TAEML: Task-Adaptive Ensemble of Meta-Learners

NeurIPS 2018 Workshop on Metalearning TAEML: Task-Adaptive Ensemble of Meta-LearnersMinseop Park, Saehoon Kim, Jungtaek Kim, Yanbin Liu, Seungjin Choi Most of meta-learning methods assume that a set of tasks in the meta-training phase is sampled from a single dataset. Thus when a new task ...

Stochastic Chebyshev Gradient Descent for Spectral Optimization

NeurIPS 2018 Stochastic Chebyshev Gradient Descent for Spectral OptimizationInsu Han, Haim Avron, Jinwoo Shin A large class of machine learning techniques requires the solution of optimization problems involving spectral functions of parametric matrices, e.g. log-determinant and nucle...

Stacked Semantic-Guided Attention Model for Fine-Grained Zero-Shot Learning

NeurIPS 2018 Stacked Semantic-Guided Attention Model for Fine-Grained Zero-Shot LearningIYunlong Yu, Zhong Ji, Yanwei Fu, Jichang Guo, Yanwei Pang, Zhongfei Zhang Zero-Shot Learning (ZSL) is achieved via aligning the semantic relationships between the global image feature vector ...

Product Quantized Translation for Fast Nearest Neighbor Search

AAAI 2018 Product Quantized Translation for Fast Nearest Neighbor SearchYoonho Hwang, Mooyeol Baek, Saehoon Kim, Bohyung Han, Hee-Kap Ahn We propose an effective filtering algorithm to eliminate nearest neighbor candidates using their distance lower bounds in nonlinear embedded spaces, con...

On the Optimal Bit Complexity of Circulant Binary Embedding

AAAI 2018 On the Optimal Bit Complexity of Circulant Binary EmbeddingSaehoon Kim, Jungtaek Kim, Seungjin Choi In this paper, to support promising empirical results of CBE, we extend the previous theoretical framework to address the optimal condition on the number of bits, achieving that CB...

MSplit LBI: Realizing Feature Selection and Dense Estimation Simultaneously in Few-shot and Zero-shot Learning

ICML 2018 MSplit LBI: Realizing Feature Selection and Dense Estimation Simultaneously in Few-shot and Zero-shot LearningBo Zhao, Xinwei Sun, Yanwei Fu, Yuan Yao, Yizhou Wang We propose the idea that the features consist of three orthogonal parts, namely sparse strong signals, den...

Lifelong Learning with Dynamically Expandable Networks

ICLR 2018 Lifelong Learning with Dynamically Expandable NetworksJaehong Yoon, Eunho Yang, Jeongtae Lee, Sungju Hwang We propose a novel deep network architecture for lifelong learning which we refer to as Dynamically Expandable Network (DEN), that can dynamically decide its network capacit...

Joint Active Feature Acquisition and Classification with Variable-Size Set Encoding

NeurIPS 2018 Joint Active Feature Acquisition and Classification with Variable-Size Set EncodingHajin Shim, Sung Ju Hwang, Eunho Yang We consider the problem of active feature acquisition, where we sequentially select the subset of features in order to achieve the maximum predict...

DropMax: Adaptive Variational Softmax

NeurIPS 2018 DropMax: Adaptive Variational SoftmaxHaebeom Lee, Juho Lee, Saehoon Kim, Eunho Yang, Sung Ju Hwang We propose DropMax, a stochastic version of softmax classifier which at each iteration drops non-target classes according to dropout probabilities adaptively decided for each ins...

Deep Asymmetric Multi-task Feature Learning

ICML 2018 Deep Asymmetric Multi-task Feature LearningHae Beom Lee, Eunho Yang, Sung Ju Hwang We propose Deep Asymmetric Multitask Feature Learning (Deep-AMTFL) which can learn deep representations shared across multiple tasks while effectively preventing negative transfer that may happen i...