Semantic Autoencoder for Zero-Shot Learning

Abstract

Existing zero-shot learning (ZSL) models typically learn a projection function from a feature space to a semantic embedding space (e.g. attribute space). However, such a projection function is only concerned with predicting the training seen class semantic representation (e.g. attribute prediction) or classification. When applied to test data, which in the context of ZSL contains different (unseen) classes without training data, a ZSL model typically suffers from the project domain shift problem. In this work, we present a novel solution to ZSL based on learning a Semantic AutoEncoder (SAE). Taking the encoder-decoder paradigm, an encoder aims to project a visual feature vector into the semantic space as in the existing ZSL models. However, the decoder exerts an additional constraint, that is, the projection/code must be able to reconstruct the original visual feature. We show that with this additional reconstruction constraint, the learned projection function from the seen classes is able to generalise better to the new unseen classes. Importantly, the encoder and decoder are linear and symmetric which enable us to develop an extremely efficient learning algorithm. Extensive experiments on six benchmark datasets demonstrate that the proposed SAE outperforms significantly the existing ZSL models with the additional benefit of lower computational cost. Furthermore, when the SAE is applied to supervised clustering problem, it also beats the state-of-the-art.

A Multi-task Learning Framework for Grasping-Position Detection and Few-Shot Classification

“…Conventional studies on transfer learning have used a method to streamline learning that involves transferring the weight of a DNN pre-trained on a large data set, such as Ima-geNet, to another task [17]. Zero-shot learning [9][10][11] [18] is an approach that can be used to classify unknown classes. Because these methods can only be applied to the same task as the pre-trained task, they cannot be used for two different tasks, as in the case of this study.…”

DeepKinZero: Zero-Shot Learning for Predicting Kinase-Phosphosite Associations Involving Understudied Kinases

“…This idea is illustrated in Figure 2. Following the work in structured output prediction [45] and prior work in zero-shot 145 learning [38,39,[46][47][48][49][50], we use a compatibility function F : X × Y → R to model the mapping between the input and output embeddings. In this model, F takes a phosphosite -kinase pair (x i , y j ) as input and returns a scalar value which is proportional to the confidence of associating the site, x i , with kinase y i .…”

On Parameter Tuning in Meta-Learning for Computer Vision

“…In [24], Kodirov et al compared their proposed method, SAE, with more than 10 highly qualified methods using small datasets, which include AwA [16], CUB [27], aP &Y [38] and SUN [39]. The authors improved the accuracy of recognising unseen images at least %6 in comparison with SS−voc and at most %20 in comparison with basic attibutebased learning method, DAP.…”

“…Algorithm 1 Implementation of optimized zero-shot learning (SAE) [24] Require: A batch set of training input (X , Y ) and training size Ensure: The best mapping matrix (W ) for zero-shot learning 1: Tuning embedded parameters…”