Zero-shot Setting Research Articles

Zero-shot learning via a specific rank-controlled semantic autoencoder

Existing embedding zero-shot learning models usually learn a projection function from the visual feature space to the semantic embedding space, e.g. attribute space or word vector space. However, the projection learned based on seen samples may not generalize well to unseen classes, which is known as the projection domain shift problem in ZSL. To address this issue, we propose a method named Low-rank Semantic Autoencoder (LSA) to consider the low-rank structure of seen samples to maintain the sparse feature of reconstruction error, which can further improve zero-shot learning capability. Moreover, to obtain a more robust projection for unseen classes, we propose a Specific Rank-controlled Semantic Autoencoder (SRSA) to accurately control of the projection’s rank. Extensive experiments on six benchmarks demonstrate the superiority of the proposed models over most existing embedding ZSL models under the standard zero-shot setting and the more realistic generalized zero-shot setting.

Relation-based Discriminative Cooperation Network for Zero-Shot Classification

Zero-shot learning (ZSL) aims to assign the category corresponding to the relevant semantic as the label of the unseen sample based on the relationship between the learned visual and semantic features. However, most typical ZSL models faced with the domain bias problem, which leads to unseen or test samples being easily misclassified into seen or training categories. To handle this problem, we propose a relation-based discriminative cooperation network (RDCN) model for ZSL in this work. The proposed model effectively utilize the robust metric space spanned by the cooperated semantics with the help of a set of relations. On the other hand, we devise the relation network to measure the relationship between the visual features and embedded semantics, and the validation information will guide the embedding module to learn more discriminative information. At last, the proposed RDCN model is validated on six benchmarks, and extensive experiments demonstrate the superiority of proposed method over most existing ZSL models on the traditional zero-shot setting and the more realistic generalized zero-shot setting.

Latent factor model based zero-shot learning

The key of zero-shot learning is to design an appropriate approach to capture the potential relevance between the image visual features and category-level semantic knowledge. There is a lot of data with matrix form in zero-shot recognition problem, such as image label matrix, visual feature matrix and semantic information matrix. Inspired by the latent factor model which tackles recommendation problems well avoid being influenced by sparse matrix, we regard an image as a user and analogize each dimension of the visual feature as an item. The key to our approach is that we model the image recognition as a special recommendation problem by aid of latent factor model. We evaluate our algorithm on three classic benchmark data sets for both conventional and generalized zero-shot setting, the classification results outperform significantly the state-of-art approaches with low calculation cost.

Zero-Shot Video Event Detection With High-Order Semantic Concept Discovery and Matching

Multimedia event detection aims to precisely retrieve videos that contain complex semantic events from a large pool. This work addresses this task under a zero-shot setting, where only brief event-specific textural information (such as event names, a few descriptive sentences, etc.) is known yet none positive video example is provided. Mainstream approaches to tackling this task are middle-level semantic concept-based, where meticulously-crafted concept banks (e.g., LSCOM) are adopted. We argue that these concept banks are still inadequate facing video semantic complexity. Existing semantic concepts are essentially first-order, mainly designed for atomic objects, scenes or human actions, etc. This work advocates the utilization of high-order concepts (such as subject-predicate-object triplets or adjective-object). The main contributions are two-fold. First, we harvest a comprehensive albeit compact high-order concept library through distilling information from three large public datasets (MS-COCO, Visual Genome, and Kinetics-600), mainly related to visual relations and human-object interactions. Secondly, zero-shot events are often only briefly and partially described via textual input. The resultant semantic ambiguity makes the pursuit of the most indicative high-order concepts challenging. We thus design a novel query-expanding scheme that enriches ambiguous event-specific keywords by searching over either large common knowledge bases ( <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">e.g.</i> , WikiHow) or top-ranked webpages retrieved from modern search engines. This way sets up a more faithful connection between zero-shot events and high-order concepts. To our best knowledge, this is the first work that strives for concept-based video search beyond first-order concepts. Extensive experiments have been conducted on several large video benchmarks (TRECVID 2013, TRECVID 2014, and ActivityNet-1.3). The evaluations clearly demonstrate the superiority of our constructed high-order concept library and its complementariness to existing concepts.

