Few-shot Learning Tasks Research Articles

Learning to Class-Adaptively Manipulate Embeddings for Few-Shot Learning

In few-shot learning (FSL), meta-learning approach (MLA) mainly focuses on learning transferable knowledge from plenty of auxiliary FSL tasks to facilitate fast generalization to a new task. For a given FSL task, due to the inter-class distribution discrepancy, each class necessitates a specific embedding (i.e., a mapping function) to map samples into an ideal semantic space where samples from this class can be well separately from other classes. Moreover, these embeddings may vary with different tasks. Hence, one crucial knowledge for MLA is how to separately construct optimal embeddings for each class based on a few training samples given in a FSL task. However, most existing MLAs rarely consider this and thus show limited generalization capacity. To mitigate this problem, instead of directly construct class-adaptive embeddings, we present a new MLA that aims at learning to class-adaptively manipulate the features of samples for accurate classification in a new FSL task. In a specific, for a new FSL task, the proposed MLA first learns to generate some class-specific weights based on training samples via exploiting the inter-class distribution discrepancy between this class and the others. Then, the generated weights are utilized to compute the Hadamard product of features produced by a task-agnostic embedding module. By doing this, the proposed MLA can dynamically enhance or depress some specific semantic dimensions of sample features depending on the distribution of each class for accurate classification, and thus equals to constructing class-adaptive embeddings for each class but in a simpler way which can appropriately avoid over-fitting and is scalable to cases with extensive classes. To show its efficacy, we test the proposed MLA on four benchmark FSL datasets under various settings and report superior performance over existing state-of-the-arts.

Prompt-Based Graph Convolution Adversarial Meta-Learning for Few-Shot Text Classification

Deep learning techniques have demonstrated significant advancements in the task of text classification. Regrettably, the majority of these techniques necessitate a substantial corpus of annotated data to achieve optimal performance. Meta-learning has yielded intriguing outcomes in few-shot learning tasks, showcasing its potential in advancing the field. However, the current meta-learning methodologies are susceptible to overfitting due to the mismatch between a small number of samples and the complexity of the model. To mitigate this concern, we propose a Prompt-based Graph Convolutional Adversarial (PGCA) meta-learning framework, aiming to improve the adaptability of complex models in a few-shot scenario. Firstly, leveraging prompt learning, we generate embedding representations that bridge the downstream tasks. Then, we design a meta-knowledge extractor based on a graph convolutional neural network (GCN) to capture inter-class dependencies through instance-level interactions. We also integrate the adversarial network architecture into a meta-learning framework to extend sample diversity through adversarial training and improve the ability of the model to adapt to new tasks. Specifically, we mitigate the impact of extreme samples by introducing external knowledge to construct a list of class prototype extensions. Finally, we conduct a series of experiments on four public datasets to demonstrate the effectiveness of our proposed method.

Amortized Bayesian Meta-Learning with Accelerated Gradient Descent Steps

Recent meta-learning models often learn priors from observed tasks using a network optimized via stochastic gradient descent (SGD), which usually takes more training steps to convergence. In this paper, we propose an accelerated Bayesian meta-learning structure with a stochastic inference network (ABML-SIN). The proposed model aims to solve the training procedure of Bayesian meta-learning to improve the training speed and efficiency. Current approaches of meta-learning hardly converge within a few descent steps, owing to the small number of training samples. Therefore, we introduce an accelerated gradient descent learning network based on teacher–student architecture to learn the meta-latent variable θt for task t. With this amortized fast inference network, the meta-learner is able to learn the task-specific latent θt within a few training steps; thus, it improves the learning speed of the meta-learner. To refine the latent variables generated from the transductive amortization network of the meta-learner, SIN—followed by a conventional SGD-optimized network—is introduced as the student–teacher network to online-update the parameters. SIN extracts the local latent variables and accelerates the convergence of the meta-learning network. Our experiments on simulation data demonstrate that the proposed method provides generalization and scalability on unseen samples, and produces competitive/superior uncertainty estimations on few-shot learning tasks on two widely adopted 2D datasets with fewer training epochs compared to the state-of-the-art meta-learning approaches. Furthermore, the parameters generated by SIN act as perturbations on latent weights, enhancing the probability of accelerating the training efficiency of the meta-learner. Extensive qualitative experiments show that our method performs well across different meta-learning tasks in both simulated and real-world circumstances.

Can we improve meta-learning model in few-shot learning by aligning data distributions?

