United States

Although deep neural networks have achieved amazing results on instance segmentation, they are still ill-equipped when they are required to learn new tasks incrementally. Concretely, they suffer from â€œcatastrophic forgettingâ€, an abrupt degradation of performance on old classes with the initial training data missing. Moreover, they are subjected to a negative transfer problem on new classes, which renders the model unable to update its knowledge while preserving the previous knowledge. To address these problems, we propose an incremental instance segmentation method that consists of three networks: Former Teacher Network (FTN), Current Student Network (CSN) and Current Teacher Network (CTN). Specifically, FTN supervises CSN to preserve the previous knowledge, and CTN supervises CSN to adapt to new classes. The supervision of two teacher networks is achieved by a distillation loss function for instances, bounding boxes, and classes. In addition, we adjust the supervision weights of different
teacher networks to balance between the knowledge preservation for former classes and the adaption to new classes. Extensive experimental results on PASCAL 2012 SBD and COCO datasets show the effectiveness of the proposed method.

AAAI 2021

Class-Incremental Instance Segmentation via Multi-Teacher Networks

The purpose of the AAAI conference is to promote research in artificial intelligence (AI) and scientific exchange among AI researchers, practitioners, scientists, and engineers in affiliated disciplines. 

technical paper

Recent work introduced progressive network growing as a promising way to ease the training for large GANs, but the model design and architecture-growing strategy still remain under-explored and needs manual design for different image data. In this paper, we propose a method to dynamically grow a GAN during training, optimizing the network architecture and its parameters together with automation. The method embeds architecture search techniques as an interleaving step with gradient-based training to periodically seek the optimal architecture-growing strategy for the generator and discriminator. It enjoys the benefits of both eased training because of progressive growing and improved performance because of broader architecture design space. Experimental results demonstrate new state-of-the-art of image generation. Observations in the search procedure also provide constructive insights into the GAN model design such as generator-discriminator balance and convolutional layer choices.

Dynamically Grown Generative Adversarial Networks

Automatic few-shot font generation is a practical and widely studied problem because manual designs are expensive and sensitive to the expertise of designers. Existing few-shot font generation methods aim to learn to disentangle the style and content element from a few reference glyphs, and mainly focus on a universal style representation for each font style. However, such approach limits the model in representing diverse local styles, and thus makes it unsuitable to the most complicated letter system, e.g., Chinese, whose characters consist of a varying number of components (often called ``radical'') with a highly complex structure. In this paper, we propose a novel font generation method by learning localized styles, namely component-wise style representations, instead of universal styles. The proposed style representations enable us to synthesize complex local details in text designs. However, learning component-wise styles solely from reference glyphs is infeasible in the few-shot font generation scenario, when a target script has a large number of components, e.g., over 200 for Chinese. To reduce the number of reference glyphs, we simplify component-wise styles by a product of component factor and style factor, inspired by low-rank matrix factorization. Thanks to the combination of strong representation and a compact factorization strategy, our method shows remarkably better few-shot font generation results (with only 8 reference glyph images) than other state-of-the-arts, without utilizing strong locality supervision, e.g., location of each component, skeleton, or strokes. The source code is available at https://github.com/clovaai/lffont.

Few-Shot Font Generation with Localized Style Representations and Factorization

Recently, Transformer has been demonstrating promising performance in many NLP tasks and showing a trend of replacing Recurrent Neural Network (RNN). Meanwhile, less attention is drawn to Convolutional Neural Network (CNN) due to its weak ability in capturing sequential and long-distance dependencies, although it has excellent local feature extraction capability. In this paper, we introduce an Attentive Convolutional Transformer (ACT) that takes the advantages of both Transformer and CNN for efficient text classification. Specifically, we propose a novel attentive convolution mechanism that utilizes the semantic meaning of convolutional filters attentively to transform text from complex word space to a more informative convolutional filter space where important n-grams are captured. ACT is able to capture both local and global dependencies effectively while preserving sequential information. Experiments on various text classification tasks and detailed analyses show that ACT is a lightweight, fast, and effective universal text classifier, outperforming CNNs, RNNs, and attentive models including Transformer.