Zero-shot Handwritten Chinese Character Recognition with hierarchical decomposition embedding

Handwritten Chinese Character Recognition (HCCR) is a challenging topic in the field of pattern recognition due to large-scale character vocabulary, complex hierarchical structure, various writing styles, and scarce training samples. In this paper, we explored the hierarchical knowledge of Chinese characters and presented a novel zero-shot HCCR method. First, we handled the relations between the characters and their primitives, such as radicals and structures, to obtain a tree layout of primitives. Then, we presented a novel zero-shot hierarchical decomposition embedding method to encode the tree layout into a semantic vector. Next, we devised a Convolutional Neural Network (CNN) based framework to learn both radicals and structures of characters via the semantic vector. As different Chinese characters share some common radicals and structures, our method is able to recognize new categories without any labeled samples from them. Moreover, our method is effective in both traditional HCCR and zero-shot HCCR tasks. It achieves competitive performance on the traditional experiment setting and significantly surpasses the state-of-the-art methods on the zero-shot experiment setting.

MA-DST: Multi-Attention-Based Scalable Dialog State Tracking

Task oriented dialog agents provide a natural language interface for users to complete their goal. Dialog State Tracking (DST), which is often a core component of these systems, tracks the system's understanding of the user's goal throughout the conversation. To enable accurate multi-domain DST, the model needs to encode dependencies between past utterances and slot semantics and understand the dialog context, including long-range cross-domain references. We introduce a novel architecture for this task to encode the conversation history and slot semantics more robustly by using attention mechanisms at multiple granularities. In particular, we use cross-attention to model relationships between the context and slots at different semantic levels and self-attention to resolve cross-domain coreferences. In addition, our proposed architecture does not rely on knowing the domain ontologies beforehand and can also be used in a zero-shot setting for new domains or unseen slot values. Our model improves the joint goal accuracy by 5% (absolute) in the full-data setting and by up to 2% (absolute) in the zero-shot setting over the present state-of-the-art on the MultiWoZ 2.1 dataset.

Zero Shot Learning with the Isoperimetric Loss

We introduce the isoperimetric loss as a regularization criterion for learning the map from a visual representation to a semantic embedding, to be used to transfer knowledge to unknown classes in a zero-shot learning setting. We use a pre-trained deep neural network model as a visual representation of image data, a Word2Vec embedding of class labels, and linear maps between the visual and semantic embedding spaces. However, the spaces themselves are not linear, and we postulate the sample embedding to be populated by noisy samples near otherwise smooth manifolds. We exploit the graph structure defined by the sample points to regularize the estimates of the manifolds by inferring the graph connectivity using a generalization of the isoperimetric inequalities from Riemannian geometry to graphs. Surprisingly, this regularization alone, paired with the simplest baseline model, outperforms the state-of-the-art among fully automated methods in zero-shot learning benchmarks such as AwA and CUB. This improvement is achieved solely by learning the structure of the underlying spaces by imposing regularity.

Multi-Level Cross-Lingual Transfer Learning With Language Shared and Specific Knowledge for Spoken Language Understanding

Recently conversational agents effectively improve their understanding capabilities by neural networks. Such deep neural models, however, do not apply to most human languages due to the lack of annotated training data for various NLP tasks. In this paper, we propose a multi-level cross-lingual transfer model with language shared and specific knowledge to improve the spoken language understanding of low-resource languages. Our method explicitly separates the model into the language-shared part and language-specific part to transfer cross-lingual knowledge and improve the monolingual slot tagging, especially for low-resource languages. To refine the shared knowledge, we add a language discriminator and employ adversarial training to reinforce information separation. Besides, we adopt novel multi-level knowledge transfer in an incremental and progressive way to acquire multi-granularity shared knowledge rather than a single layer. To mitigate the discrepancies between the feature distributions of language specific and shared knowledge, we propose the neural adapters to fuse knowledge automatically. Experiments show that our proposed model consistently outperforms monolingual baseline with a statistically significant margin up to 2.09%, even higher improvement of 12.21% in the zero-shot setting.

Deep Ranking for Image Zero-Shot Multi-Label Classification.

During the past decade, both multi-label learning and zero-shot learning have attracted huge research attention, and significant progress has been made. Multi-label learning algorithms aim to predict multiple labels given one instance, while most existing zero-shot learning approaches target at predicting a single testing label for each unseen class via transferring knowledge from auxiliary seen classes to target unseen classes. However, relatively less effort has been made on predicting multiple labels in the zero-shot setting, which is nevertheless a quite challenging task. In this work, we investigate and formalize a flexible framework consisting of two components, i.e., visual-semantic embedding and zero-shot multi-label prediction. First, we present a deep regression model to project the visual features into the semantic space, which explicitly exploits the correlations in the intermediate semantic layer of word vectors and makes label prediction possible. Then, we formulate the label prediction problem as a pairwise one and employ Ranking SVM to seek the unique multi-label correlations in the embedding space. Furthermore, we provide a transductive multi-label zeroshot prediction approach that exploits the testing data manifold structure. We demonstrate the effectiveness of the proposed approach on three popular multi-label datasets with state-of-theart performance obtained on both conventional and generalized ZSL settings.