Meta-learning becomes a promising way to solve the few-shot learning problem in recent research. This paradigm mainly relies on hierarchical architecture and episodic training to achieve good generalization on the new learning task. However, as we know, support and query sets in each training task in episodic training only contain a few data points. Too limited training data can lead to data discrepancy between the training and test data which further hurts the performance of learning models in classical supervised and semi-supervised learning. Hence, there is an interesting question: Can we also improve the meta-learning model in few-shot learning by aligning the data distributions of support and query sets? In this work, we try to answer this question. Firstly, we give a theoretical analysis of the generalization ability of meta-learning to show that eliminating the data-distribution discrepancy between the support and query sets in each training task can improve the meta-learning model in few-shot learning. Next, we design an experiment to further verify our analysis. The experimental results show that aligning the data distributions is crucial and even can help the learning model trained by the classical batch-based training achieve better performance on unseen few-shot learning tasks. Furthermore, we also propose two simple meta-batch normalization methods following our result which can be integrated into recent meta-learning methods to improve their performance in few-shot learning.

PatchMix Augmentation to Identify Causal Features in Few-shot Learning.

The task of Few-shot learning (FSL) aims to transfer the knowledge learned from base categories with sufficient labelled data to novel categories with scarce known information. It is currently an important research question and has great practical values in the real-world applications. Despite extensive previous efforts are made on few-shot learning tasks, we emphasize that most existing methods did not take into account the distributional shift caused by sample selection bias in the FSL scenario. Such a selection bias can induce spurious correlation between the semantic causal features, that are causally and semantically related to the class label, and the other non-causal features. Critically, the former ones should be invariant across changes in distributions, highly related to the classes of interest, and thus well generalizable to novel classes, while the latter ones are not stable to changes in the distribution. To resolve this problem, we propose a novel data augmentation strategy dubbed as PatchMix that can break this spurious dependency by replacing the patch-level information and supervision of the query images with random gallery images from different classes from the query ones. We theoretically show that such an augmentation mechanism, different from existing ones, is able to identify the causal features. To further make these features to be discriminative enough for classification, we propose Correlation-guided Reconstruction (CGR) and Hardness-Aware module for instance discrimination and easier discrimination between similar classes. Moreover, such a framework can be adapted to the unsupervised FSL scenario. The utility of our method is demonstrated on the state-of-the-art results consistently achieved on several benchmarks including miniImageNet, tieredImageNet, CIFAR-FS , CUB, Cars, Places and Plantae, in all settings of single-domain, cross-domain and unsupervised FSL. By studying the intra-variance property of learned features and visualizing the learned features, we further quantitatively and qualitatively show that such a promising result is due to the effectiveness in learning causal features.

Bidirectional meta-Kronecker factored optimizer and Hausdorff distance loss for few-shot medical image segmentation

To increase the accuracy of medical image analysis using supervised learning-based AI technology, a large amount of accurately labeled training data is required. However, the supervised learning approach may not be applicable to real-world medical imaging due to the lack of labeled data, the privacy of patients, and the cost of specialized knowledge. To handle these issues, we utilized Kronecker-factored decomposition, which enhances both computational efficiency and stability of the learning process. We combined this approach with a model-agnostic meta-learning framework for the parameter optimization. Based on this method, we present a bidirectional meta-Kronecker factored optimizer (BM-KFO) framework to quickly optimize semantic segmentation tasks using just a few magnetic resonance imaging (MRI) images as input. This model-agnostic approach can be implemented without altering network components and is capable of learning the learning process and meta-initial points while training on previously unseen data. We also incorporated a combination of average Hausdorff distance loss (AHD-loss) and cross-entropy loss into our objective function to specifically target the morphology of organs or lesions in medical images. Through evaluation of the proposed method on the abdominal MRI dataset, we obtained an average performance of 78.07% in setting 1 and 79.85% in setting 2. Our experiments demonstrate that BM-KFO with AHD-loss is suitable for general medical image segmentation applications and achieves superior performance compared to the baseline method in few-shot learning tasks. In order to replicate the proposed method, we have shared our code on GitHub. The corresponding URL can be found: https://github.com/YeongjoonKim/BMKFO.git.

