Learning Downstream Task by Selectively Capturing Complementary Knowledge from Multiple Self-supervisedly Learning Pretexts

no code implementations • 11 Apr 2022 • Jiayu Yao, Qingyuan Wu, Quan Feng, Songcan Chen

Self-supervised learning (SSL), as a newly emerging unsupervised representation learning paradigm, generally follows a two-stage learning pipeline: 1) learning invariant and discriminative representations with auto-annotation pretext(s), then 2) transferring the representations to assist downstream task(s).

Representation Learning Self-Supervised Learning

Topic Driven Adaptive Network for Cross-Domain Sentiment Classification

no code implementations • 28 Nov 2021 • Yicheng Zhu, Yiqiao Qiu, Qingyuan Wu, Fu Lee Wang, Yanghui Rao

In this vein, most approaches utilized domain adaptation that maps data from different domains into a common feature space.

Classification Domain Adaptation +3

Kuramoto model based analysis reveals oxytocin effects on brain network dynamics

no code implementations • 18 May 2021 • Shuhan Zheng, Zhichao Liang, Youzhi Qu, Qingyuan Wu, Haiyan Wu, Quanying Liu

Here, we propose a physics-based framework of Kuramoto model to investigate oxytocin effects on the phase dynamic neural coupling in DMN and FPN.

Greedy-Step Off-Policy Reinforcement Learning

no code implementations • 23 Feb 2021 • Yuhui Wang, Qingyuan Wu, Pengcheng He, Xiaoyang Tan

Most of the policy evaluation algorithms are based on the theories of Bellman Expectation and Optimality Equation, which derive two popular approaches - Policy Iteration (PI) and Value Iteration (VI).

Q-Learning reinforcement-learning +1