Asymmetric graph based zero shot learning

Zero-shot learning (ZSL) now has gained a great deal of focus due to its ability of recognizing unseen categories by training with samples of only seen categories. Existing efforts have been devoted to learn a projection between semantic space and feature space, which has made a big progress in ZSL. However, simply establishing a projection often suffers from the visual semantic ambiguity problem and hubness problem. Specifically, visual patterns and semantic concepts often can not properly match each other, and lead to inaccurate recognition result. To this end, in this paper, we propose a novel ZSL model, namely Asymmetric Graph-based Zero Shot Learning (AGZSL), to simultaneously preserve class level semantic manifold and instance level visual manifold in a latent space. In addition, to make the model more discriminative, we also constrain the latent space to be orthogonal, which means that the projected visual features and semantic embeddings in the latent space are orthogonal when they belong to different categories. We test our approach on four benchmark datasets under both standard zero-shot setting and more realistic generalized zero-shot learning (GZSL) setting, and the results show that our AGZSL can significantly improve the final performance comparing to the state-of-the-art methods.

Attribute-Guided Network for Cross-Modal Zero-Shot Hashing.

Zero-shot hashing (ZSH) aims at learning a hashing model that is trained only by instances from seen categories but can generate well to those of unseen categories. Typically, it is achieved by utilizing a semantic embedding space to transfer knowledge from seen domain to unseen domain. Existing efforts mainly focus on single-modal retrieval task, especially image-based image retrieval (IBIR). However, as a highlighted research topic in the field of hashing, cross-modal retrieval is more common in real-world applications. To address the cross-modal ZSH (CMZSH) retrieval task, we propose a novel attribute-guided network (AgNet), which can perform not only IBIR but also text-based image retrieval (TBIR). In particular, AgNet aligns different modal data into a semantically rich attribute space, which bridges the gap caused by modality heterogeneity and zero-shot setting. We also design an effective strategy that exploits the attribute to guide the generation of hash codes for image and text within the same network. Extensive experimental results on three benchmark data sets (AwA, SUN, and ImageNet) demonstrate the superiority of AgNet on both cross-modal and single-modal zero-shot image retrieval tasks.

Inductive Zero-Shot Image Annotation via Embedding Graph

Conventional image annotation systems can only handle those images having labels within the exist library, but cannot recognize those novel labels. In order to learn new concepts, one has to gather large amount of labeled images and train the model from scratch. More importantly, it can come with a high price to collect those labeled images. For these reasons, we put forward a zero-shot image annotation model, to reduce the demand for the images with novel labels. In this paper, we focus on the two big challenges of zero-shot image annotation: polysemous words and a strong bias in the generalized zero-shot setting. For the first problem, instead of training on large corpus datasets as previous methods, we propose to adopt Node2Vec to obtain contextualized word embeddings, which can easily produce word vectors of the polysemous words. For the second problem, we alleviate the strong bias in two ways: on one hand, we utilize a model based on graph convolutional network (GCN) to make target images involved in the training process; on the other hand, we put forward a novel semantic coherent (SC) loss to capture the semantic relations of the source and target labels. The extensive experiments on NUSWIDE, COCO, IAPR TC-12, and Core15k datasets show the superiority of the proposed model and the annotation performance get improved by 4%-6% comparing with state-of-the-art methods.

Generalized zero-shot learning for action recognition with web-scale video data

Action recognition in surveillance video makes our life safer by detecting the criminal events or predicting violent emergencies. However, efficient action recognition is not free of difficulty. First, there are so many action classes in daily life that we cannot pre-define all possible action classes beforehand. Moreover, it is very hard to collect real-word videos for certain particular actions such as steal and street fight due to legal restrictions and privacy protection. These challenges make existing data-driven recognition methods insufficient to attain desired performance. Zero-shot learning is potential to be applied to solve these issues since it can perform classification without positive example. Nevertheless, current zero-shot learning algorithms have been studied under the unreasonable setting where seen classes are absent during the testing phase. Motivated by this, we study the task of action recognition in surveillance video under a more realistic generalized zero-shot setting, where testing data contains both seen and unseen classes. To our best knowledge, this is one of the first works to study video action recognition under the generalized zero-shot setting. We firstly perform extensive empirical studies on several existing zero-shot leaning approaches under this new setting on a web-scale video data. Our experimental results demonstrate that, under the generalize setting, typical zero-shot learning methods are no longer effective for the dataset we applied. Then, we propose to deploy generalized zero-shot learning which transfers the knowledge of Web video to detect the anomalous actions in surveillance videos. To verify the effectiveness of methods, we further construct a new surveillance video dataset consisting of nine action classes related to the public safety situation.