Multiscale Local Feature Fusion: Marine Microalgae Classification for Few-Shot Learning

In the marine ecological environment, marine microalgae is an important photosynthetic autotrophic organism, which can carry out photosynthesis and absorb carbon dioxide. With the increasingly serious eutrophication of the water body, under certain environmental factors, the rapid propagation of some algae in the water body gradually forms a harmful bloom, which damages the water environment. Therefore, how to identify the beneficial algae and harmful algae quickly and accurately has gradually become the key to solve the problem. There are more than 30,000 species of marine microalgae in the world, and the sample data are few and the characteristics are not obvious. Many of the algae are similar in shape and difficult to distinguish. The few-shot learning task is very challenging. By training very few labeled samples, the deep learning model has excellent recognition ability. Meanwhile, the few-shot classification method based on metric learning has attracted considerable attention. In this paper, in order to make full use of image features and improve the generalization ability of the model, a multi-scale local feature fusion algorithm was proposed to classify marine microalgae with few shots. First, the input image is gridded and multiscale processed, and then it is sent to the CTM category traversal module for feature extraction to obtain local features. A local feature fusion module based on the SE-NET self-attention mechanism is designed to obtain local enhanced features containing global information to improve the generalization ability of the model. Classification is realized by calculating the distance between the sample feature vector of the query set and the prototype of the support set. Under the settings of 5-way 1-shot and 5-way 5-shot, the classification accuracy of the proposed method is improved by 6.08% and 5.5%, respectively. It provides a new idea for microalgae identification and a new opportunity for the sustainable development of new energy.

Towards Retrieval-Based Neural Code Summarization: A Meta-Learning Approach

Code summarization aims to generate code summaries automatically, and has attracted a lot of research interest lately. Recent approaches to it commonly adopt neural machine translation techniques, which train a Seq2Seq model on a large corpus and assume it could work on various new code snippets. However, codes are highly varied in practice due to different domains, businesses or programming styles. Therefore, it is challenging to learn such a variety of patterns into a single model. In this paper, we propose a brand-new framework for code summarization based on meta-learning and code retrieval, named MLCS to tackle this issue. In this framework, the summarization of each target code is formalized as a few-shot learning task, where its similar examples are used as training data and the testing example is itself. We retrieve examples similar to the target code in a rank-and-filter manner. Given a neural code summarizer, we optimize it into a meta-learner via Model-Agnostic Meta-Learning (MAML). During inference, the meta-learner first adapts to the retrieved examples and yields an exclusive model for the target code, and then generates its summary. Extensive experiments on real-world datasets show: (1) Utilizing MLCS, a standard Seq2Seq model is able to outperform previous state-of-the-art approaches, including both neural models and retrieval-based neural models; (2) MLCS can flexibly adapt to existing neural code summarizers without modifying their architecture, and could significantly improve their performance with the relative gain of up to 112.7% on BLEU-4, 23.2% on ROUGE-L, and 31.5% on METEOR; (3) Compared to the existing retrieval-based neural approaches, MLCS can better leverage multiple similar examples, and shows better generalization ability on different retrievers, unseen retrieval corpus and low-frequency words.

BuresNet: Conditional Bures Metric for Transferable Representation Learning.

As a fundamental manner for learning and cognition, transfer learning has attracted widespread attention in recent years. Typical transfer learning tasks include unsupervised domain adaptation (UDA) and few-shot learning (FSL), which both attempt to sufficiently transfer discriminative knowledge from the training environment to the test environment to improve the model's generalization performance. Previous transfer learning methods usually ignore the potential conditional distribution shift between environments. This leads to the discriminability degradation in the test environments. Therefore, how to construct a learnable and interpretable metric to measure and then reduce the gap between conditional distributions is very important in the literature. In this article, we design the Conditional Kernel Bures (CKB) metric for characterizing conditional distribution discrepancy, and derive an empirical estimation with convergence guarantee. CKB provides a statistical and interpretable approach, under the optimal transportation framework, to understand the knowledge transfer mechanism. It is essentially an extension of optimal transportation from the marginal distributions to the conditional distributions. CKB can be used as a plug-and-play module and placed onto the loss layer in deep networks, thus, it plays the bottleneck role in representation learning. From this perspective, the new method with network architecture is abbreviated as BuresNet, and it can be used extract conditional invariant features for both UDA and FSL tasks. BuresNet can be trained in an end-to-end manner. Extensive experiment results on several benchmark datasets validate the effectiveness of BuresNet.

