arXiv subjects

Zhengjie Huang

Publications and source records attributed to Zhengjie Huang.

27 records · Page 2Linked to original sources

arXiv2021

NOTE: Solution for KDD-CUP 2021 WikiKG90M-LSC

WikiKG90M in KDD Cup 2021 is a large encyclopedic knowledge graph, which could benefit various downstream applications such as question answering and recommender systems. Participants are invited to complete the knowledge graph by predicting missing triplets. Recent representation learning methods have achieved great success on standard datasets like FB15k-237. Thus, we train the advanced algorithms in different domains to learn the triplets, including OTE, QuatE, RotatE and TransE. Significantly, we modified OTE into NOTE (short for Norm-OTE) for better performance. Besides, we use both the DeepWalk and the post-smoothing technique to capture the graph structure for supplementation. In addition to the representations, we also use various statistical probabilities among the head entities, the relations and the tail entities for the final prediction. Experimental results show that the ensemble of state-of-the-art representation learning methods could draw on each others strengths. And we develop feature engineering from validation candidates for further improvements. Please note that we apply the same strategy on the test set for final inference. And these features may not be practical in the real world when considering ranking against all the entities.

cs.IR↗

arXiv2021

LiteGEM: Lite Geometry Enhanced Molecular Representation Learning for Quantum Property Prediction

In this report, we (SuperHelix team) present our solution to KDD Cup 2021-PCQM4M-LSC, a large-scale quantum chemistry dataset on predicting HOMO-LUMO gap of molecules. Our solution, Lite Geometry Enhanced Molecular representation learning (LiteGEM) achieves a mean absolute error (MAE) of 0.1204 on the test set with the help of deep graph neural networks and various self-supervised learning tasks. The code of the framework can be found in https://github.com/PaddlePaddle/PaddleHelix/tree/dev/competition/kddcup2021-PCQM4M-LSC/.

physics.chem-ph↗

arXiv2021

Masked Label Prediction: Unified Message Passing Model for Semi-Supervised Classification

Graph neural network (GNN) and label propagation algorithm (LPA) are both message passing algorithms, which have achieved superior performance in semi-supervised classification. GNN performs feature propagation by a neural network to make predictions, while LPA uses label propagation across graph adjacency matrix to get results. However, there is still no effective way to directly combine these two kinds of algorithms. To address this issue, we propose a novel Unified Message Passaging Model (UniMP) that can incorporate feature and label propagation at both training and inference time. First, UniMP adopts a Graph Transformer network, taking feature embedding and label embedding as input information for propagation. Second, to train the network without overfitting in self-loop input label information, UniMP introduces a masked label prediction strategy, in which some percentage of input label information are masked at random, and then predicted. UniMP conceptually unifies feature propagation and label propagation and is empirically powerful. It obtains new state-of-the-art semi-supervised classification results in Open Graph Benchmark (OGB).

cs.LG↗

arXiv2020

kk2018 at SemEval-2020 Task 9: Adversarial Training for Code-Mixing Sentiment Classification

Code switching is a linguistic phenomenon that may occur within a multilingual setting where speakers share more than one language. With the increasing communication between groups with different languages, this phenomenon is more and more popular. However, there are little research and data in this area, especially in code-mixing sentiment classification. In this work, the domain transfer learning from state-of-the-art uni-language model ERNIE is tested on the code-mixing dataset, and surprisingly, a strong baseline is achieved. Furthermore, the adversarial training with a multi-lingual model is used to achieve 1st place of SemEval-2020 Task 9 Hindi-English sentiment classification competition.

cs.CL↗

arXiv2020

ERNIE at SemEval-2020 Task 10: Learning Word Emphasis Selection by Pre-trained Language Model

This paper describes the system designed by ERNIE Team which achieved the first place in SemEval-2020 Task 10: Emphasis Selection For Written Text in Visual Media. Given a sentence, we are asked to find out the most important words as the suggestion for automated design. We leverage the unsupervised pre-training model and finetune these models on our task. After our investigation, we found that the following models achieved an excellent performance in this task: ERNIE 2.0, XLM-ROBERTA, ROBERTA and ALBERT. We combine a pointwise regression loss and a pairwise ranking loss which is more close to the final M atchm metric to finetune our models. And we also find that additional feature engineering and data augmentation can help improve the performance. Our best model achieves the highest score of 0.823 and ranks first for all kinds of metrics