Zero-Shot Learning-A Comprehensive Evaluation of the Good, the Bad and the Ugly.

Due to the importance of zero-shot learning, i.e., classifying images where there is a lack of labeled training data, the number of proposed approaches has recently increased steadily. We argue that it is time to take a step back and to analyze the status quo of the area. The purpose of this paper is three-fold. First, given the fact that there is no agreed upon zero-shot learning benchmark, we first define a new benchmark by unifying both the evaluation protocols and data splits of publicly available datasets used for this task. This is an important contribution as published results are often not comparable and sometimes even flawed due to, e.g., pre-training on zero-shot test classes. Moreover, we propose a new zero-shot learning dataset, the Animals with Attributes 2 (AWA2) dataset which we make publicly available both in terms of image features and the images themselves. Second, we compare and analyze a significant number of the state-of-the-art methods in depth, both in the classic zero-shot setting but also in the more realistic generalized zero-shot setting. Finally, we discuss in detail the limitations of the current status of the area which can be taken as a basis for advancing it.

Zero-Shot Image Classification Based on Deep Feature Extraction

The attribute-based zero-shot learning methods generally use low-level features of images to train attribute classifiers, and the corresponding classification accuracy heavily depends on specific low-level features. Because deep networks can automatically extract features from original unlabeled images and the extracted features can better represent the nature of original images, we proposed a zero-shot image classification method based on deep feature extraction. In the image preprocessing step, in order to reduce the computational complexity and the correlations between pixels, image patches extraction and zero-phase component analysis whitening are performed. The compressed feature representations of unlabeled image patches are learned through a stacked sparse autoencoder and a feature mapping matrix can be obtained. Further, we use the feature mapping matrix as a convolution kernel to convolve with image patches. Since the convolution operation results in the feature vector with huge dimensionality, the convolution features will be pooled to reduce the number of network parameters and to reduce the spatial resolution of the network to prevent over-fitting. Finally, the exacted image features are used to train the conventional indirect attribute prediction model to predict image attributes and classify images under the zero-shot setting. Experimental results on the shoes, outdoor scene recognition, and a-Yahoo datasets show that, compared with several popular zero-shot learning methods, the proposed method can yield more accurate attribute prediction and better zero-shot image classification.

Representation Learning of Knowledge Graphs with Entity Descriptions

Representation learning (RL) of knowledge graphs aims to project both entities and relations into a continuous low-dimensional space. Most methods concentrate on learning representations with knowledge triples indicating relations between entities. In fact, in most knowledge graphs there are usually concise descriptions for entities, which cannot be well utilized by existing methods. In this paper, we propose a novel RL method for knowledge graphs taking advantages of entity descriptions. More specifically, we explore two encoders, including continuous bag-of-words and deep convolutional neural models to encode semantics of entity descriptions. We further learn knowledge representations with both triples and descriptions. We evaluate our method on two tasks, including knowledge graph completion and entity classification. Experimental results on real-world datasets show that, our method outperforms other baselines on the two tasks, especially under the zero-shot setting, which indicates that our method is capable of building representations for novel entities according to their descriptions. The source code of this paper can be obtained from https://github.com/xrb92/DKRL.

All-in Text: Learning Document, Label, and Word Representations Jointly

Conventional multi-label classification algorithms treat the target labels of the classification task as mere symbols that are void of an inherent semantics. However, in many cases textual descriptions of these labels are available or can be easily constructed from public document sources such as Wikipedia. In this paper, we investigate an approach for embedding documents and labels into a joint space while sharing word representations between documents and labels. For finding such embeddings, we rely on the text of documents as well as descriptions for the labels. The use of such label descriptions not only lets us expect an increased performance on conventional multi-label text classification tasks, but can also be used to make predictions for labels that have not been seen during the training phase. The potential of our method is demonstrated on the multi-label classification task of assigning keywords from the Medical Subject Headings (MeSH) to publications in biomedical research, both in a conventional and in a zero-shot learning setting.