A metric-learning method for few-shot cross-event rumor detection

With the rapid development of social media, quickly detecting rumors on social media has become vitally crucial. However, there exists the following challenges. 1) Rumors are always flooded with time-critical events, where large-scale labeled datasets are difficult to obtain. 2) Although historic events have sufficient labels, the performance of models tends to degrade on the newly emergent events due to event difference shift. Facing these challenges, in this study, we attempt to leverage the idea of few-shot learning which aims to quickly acquire knowledge on unseen events with a few labeled samples by sufficiently learning from old events with a number of verified samples. Different from few-shot learning tasks in the literature, rumors in the same class of different events (no matter historical events or new events) are likely to contain coincident features in their embedding space. Therefore, we validate two classical metric learning methods, the prototypical network (PN) and the relation network (RN) which are able to capture the class-level representations in few-shot learning settings, to explore the effectiveness of metric learning methods for cross-event rumor detection. Our proposed model contains two stages corresponding to Base-Classifier pre-trained module and Base-Meta training module. The Base-Classifier pre-trained module is a classification model trained on old events. Then its last layer for class prediction is removed and the remains of the module is viewed as an encoder. The Base-Meta training module fine-tunes the encoder, meanwhile trains a selected metric learning model with episodic training strategies on old events. Empirical tests on novel events have showed that our model can outperform the state-of-the-art baseline models on the benchmark cross-event rumor datasets PHEME5 and PHEME9. What’s more, the model can improve the performance of cross-data rumor detection, where the model is trained on Twitter15, but tested on Twitter16.

Self-taught cross-domain few-shot learning with weakly supervised object localization and task-decomposition

The shift between the source and target domains is the main challenge in cross-domain few-shot learning (CD-FSL) tasks. However, the target domain is unknown during training in the source domain, which results in a lack of directed guidance for target tasks. We observe that, owing to similar backgrounds in the target domains, self-labelled samples can be applied as prior tasks to transfer knowledge onto target tasks. Accordingly, we propose a task-expansion-decomposition framework for CD-FSL called the self-taught (ST) approach, which alleviates the problem of non-target guidance by constructing task-oriented metric spaces. Specifically, weakly supervised object localization (WSOL) and self-supervised technologies are employed to enrich task-oriented samples by exchanging and rotating discriminative regions, which generates a more abundant task set. Then, these tasks are decomposed into several tasks to complete few-shot recognition and rotation classification tasks. Transferring the source knowledge onto the target tasks and focusing on discriminative regions is beneficial to the task completion. We conducted extensive experiments under a cross-domain setting, including eight target domains: CUB, Cars, Places, Plantae, CropDieases, EuroSAT, ISIC, and ChestX. The experimental results demonstrate that the proposed ST approach is applicable to various metric-based models and provides promising improvements to CD-FSL.

A Primer on Contrastive Pretraining in Language Processing: Methods, Lessons Learned, and Perspectives

Modern natural language processing (NLP) methods employ self-supervised pretraining objectives such as masked language modeling to boost the performance of various downstream tasks. These pretraining methods are frequently extended with recurrence, adversarial, or linguistic property masking. Recently, contrastive self-supervised training objectives have enabled successes in image representation pretraining by learning to contrast input-input pairs of augmented images as either similar or dissimilar. In NLP however, a single token augmentation can invert the meaning of a sentence during input-input contrastive learning, which led to input-output contrastive approaches that avoid the issue by instead contrasting over input-label pairs. In this primer, we summarize recent self-supervised and supervised contrastive NLP pretraining methods and describe where they are used to improve language modeling, zero to few-shot learning, pretraining data-efficiency, and specific NLP tasks. We overview key contrastive learning concepts with lessons learned from prior research and structure works by applications. Finally, we point to open challenges and future directions for contrastive NLP to encourage bringing contrastive NLP pretraining closer to recent successes in image representation pretraining.

Generalized Knowledge Distillation via Relationship Matching.

The knowledge of a well-trained deep neural network (a.k.a. the teacher) is valuable for learning similar tasks. Knowledge distillation extracts knowledge from the teacher and integrates it with the target model (a.k.a. the student), which expands the student's knowledge and improves its learning efficacy. Instead of restricting the teacher from working on the same task as the student, we borrow the knowledge of a teacher trained from a general label space --- in this Generalized Knowledge Distillation (GKD), the classes of the teacher and the student may be the same, completely different, or partially overlapped. We claim that the comparison ability between instances acts as an essential factor threading knowledge across tasks, and propose the Relationship Facilitated Local Classifier Distillation (ReFilled) approach, which decouples the GKD flow of the embedding and the top-layer classifier. In particular, different from reconciling the instance-label confidence between models, ReFilled requires the teacher to reweight the hard tuples push forwarded by the student adaptively and then matches the similarity comparison levels between instances. ReFilled demonstrates strong discriminative ability when the classes of the teacher vary from the same to a fully non-overlapped set w.r.t. the student.

