.