arXiv subjects

Meng Zhou

Publications and source records attributed to Meng Zhou.

49 records · Page 3Linked to original sources

arXiv2021

Self-supervised Regularization for Text Classification

Text classification is a widely studied problem and has broad applications. In many real-world problems, the number of texts for training classification models is limited, which renders these models prone to overfitting. To address this problem, we propose SSL-Reg, a data-dependent regularization approach based on self-supervised learning (SSL). SSL is an unsupervised learning approach which defines auxiliary tasks on input data without using any human-provided labels and learns data representations by solving these auxiliary tasks. In SSL-Reg, a supervised classification task and an unsupervised SSL task are performed simultaneously. The SSL task is unsupervised, which is defined purely on input texts without using any human-provided labels. Training a model using an SSL task can prevent the model from being overfitted to a limited number of class labels in the classification task. Experiments on 17 text classification datasets demonstrate the effectiveness of our proposed method.

cs.CL↗

arXiv2021

Robust Intensity Mapping Analysis against Foregrounds for the Epoch of Reionization

astro-ph.CO↗

arXiv2021

Antisymmetric Cross-correlation between H I and CO Line Intensity Maps as a New Probe of Cosmic Reionization

Intensity mapping of the H I 21 cm line and the CO 2.61 mm line from the epoch of reionization has emerged as powerful, complementary, probes of the high-redshift Universe. However, both maps and their cross-correlation are dominated by foregrounds. We propose a new analysis by which the signal is unbiased by foregrounds, i.e. it can be measured without foreground mitigation. We construct the antisymmetric part of two-point cross-correlation between intensity maps of the H I 21 cm line and the CO 2.61 mm line, arising because the statistical fluctuations of two fields have different evolution in time. We show that the sign of this new signal can distinguish model-independently whether inside-out reionization happens during some interval of time. More importantly, within the framework of the excursion set model of reionization, we demonstrate that the slope of the dipole of H I-CO cross-power spectrum at large scales is linear to the rate of change of global neutral fraction of hydrogen in a manner independent of reionization parameters, until the slope levels out near the end of reionization, but this trend might possibly depend on the framework of reionization modelling. The H I-CO dipole may be a smoking-gun probe for the speed of reionization, or "standard speedometer" for cosmic reionization. Observations of this new signal will unveil the global reionization history from the midpoint to near the completion of reionization.

astro-ph.CO↗

arXiv2020

Spatiotemporal Characteristics of Ride-sourcing Operation in Urban Area

The emergence of ride-sourcing platforms has brought an innovative alternative in transportation, radically changed travel behaviors, and suggested new directions for transportation planners and operators. This paper provides an exploratory analysis on the operations of a ride-sourcing service using large-scale data on service performance. Observations over multiple days in Singapore suggest reproducible demand patterns and provide empirical estimates of fleet operations over time and space. During peak periods, we observe significant increases in the service rate along with surge price multipliers. We perform an in-depth analysis of fleet utilization rates and are able to explain daily patterns based on drivers' behavior by involving the number of shifts, shift duration, and shift start and end time choices. We also evaluate metrics of user experience, namely waiting and travel time distribution, and explain our empirical findings with distance metrics from driver trajectory analysis and congestion patterns. Our results of empirical observations on actual service in Singapore can help to understand the spatiotemporal characteristics of ride-sourcing services and provide important insights for transportation planning and operations.

eess.SY↗

arXiv2020

Learning Implicit Credit Assignment for Cooperative Multi-Agent Reinforcement Learning

We present a multi-agent actor-critic method that aims to implicitly address the credit assignment problem under fully cooperative settings. Our key motivation is that credit assignment among agents may not require an explicit formulation as long as (1) the policy gradients derived from a centralized critic carry sufficient information for the decentralized agents to maximize their joint action value through optimal cooperation and (2) a sustained level of exploration is enforced throughout training. Under the centralized training with decentralized execution (CTDE) paradigm, we achieve the former by formulating the centralized critic as a hypernetwork such that a latent state representation is integrated into the policy gradients through its multiplicative association with the stochastic policies; to achieve the latter, we derive a simple technique called adaptive entropy regularization where magnitudes of the entropy gradients are dynamically rescaled based on the current policy stochasticity to encourage consistent levels of exploration. Our algorithm, referred to as LICA, is evaluated on several benchmarks including the multi-agent particle environments and a set of challenging StarCraft II micromanagement tasks, and we show that LICA significantly outperforms previous methods.

cs.LG↗

arXiv2020

Intelligent Exploration for User Interface Modules of Mobile App with Collective Learning

A mobile app interface usually consists of a set of user interface modules. How to properly design these user interface modules is vital to achieving user satisfaction for a mobile app. However, there are few methods to determine design variables for user interface modules except for relying on the judgment of designers. Usually, a laborious post-processing step is necessary to verify the key change of each design variable. Therefore, there is a only very limited amount of design solutions that can be tested. It is timeconsuming and almost impossible to figure out the best design solutions as there are many modules. To this end, we introduce FEELER, a framework to fast and intelligently explore design solutions of user interface modules with a collective machine learning approach. FEELER can help designers quantitatively measure the preference score of different design solutions, aiming to facilitate the designers to conveniently and quickly adjust user interface module. We conducted extensive experimental evaluations on two real-life datasets to demonstrate its applicability in real-life cases of user interface module design in the Baidu App, which is one of the most popular mobile apps in China.