Multiscale Wavelet Prototypical Network for Cross-Component Few-Shot Intelligent Fault Diagnosis

The techniques of machine learning, as well as deep learning (DL) methods, have seen a wide application in the intelligent fault diagnosis field these years. However, contemporary methods are still restricted under some drawbacks: 1) conventional DL-based models always rely on the quality and amount of the data. However, there are usually insufficient samples in practical scenarios because of suddenly happened failures and 2) the existing DL models cannot be well implemented in different rotating components, which have different distributions and label space, such as from bearings to gears. To address these problems, a novel multiscale wavelet prototypical network (MWPN) is proposed in this study. It is designed to solve the few-shot fault diagnosis of the cross-component problem in rotating machines: the model is trained by one component with sufficient data and tested in another component with little data. First, a multiscale wavelet convolution module is designed to extract abundant features. Second, a metric meta-learner module is applied to measure the distance distribution between the labeled and unlabeled data. With the episode training strategy, the model is optimized and can adapt to similar tasks in a new machine and classify the unknown fault categories with few labeled samples. Experiments on three datasets are carried out to demonstrate the effectiveness of MWPN. Extensive experimental results show that MWPN outperforms many baseline methods on few-shot learning tasks in different working conditions and components.

SS-VPE: Semi-Supervised Variational Prototyping Encoder With Student’s-t Mixture Model

As humans can identify new targets with one or a few images, machine learning can use a few images to generalize effective representations so as to achieve few-shot learning. Previous research has established relationships between visual prototype in input space and feature representations in latent space to capture data distributions. However, such visual prototype that can compactly represent class semantics are often lacking in most datasets. This study is aimed at distinguishing samples by using a latent prototype that represents class information in latent space. However, a latent prototype is mostly sensitive to outliers; therefore, a general method using a semi-supervised variational prototyping encoder (SS-VPE) is proposed. The contributions of this study are three-fold. First, a generative unsupervised learning is employed to optimize the prototypes in latent space, such that the data distribution can be captured accurately. Second, the Student’s-t mixture model is applied to the above clustering in the latent space, resulting in high robustness against outliers. Finally, the conventional variational autoencoder (VAE) is advanced by few-shot semi-supervised learning. Intensive experiments are conducted on four popular benchmark datasets, and the experimental results show that our proposed SS-VPE outperforms the state-of-the-art methods, significantly improving both the clustering and classification ability of few-shot learning tasks.

Graph Neural Networks With Triple Attention for Few-Shot Learning

Recent advances in Graph Neural Networks (GNNs) have achieved superior results in many challenging tasks, such as few-shot learning. Despite its capacity to learn and generalize a model from only a few annotated samples, GNN is limited in scalability, as deep GNN models usually suffer from severe over-fitting and over-smoothing. In this work, we propose a novel GNN framework with a <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">triple-attention mechanism</i> , <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">i.e.</i> node self-attention, neighbor attention, and layer memory attention, to tackle these challenges. We provide both theoretical analysis and illustrations to explain why the proposed attentive modules can improve GNN scalability for few-shot learning tasks. Our experiments show that the proposed Attentive GNN model outperforms the state-of-the-art few-shot learning methods using both GNN and non-GNN approaches. The improvement is consistent over the mini-ImageNet, tiered-ImageNet, CUB-200-2011, and Flowers-102 benchmarks, using both ConvNet-4 and ResNet-12 backbones, and under both the inductive and transductive settings. Furthermore, we demonstrate the superiority of our method for few-shot fine-grained and semi-supervised classification tasks with extensive experiments. The code for this work is publicly available at <uri xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">https://github.com/chenghao-ch94/AGNN</uri> .

Consistent Meta-Regularization for Better Meta-Knowledge in Few-Shot Learning.