ACT: An Attentive Convolutional Transformer for Efficient Text Classification

Zero-Shot Learning (ZSL) aims to recognize images belonging to unseen classes that are unavailable in the training
process, while Generalized Zero-Shot Learning (GZSL) is a more realistic variant that both seen and unseen classes appear during testing. Most GZSL approaches achieve knowledge transfer based on the features of samples that inevitably contain information irrelevant to recognition, bringing negative influence for the performance. In this work, we propose a novel method, dubbed Disentangled-VAE, which aims to disentangle category-distilling factors and category-dispersing factors from visual as well as semantic features, respectively. In addition, a batch re-combining strategy on latent features is introduced to guide the disentanglement, encouraging the distilling latent features to be more discriminative for recognition. Extensive experiments demonstrate that our method outperforms the state-of-the-art approaches on four challenging benchmark datasets

Generalized Zero-Shot Learning via Disentangled Representation

We study the streaming model for approximate matrix multiplication (AMM). We are interested in the scenario that the algorithm can only take one pass over the data with limited memory. The state-of-the-art deterministic sketching algorithm for streaming AMM is the co-occurring directions (COD), which has much smaller approximation errors than randomized algorithms and outperforms other deterministic sketching methods empirically. In this paper, we provide a tighter error bound for COD whose leading term considers the potential approximate low-rank structure and the correlation of input matrices.  We prove COD is space optimal with respect to our improved error bound. We also propose a variant of COD for sparse matrices with theoretical guarantees. The experiments on real-world sparse datasets show that the proposed algorithm is more efficient than baseline methods.

Revisiting Co-Occurring Directions: Sharper Analysis and Efficient Algorithm for Sparse Matrices

Narrative Generation has attracted significant interest as a novel application of Automated Planning techniques. However, the vast amount of narrative material available opens the way to the use of Deep Learning techniques. In this paper, we explore the feasibility of narrative generation through self-supervised learning, using sequence embedding techniques or auto-encoders to produce narrative sequences. We use datasets of well-formed plots generated by a narrative planning approach, using pre-existing, published, narrative planning domains, to train generative models.
Our experiments demonstrate the ability of generative sequence models to produce narrative plots with similar structure to those obtained with planning techniques, but with significant plot novelty in comparison with the training set. Most importantly, generated plots share structural properties associated with narrative quality measures used in Planning-based methods. As plan-based structures account for a higher level of causality and narrative consistency, this suggests that our approach is able to extend a set of narratives with novel sequences that display the same high-level narrative properties.
Unlike methods developed to extend sets of textual narratives, ours operates at the level of plot structure. Thus, it has the potential to be used across various media for plots of significant complexity, being initially limited to training and generation operating in the same narrative genre.

Narrative Plan Generation with Self-Supervised Learning

Double Q-learning is a popular reinforcement learning algorithm in Markov decision process (MDP) problems. Clipped Double Q-learning, as an effective variant of Double Q-learning, employs the clipped double estimator to approximate the maximum expected action value. Due to the underestimation bias of the clipped double estimator, performance of clipped Double Q-learning may be degraded in some stochastic environments. In this paper, in order to reduce the underestimation bias, we propose an action candidate based clipped double estimator for Double Q-learning. Specifically, we first select a set of elite action candidates with the high action values from one set of estimators. Then, among these candidates, we choose the highest valued action from the other set of estimators. Finally, we use the maximum value in the second set of estimators to clip the action value of the chosen action in the first set of estimators and the clipped value is used for approximating the maximum expected action value.  Theoretically, the underestimation bias in our clipped Double Q-learning decays monotonically as the number of the action candidates decreases.  Moreover, the number of action candidates controls the trade-off between the overestimation and underestimation biases. In addition, we also extend our clipped Double Q-learning to continuous action tasks via approximating the elite continuous action candidates. We empirically verify that our algorithm can more accurately estimate the maximum expected action value on some toy environments and yield good performance on several benchmark problems.