cs.SE↗

arXiv2020

MedDialog: Two Large-scale Medical Dialogue Datasets

Medical dialogue systems are promising in assisting in telemedicine to increase access to healthcare services, improve the quality of patient care, and reduce medical costs. To facilitate the research and development of medical dialogue systems, we build two large-scale medical dialogue datasets: MedDialog-EN and MedDialog-CN. MedDialog-EN is an English dataset containing 0.3 million conversations between patients and doctors and 0.5 million utterances. MedDialog-CN is an Chinese dataset containing 1.1 million conversations and 4 million utterances. To our best knowledge, MedDialog-(EN,CN) are the largest medical dialogue datasets to date. The dataset is available at https://github.com/UCSD-AI4H/Medical-Dialogue-System

cs.LG↗

arXiv2020

CERT: Contrastive Self-supervised Learning for Language Understanding

Pretrained language models such as BERT, GPT have shown great effectiveness in language understanding. The auxiliary predictive tasks in existing pretraining approaches are mostly defined on tokens, thus may not be able to capture sentence-level semantics very well. To address this issue, we propose CERT: Contrastive self-supervised Encoder Representations from Transformers, which pretrains language representation models using contrastive self-supervised learning at the sentence level. CERT creates augmentations of original sentences using back-translation. Then it finetunes a pretrained language encoder (e.g., BERT) by predicting whether two augmented sentences originate from the same sentence. CERT is simple to use and can be flexibly plugged into any pretraining-finetuning NLP pipeline. We evaluate CERT on 11 natural language understanding tasks in the GLUE benchmark where CERT outperforms BERT on 7 tasks, achieves the same performance as BERT on 2 tasks, and performs worse than BERT on 2 tasks. On the averaged score of the 11 tasks, CERT outperforms BERT. The data and code are available at https://github.com/UCSD-AI4H/CERT

cs.CL↗

arXiv2019

Towards Understanding Chinese Checkers with Heuristics, Monte Carlo Tree Search, and Deep Reinforcement Learning

The game of Chinese Checkers is a challenging traditional board game of perfect information that differs from other traditional games in two main aspects: first, unlike Chess, all checkers remain indefinitely in the game and hence the branching factor of the search tree does not decrease as the game progresses; second, unlike Go, there are also no upper bounds on the depth of the search tree since repetitions and backward movements are allowed. Therefore, even in a restricted game instance, the state-space of the game can still be unbounded, making it challenging for a computer program to excel. In this work, we present an approach that effectively combines the use of heuristics, Monte Carlo tree search, and deep reinforcement learning for building a Chinese Checkers agent without the use of any human game-play data. Experiment results show that our agent is competent under different scenarios and reaches the level of experienced human players.

cs.LG↗

arXiv2018

Cascading Sensor Network Clock Synchronization Scheme

In this paper, we propose a hybrid clock synchronization architecture for a cascaded sensor network based on GPS time service and synchronous frame protocol. The sensor's upper-level unit is called sensor management unit (SMU) which synchronizes the local clock with the GPS and then use the broadcast sync frame to synchronize the clock of the sensor chain. In this paper, we build a link of 20 cascaded sensors to measure the center frequency and clock stability. The results showed that the center frequency deviation was 0.037 Hz and the stability was 0.045 ppm to 0.066 ppm .The stability of GPS is 0.06ppm, the scheme has achieved the effect similar to installing GPS chip on every sensor.

physics.ins-det↗

arXiv2018

Using Adjacent Data Retransmission to Improve the Transmission Efficiency of Multi-hop Relay Networks

Data transmission systems are widely used in various aspects of life, industry, research and other fields. Transmission systems have different characteristics for different application scenarios. Many of them require multi-hop transmission to extend the transmission range, such as Wireless Mesh Network (WMN), Mobile Ad-hoc Network (MANET), Wireless Sensor Networks (WSN) and so on. The biggest problem of multi-hop transmission lies in that packet loss caused by bit error ratio (BER) in multiple transmission processes increases greatly, which makes end-to-end transmission reliability and bandwidth utilization decrease. This paper analyzes the impact of BER on bandwidth utilization in multi-hop relay transmission and proposes a simple and flexible transmission mechanism that is generally applicable to multi-hop transmission scenarios.

eess.SP↗

arXiv2007

Existence of two electronic states in Sr4Ru3O10 at low temperatures

We report measurements on in-plane resistivity, thermopower, and magnetization as a function of temperature and magnetic fields on single crystalline Sr4Ru3O10 grown by the floating zone method. As the temperature was lowered to below around 30 K, the in-plane and c-axis resistivities and the thermopower were found to exhibit a step feature accompanied by hysteresis behavior when the in-plane field was swept up and down from below 10 kOe to above 20 kOe. The sharp increase in the thermopower with increasing in-plane magnetic field at low temperatures has not been observed previously in layered transition metal oxides. Comparing with magnetization data, we propose that the step feature marks the transition between the two different electronic states in Sr4Ru3O10. We propose that the alignment of domains by the in-plane magnetic field is responsible to the emergence of the new electronic states in high applied in-plane magnetic field.

cond-mat.str-el↗

arXiv2005

Electronic phase separation in the itinerant metamagnetic transition of Sr$_4$Ru$_3$O$_{10}$

Triple-layered ruthenate Sr$_4$Ru$_3$O$_{10}$ shows a first-order itinerant metamagnetic transition for in-plane magnetic fields. Our experiments revealed rather surprising behavior in the low-temperature transport properties near this transition. The in-plane magnetoresistivity $ρ$$_{ab}$(H) exhibits ultrasharp steps as the magnetic field sweeps down through the transition. Temperature sweeps of $ρ$$_{ab}$ for fields within the transition regime show non-metallic behavior in the up-sweep cycle of magnetic field, but show a significant drop in the down-sweep cycle. These observations indicate that the transition occurs via a new electronic phase separation process; a lowly polarized state is mixed with a ferromagnetic state within the transition regime.

cond-mat.str-el↗