Recently, meta-learning provides a powerful paradigm to deal with the few-shot learning problem. However, existing meta-learning approaches ignore the prior fact that good meta-knowledge should alleviate the data inconsistency between training and test data, caused by the extremely limited data, in each few-shot learning task. Moreover, legitimately utilizing the prior understanding of meta-knowledge can lead us to design an efficient method to improve the meta-learning model. Under this circumstance, we consider the data inconsistency from the distribution perspective, making it convenient to bring in the prior fact, and propose a new consistent meta-regularization (Con-MetaReg) to help the meta-learning model learn how to reduce the data-distribution discrepancy between the training and test data. In this way, the ability of meta-knowledge on keeping the training and test data consistent is enhanced, and the performance of the meta-learning model can be further improved. The extensive analyses and experiments demonstrate that our method can indeed improve the performances of different meta-learning models in few-shot regression, classification, and fine-grained classification.

Re-FeMAT: A Reconfigurable Multifunctional FeFET-Based Memory Architecture

Most of current processing-in-memory (PIM) architectures are application specific, that is, they can only accelerate particular functions, e.g., matrix-vector dot product for neural network acceleration. However, practical applications usually involve various functions. In order to accelerate different functions, various accelerators, and dedicated circuits have been proposed. In this work, by exploring the similarities among some commonly used dedicated circuits, we adopt ferroelectric field-effect transistors (FeFETs) to build a reconfigurable multifunctional memory architecture named Re-FeMAT. Re-FeMAT is composed of multiple processing elements (PEs). Each PE is not only a nonvolatile memory array, but also can perform logic operations (i.e., the PIM mode), convolutions (i.e., the binary convolutional neural network and the convolutional neural network (CNN) acceleration mode) and content search (i.e., the ternary content-addressable memory (TCAM) mode) without changing the circuit structure. Re-FeMAT can support applications that require multiple functions. As an example, by configuring different PEs to different working modes and using a simulated annealing algorithm or a tabu search algorithm to optimize the task-PE assignment, Re-FeMAT can completely accelerate few-shot learning applications. Our simulation results based on a calibrated FeFET model show that the proposed Re-FeMAT architecture achieves better performance and power efficiency than the previous FeMAT architecture. Compared with FeFET-based single-functional circuits, though the power dissipation of Re-FeMAT is higher in some modes, the power-delay product is still smaller. Compared with a state-of-the-art FeFET-based multifunctional accelerator named attention-in-memory, Re-FeMAT achieves lower power, latency, and energy when accelerating a complete few-shot learning task.

Efficient Personalized Speech Enhancement Through Self-Supervised Learning

This work presents self-supervised learning methods for monaural speaker-specific (i.e., personalized) speech enhancement models. While general-purpose models must broadly address many speakers, personalized models can adapt to a particular speaker's voice, expecting to solve a narrower problem. Hence, personalization can achieve more optimal performance in addition to reducing computational complexity. However, naive personalization methods can inconveniently require clean speech from the target user, e.g., due to subpar recording conditions. To this end, we pose personalization as either a zero-shot task, in which no clean speech of the target speaker is used, or a few-shot learning task, which is to minimize the duration of the clean speech used for transfer learning. With this paper, we propose self-supervised learning methods as a solution to both zero- and few-shot personalization tasks. The proposed methods learn the personalized speech features from unlabeled data (i.e., in-the-wild noisy recordings from the target user) rather than from the clean sources. We investigate three different self-supervised learning mechanisms. We set up a pseudo speech enhancement problem as a pretext task, which pretrains the models to estimate noisy speech as if it were the clean target. Contrastive learning and data purification methods regularize the loss function of the pseudo enhancement problem, overcoming the limitations of learning from unlabeled data. We assess our methods by personalizing the well-known ConvTasNet architecture to twenty different target speakers. The results show that self-supervision-based personalization improves the original ConvTasNet's enhancement quality with fewer model parameters and less clean data from the target user.

Self-Supervised Learning on Point Clouds by Contrasting Global and Local Features

Traditional supervised learning methods rely on large amounts of labeled data, while collecting labeled data with high quality is expensive. To this end, a self-supervised learning method based on contrasting global and local features of point clouds is proposed, which includes two stages: data construction and contrastive learning. In the data construction phase, local regions are generated by partial scanning from different views and cropping local substructures. In the contrastive learning phase, the global object and the local region are sent to the encoder, the projection layer and the predictor in sequence to get the global and local features respectively, and the objective function based on the contrastive learning is used to enhance the similarity between global and local features. The experimental results on two public datasets including ModelNet40 and ShapeNet show that compared with other methods such as Info3D, the proposed method can significantly improve the classification accuracy in both unsupervised point cloud classification and few-shot learning tasks, and it is more robust than existing methods when training data is scarce.