cs.CL↗

arXiv2017

Online Robot Introspection via Wrench-based Action Grammars

Robotic failure is all too common in unstructured robot tasks. Despite well-designed controllers, robots often fail due to unexpected events. How do robots measure unexpected events? Many do not. Most robots are driven by the sense-plan act paradigm, however more recently robots are undergoing a sense-plan-act-verify paradigm. In this work, we present a principled methodology to bootstrap online robot introspection for contact tasks. In effect, we are trying to enable the robot to answer the question: what did I do? Is my behavior as expected or not? To this end, we analyze noisy wrench data and postulate that the latter inherently contains patterns that can be effectively represented by a vocabulary. The vocabulary is generated by segmenting and encoding the data. When the wrench information represents a sequence of sub-tasks, we can think of the vocabulary forming a sentence (set of words with grammar rules) for a given sub-task; allowing the latter to be uniquely represented. The grammar, which can also include unexpected events, was classified in offline and online scenarios as well as for simulated and real robot experiments. Multiclass Support Vector Machines (SVMs) were used offline, while online probabilistic SVMs were are used to give temporal confidence to the introspection result. The contribution of our work is the presentation of a generalizable online semantic scheme that enables a robot to understand its high-level state whether nominal or abnormal. It is shown to work in offline and online scenarios for a particularly challenging contact task: snap assemblies. We perform the snap assembly in one-arm simulated and real one-arm experiments and a simulated two-arm experiment. This verification mechanism can be used by high-level planners or reasoning systems to enable intelligent failure recovery or determine the next most optima manipulation skill to be used.

cs.RO↗

arXiv2016

Chinese/English mixed Character Segmentation as Semantic Segmentation

OCR character segmentation for multilingual printed documents is difficult due to the diversity of different linguistic characters. Previous approaches mainly focus on monolingual texts and are not suitable for multilingual-lingual cases. In this work, we particularly tackle the Chinese/English mixed case by reframing it as a semantic segmentation problem. We take advantage of the successful architecture called fully convolutional networks (FCN) in the field of semantic segmentation. Given a wide enough receptive field, FCN can utilize the necessary context around a horizontal position to determinate whether this is a splitting point or not. As a deep neural architecture, FCN can automatically learn useful features from raw text line images. Although trained on synthesized samples with simulated random disturbance, our FCN model generalizes well to real-world samples. The experimental results show that our model significantly outperforms the previous methods.

cs.CV↗

arXiv2016

Robot Introspection via Wrench-based Action Grammars

Robotic failure is all too common in unstructured robot tasks. Despite well designed controllers, robots often fail due to unexpected events. How do robots measure unexpected events? Many do not. Most robots are driven by the senseplan- act paradigm, however more recently robots are working with a sense-plan-act-verify paradigm. In this work we present a principled methodology to bootstrap robot introspection for contact tasks. In effect, we are trying to answer the question, what did the robot do? To this end, we hypothesize that all noisy wrench data inherently contains patterns that can be effectively represented by a vocabulary. The vocabulary is generated by meaningfully segmenting the data and then encoding it. When the wrench information represents a sequence of sub-tasks, we can think of the vocabulary forming sets of words or sentences, such that each subtask is uniquely represented by a word set. Such sets can be classified using statistical or machine learning techniques. We use SVMs and Mondrian Forests to classify contacts tasks both in simulation and in real robots for one and dual arm scenarios showing the general robustness of the approach. The contribution of our work is the presentation of a simple but generalizable semantic scheme that enables a robot to understand its high level state. This verification mechanism can provide feedback for high-level planners or reasoning systems that use semantic descriptors as well. The code, data, and other supporting documentation can be found at: http://www.juanrojas.net/2017icra_wrench_introspection.

cs.RO↗

arXiv2016

Robot Contact Task State Estimation via Action Grammars

Uncertainty is a major difficulty in endowing robots with autonomy. Robots often fail due to unexpected events. In robot contact tasks are often design to empirically look for force thresholds to define state transitions in a Markov chain or finite state machines. Such design is prone to failure in unstructured environments, when due to external disturbances or erroneous models, such thresholds are met, and lead to state transitions that are false-positives. The focus of this paper is to perform high-level state estimation of robot behaviors and task output for robot contact tasks. Our approach encodes raw low-level 3D cartesian trajectories and converts them into a high level (HL) action grammars. Cartesian trajectories can be segmented and encoded in a way that their dynamic properties, or "texture" are preserved. Once an action grammar is generated, a classifier is trained to detect current behaviors and ultimately the task output. The system executed HL state estimation for task output verification with an accuracy of 86%, and behavior monitoring with an average accuracy of: 72%. The significance of the work is the transformation of difficult-to-use raw low-level data to HL data that enables robust behavior and task monitoring. Monitoring is useful for failure correction or other deliberation in high-level planning, programming by demonstration, and human-robot interaction to name a few.

cs.RO↗