基于深度强化学习的国内金融市场投资比较研究

近年来,随着全球经济的迅速发展,参与金融投资的投资者增多,如何在复杂的金融市场中自动选择交易策略使收益最大化成为研究热点.强化学习可以通过与实际环境的交互来寻找最优的交易策略,使投资收益最大化.现有的方法大都是将一到两个强化学习算法应用于金融市场并比较算法在单一交易任务上的表现,此外,这些研究大都针对国外的股票、证券市场或加密货币市场,对国内金融市场的研究甚少.针对上述问题,面向国内金融投资市场,系统性地验证了不同类型的多种深度强化学习代表性算法在单只股票交易、多只股票交易和投资组合分配三个投资任务上的有效性.通过观察在累计收益率、夏普比率、最大回撤等评价指标上的回测结果对算法进行比较,结果显示在不同的投资任务中选取合适的强化学习算法可以有效地提升收益.

Graph CA: Learning From Graph Counterfactual Augmentation for Knowledge Tracing

With the popularity of online learning in educational settings, knowledge tracing(KT) plays an increasingly significant role. The task of KT is to help students learn more effectively by predicting their next mastery of knowledge based on their historical exercise sequences. Nowadays, many related works have emerged in this field, such as Bayesian knowledge tracing and deep knowledge tracing methods. Despite the progress that has been made in KT, existing techniques still have the following limitations: 1) Previous studies address KT by only exploring the observational sparsity data distribution, and the counterfactual data distribution has been largely ignored. 2) Current works designed for KT only consider either the entity relationships between questions and concepts, or the relations between two concepts, and none of them investigates the relations among students, questions, and concepts, simultaneously, leading to inaccurate student modeling. To address the above limitations,we propose a graph counterfactual augmentation method for knowledge tracing. Concretely, to consider the multiple relationships among different entities, we first uniform students, questions, and concepts in graphs, and then leverage a heterogeneous graph convolutional network to conduct representation learning.To model the counterfactual world, we conduct counterfactual transformations on students’ learning graphs by changing the corresponding treatments and then exploit the counterfactual outcomes in a contrastive learning framework. We conduct extensive experiments on three real-world datasets, and the experimental results demonstrate the superiority of our proposed Graph CA method compared with several state-of-the-art baselines.

基于多尺度注意力融合的知识追踪方法

互联网的普及使线上教育迅速发展,在缓解教育资源不均衡问题的同时,也为科研人员提供了大量的研究数据.教育数据挖掘是一个新兴学科,通过分析海量数据来理解学生的学习行为,为学生提供个性化学习建议.知识追踪是教育数据挖掘中的重要任务,其利用学生的历史答题序列预测学生下一次的答题表现.已有的知识追踪模型没有区分历史序列中的长期交互信息和短期交互信息,忽略了不同时间尺度的序列信息对未来预测的不同影响.针对该问题,提出一种基于多尺度注意力融合的知识追踪模型,使用时间卷积网络捕获历史交互序列的不同时间尺度信息,并基于注意力机制进行多尺度信息融合.针对不同学生及答题序列,该模型能自适应地确定不同时间尺度信息的重要性.实验结果表明,提出模型的性能优于已有的知识追踪模型.

Mesoscale wind stress-SST coupling induced feedback to the ocean in the western coast of South America

The feedback induced by mesoscale wind stress-SST coupling to the ocean in the western coast of South America was studied using the Regional Ocean Modeling System(ROMS).To represent the feedback,an empirical mesoscale wind stress perturbation model was constructed from satellite observations,and was incorporated into the ocean model.Comparing two experiments with and without the mesoscale wind stress-SST coupling,it was found that SST in the mesoscale coupling experiment was reduced in the western coast of South America,with the maximum values of 0.5℃in the Peru Sea and 0.7℃in the Chile Sea.A mixed layer heat budget analysis indicates that horizontal advection is the main term that explains the reduction in SST.Specifically,the feedback induced by mesoscale wind stress-SST coupling to the ocean can enhance vertical velocity in the nearshore area through the Ekman pumping,which brings subsurface cold water to the sea surface.These results indicate that the feedback due to the mesoscale wind stress-SST coupling to the ocean has the potential for reducing the warm SST bias often seen in the large-scale climate model simulations in this region.

