arXiv subjects

Ke Lin

Publications and source records attributed to Ke Lin.

24 records · Page 2Linked to original sources

arXiv2020

A Semantics-Assisted Video Captioning Model Trained with Scheduled Sampling

Given the features of a video, recurrent neural networks can be used to automatically generate a caption for the video. Existing methods for video captioning have at least three limitations. First, semantic information has been widely applied to boost the performance of video captioning models, but existing networks often fail to provide meaningful semantic features. Second, the Teacher Forcing algorithm is often utilized to optimize video captioning models, but during training and inference, different strategies are applied to guide word generation, leading to poor performance. Third, current video captioning models are prone to generate relatively short captions that express video contents inappropriately. Toward resolving these three problems, we suggest three corresponding improvements. First of all, we propose a metric to compare the quality of semantic features, and utilize appropriate features as input for a semantic detection network (SDN) with adequate complexity in order to generate meaningful semantic features for videos. Then, we apply a scheduled sampling strategy that gradually transfers the training phase from a teacher-guided manner toward a more self-teaching manner. Finally, the ordinary logarithm probability loss function is leveraged by sentence length so that the inclination of generating short sentences is alleviated. Our model achieves better results than previous models on the YouTube2Text dataset and is competitive with the previous best model on the MSR-VTT dataset.

cs.CV↗

arXiv2020

Optimally Combining Classifiers for Semi-Supervised Learning

This paper considers semi-supervised learning for tabular data. It is widely known that Xgboost based on tree model works well on the heterogeneous features while transductive support vector machine can exploit the low density separation assumption. However, little work has been done to combine them together for the end-to-end semi-supervised learning. In this paper, we find these two methods have complementary properties and larger diversity, which motivates us to propose a new semi-supervised learning method that is able to adaptively combine the strengths of Xgboost and transductive support vector machine. Instead of the majority vote rule, an optimization problem in terms of ensemble weight is established, which helps to obtain more accurate pseudo labels for unlabeled data. The experimental results on the UCI data sets and real commercial data set demonstrate the superior classification performance of our method over the five state-of-the-art algorithms improving test accuracy by about $3\%-4\%$. The partial code can be found at https://github.com/hav-cam-mit/CTO.

cs.LG↗

arXiv2020

Multi-modal Feature Fusion with Feature Attention for VATEX Captioning Challenge 2020

This report describes our model for VATEX Captioning Challenge 2020. First, to gather information from multiple domains, we extract motion, appearance, semantic and audio features. Then we design a feature attention module to attend on different feature when decoding. We apply two types of decoders, top-down and X-LAN and ensemble these models to get the final result. The proposed method outperforms official baseline with a significant gap. We achieve 76.0 CIDEr and 50.0 CIDEr on English and Chinese private test set. We rank 2nd on both English and Chinese private test leaderboard.

cs.CV↗

arXiv2020

On boundedness, gradient estimate, blow-up and convergence in a two-species and two-stimuli chemotaxis system with/without loop

In this work, we study dynamic properties of classical solutions to a homogenous Neumann initial-boundary value problem (IBVP) for a two-species and two-stimuli chemotaxis model with/without chemical signalling loop in a 2D bounded and smooth domain. We successfully detect the product of two species masses as a feature to determine boundedness, gradient estimates, blow-up and $W^{j,\infty}(1\leq j\leq 3)$-exponential convergence of classical solutions for the corresponding IBVP. More specifically, we first show generally a smallness on the product of both species masses, thus allowing one species mass to be suitably large, is sufficient to guarantee global boundedness, higher order gradient estimates and $W^{j,\infty}$-convergence with rates of convergence to constant equilibria; and then, in a special case, we detect a straight line of masses on which blow-up occurs for large product of masses. Our findings provide new understandings about the underlying model, and thus, improve and extend greatly the existing knowledge relevant to this model.

math.AP↗

arXiv2020

Exploration-efficient Deep Reinforcement Learning with Demonstration Guidance for Robot Control

Although deep reinforcement learning (DRL) algorithms have made important achievements in many control tasks, they still suffer from the problems of sample inefficiency and unstable training process, which are usually caused by sparse rewards. Recently, some reinforcement learning from demonstration (RLfD) methods have shown to be promising in overcoming these problems. However, they usually require considerable demonstrations. In order to tackle these challenges, on the basis of the SAC algorithm we propose a sample efficient DRL-EG (DRL with efficient guidance) algorithm, in which a discriminator D(s) and a guider G(s) are modeled by a small number of expert demonstrations. The discriminator will determine the appropriate guidance states and the guider will guide agents to better exploration in the training phase. Empirical evaluation results from several continuous control tasks verify the effectiveness and performance improvements of our method over other RL and RLfD counterparts. Experiments results also show that DRL-EG can help the agent to escape from a local optimum.

cs.RO↗

arXiv2011

Laser-launched evanescent surface plasmon polariton field utilized as a direct coherent pumping source to generate emitted nonlinear four-wave mixing radiation

We develop a concept of surface plasmon polaritons (SPPs) based four-wave mixing (4WM), in which a laser-launched evanescent SPP field is utilized as a coherent pumping source to involve directly in a nonlinear 4WM process at the dielectric/metal interface. Conversion efficiency of the resulting 4WM radiation is expected to be dramatically increased due to the local-field enhancement effect. Feasibility of implementing this concept at the air/gold film and graphene flake/gold film interfaces is further examined by numerical simulations. The concept shows intriguing promise for applications in newly emerging nanophotonics, optoelectronics, and active plasmonics.

cond-mat.mes-hall↗