Action Candidate Based Clipped Double Q-Learning for Discrete and Continuous Action Tasks

Imagining a colored realistic image from an arbitrary-drawn sketch is one of human capabilities that we eager machines to mimic.  Unlike previous methods that either require the sketch-image pairs or utilize low-quantity detected edges as sketches, we study the exemplar-based sketch-to-image (s2i) synthesis task in a self-supervised learning manner, eliminating the necessity of the paired sketch data. To this end,  we first propose an unsupervised method to efficiently synthesize line-sketches for general RGB-only datasets. With the synthetic paired-data, we then present a self-supervised Auto-Encoder (AE) to decouple the content/style features from sketches and RGB-images, and synthesize images both content-faithful to the sketches and style-consistent to the RGB-images. While prior works employ either the cycle-consistence loss or dedicated attentional modules to enforce the content/style fidelity, we show AE's superior performance with pure self-supervisions.  To further improve the synthesis quality in high resolution, we also leverage an adversarial network to refine the details of synthetic images. Extensive experiments on $1024^2$ resolution demonstrate a new state-of-art-art performance of the proposed model on CelebA-HQ and Wiki-Art datasets. Moreover, with the proposed sketch generator, the model shows a promising performance on style mixing and style transfer, which the synthesized images are not only style-consistent but also semantically meaningful.

Self-Supervised Sketch-to-Image Synthesis

In an order-driven financial market, the price of a financial asset is discovered through the interaction of orders - requests to buy or sell at a particular price - that are posted to the public limit order book (LOB). Therefore, LOB data is extremely valuable for modelling market dynamics. However, LOB data is not freely accessible, which poses a challenge to market participants and researchers wishing to exploit this information. Fortunately, trades and quotes (TAQ) data - orders arriving at the top of the LOB, and trades executing in the market - are more readily available. In this paper, we present the LOB recreation model, a first attempt from a deep learning perspective to recreate the top five price levels of the LOB for small-tick stocks using only TAQ data. Volumes of orders sitting deep in the LOB are predicted by combining outputs from: (1) a history compiler that uses a Gated Recurrent Unit (GRU) module to selectively compile prediction relevant quote history; (2) a market events simulator, which uses an Ordinary Differential Equation Recurrent Neural Network (ODE-RNN) to simulate the accumulation of net order arrivals; and (3) a weighting scheme to adaptively combine the predictions generated by (1) and (2). By the paradigm of transfer learning, the core encoder trained on one stock can be fine-tuned to enable application to other financial assets of the same class with much lower demand on additional data. Comprehensive experiments conducted on two real world intraday LOB datasets demonstrate that the proposed model can efficiently recreate the LOB with high accuracy using only TAQ data as input.

The LOB Recreation Model: Predicting the Limit Order Book from TAQ History Using an Ordinary Differential Equation Recurrent Neural Network

Focusing on text-to-image (T2I) generation, we propose Text and Image Mutual-Translation Adversarial Networks (TIME), a lightweight but effective model that jointly learns a T2I generator G and an image captioning discriminator D under the Generative Adversarial Network framework. While previous methods tackle the T2I problem as a uni-directional task and use pre-trained language models to enforce the image--text consistency, TIME requires neither extra modules nor pre-training. We show that the performance of G can be boosted substantially by training it jointly with D as a language model. Specifically, we adopt Transformers to model the cross-modal connections between the image features and word embeddings, and design an annealing conditional hinge loss that dynamically balances the adversarial learning. In our experiments, TIME achieves state-of-the-art (SOTA) performance on the CUB dataset (Inception Score of 4.91 and FrÃ©chet Inception Distance of 14.3 on CUB), and shows promising performance on MS-COCO dataset on image captioning and downstream vision-language tasks.

Downloads

Next from AAAI 2021

Dynamically Grown Generative Adversarial Networks

Similar lecture

Robust Knowledge Transfer via Hybrid Forward on the Teacher-Student Model

.css-70qvj9{display:-webkit-box;display:-webkit-flex;display:-ms-flexbox;display:flex;-webkit-align-items:center;-webkit-box-align:center;-ms-flex-align:center;align-items:center;}Downloads

Next from AAAI 2021

Dynamically Grown Generative Adversarial Networks

Similar lecture

Robust Knowledge Transfer via Hybrid Forward on the Teacher-Student Model

Downloads