Robust Core Tensor Dictionary Learning with Modified Gaussian Mixture Model for Multispectral Image Restoration

The multispectral remote sensing image(MS-RSI)is degraded existing multi-spectral camera due to various hardware limitations.In this paper,we propose a novel core tensor dictionary learning approach with the robust modified Gaussian mixture model for MS-RSI restoration.First,the multispectral patch is modeled by three-order tensor and high-order singular value decomposition is applied to the tensor.Then the task of MS-RSI restoration is formulated as a minimum sparse core tensor estimation problem.To improve the accuracy of core tensor coding,the core tensor estimation based on the robust modified Gaussian mixture model is introduced into the proposed model by exploiting the sparse distribution prior in image.When applied to MS-RSI restoration,our experimental results have shown that the proposed algorithm can better reconstruct the sharpness of the image textures and can outperform several existing state-of-the-art multispectral image restoration methods in both subjective image quality and visual perception.

Multi-task MIML learning for pre-course student performance prediction

In higher education,the initial studying period of each course plays a crucial role for students,and seriously influences the subsequent learning activities.However,given the large size of a course’s students at universities,it has become impossible for teachers to keep track of the performance of individual students.In this circumstance,an academic early warning system is desirable,which automatically detects students with difficulties in learning(i.e.,at-risk students)prior to a course starting.However,previous studies are not well suited to this purpose for two reasons:1)they have mainly concentrated on e-learning platforms,e.g.,massive open online courses(MOOCs),and relied on the data about students’online activities,which is hardly accessed in traditional teaching scenarios;and 2)they have only made performance prediction when a course is in progress or even close to the end.In this paper,for traditional classroom-teaching scenarios,we investigate the task of pre-course student performance prediction,which refers to detecting at-risk students for each course before its commencement.To better represent a student sample and utilize the correlations among courses,we cast the problem as a multi-instance multi-label(MIML)problem.Besides,given the problem of data scarcity,we propose a novel multi-task learning method,i.e.,MIML-Circle,to predict the performance of students from different specialties in a unified framework.Extensive experiments are conducted on five real-world datasets,and the results demonstrate the superiority of our approach over the state-of-the-art methods.

Evaluating and improving the interpretability of item embeddings using item-tag relevance information

Matrix factorization(MF)methods have superior recommendation performance and are flexible to incorporate other side information,but it is hard for humans to interpret the derived latent factors.Recently,the item-item cooccurrence information is exploited to learn item embeddings and enhance the recommendation performance.However,the item-item co-occurrence information,constructed from the sparse and long-tail distributed user-item interaction matrix,is over-estimated for rare items,which could lead to bias in learned item embeddings.In this paper,we seek to evaluate and improve the interpretability of item embeddings by leveraging a dense item-tag relevance matrix.Specifically,we design two metrics to quantitatively evaluate the interpretability of item embeddings from different viewpoints:interpretability of individual dimensions of item embeddings and semantic coherence of local neighborhoods in the latent space.We also propose a tag-informed item embedding(TIE)model that jointly factorizes the user-item interaction matrix,the item-item co-occurrence matrix and the item-tag relevance matrix with shared item embeddings so that different forms of information can co-operate with each other to learn better item embeddings.Experiments on the MovieLens20M dataset demonstrate that compared with other state-of-the-art MF methods,TIE achieves better top-N recommendations,and the relative improvement is larger when the user-item interaction matrix becomes sparser.By leveraging the itemtag relevance information,individual dimensions of item embeddings are more interpretable and local neighborhoods in the latent space are more semantically coherent;the bias in learned item embeddings are also mitigated to some extent.