Implicit feedback, which indirectly reflects opinion through user behaviors, has gained increasing attention in recommender system communities due to its accessibility and richness in real-world applications. A major ...Implicit feedback, which indirectly reflects opinion through user behaviors, has gained increasing attention in recommender system communities due to its accessibility and richness in real-world applications. A major way of exploiting implicit feedback is to treat the data as an indication of positive and negative preferences associated with vastly varying confidence levels. Such algorithms assume that the numerical value of implicit feedback, such as time of watching, indicates confidence, rather than degree of preference, and a larger value indicates a higher confidence, although this works only when just one type of implicit feedback is available. However, in real-world applications, there are usually various types of implicit feedback, which can be referred to as heterogeneous implicit feedback. Existing methods cannot efficiently infer confidence levels from heterogeneous implicit feedback. In this paper, we propose a novel confidence estimation approach to infer the confidence level of user preference based on heterogeneous implicit feedback. Then we apply the inferred confidence to both point-wise and pair-wise matrix factorization models, and propose a more generic strategy to select effective training samples for pair-wise methods. Experiments on real-world e-commerce datasets from Tmall.com show that our methods outperform the state-of-the-art approaches, consid- ering several commonly used ranking-oriented evaluation criteria.展开更多
Humans are able to overcome sensory perturbations imposed on their movements through motor learning. One of the key mechanisms to accomplish this is sensorimotor adaptation, an implicit, error-driven learning mechanis...Humans are able to overcome sensory perturbations imposed on their movements through motor learning. One of the key mechanisms to accomplish this is sensorimotor adaptation, an implicit, error-driven learning mechanism. Past work on sensorimotor adaptation focused mainly on adaptation to rotated visual feedback—A paradigm known as visuomotor rotation. Recent studies have shown that sensorimotor adaptation can also occur under mirror-reversed visual feedback. In visuomotor rotation, sensorimotor adaptation can be driven by both endpoint and online feedback [1] [2]. However, it’s not been clear whether both kinds of feedback can similarly drive adaptation under a mirror reversed perturbation. We performed a study to establish what kinds of feedback can drive adaptation under mirror reversal. In the first two conditions, the participants were asked to ignore visual feedback. In the first condition, we provided mirror reversed online feedback and endpoint feedback. We reproduced previous findings showing that online feedback elicited adaptation under mirror reversal. In a second condition, we provided mirror reversed endpoint feedback. However, in the second condition, we found that endpoint feedback alone failed to elicit adaptation. In a third condition, we provided both types of feedback at the same time, but in a conflicting way: endpoint feedback was non-reversed while online feedback was mirror reversed. The participants were asked to ignore online visual feedback and try to hit the target with help from veridical endpoint feedback. In the third condition, in which veridical endpoint feedback and mirror reversed online feedback were provided, adaptation still occurred. Our results showed that endpoint feedback did not elicit adaptation under mirror reversal but online feedback did. This dissociation between effects of endpoint feedback and online feedback on adaptation under mirror reversal suggests that adaptation under these different kinds of feedback might in fact operate via distinct mechanisms.展开更多
The paper introduces effective and straightforward algorithms of both explicit and implicit model-following designs with state derivative measurement feedback in novel reciprocal state space form (RSS) to handle state...The paper introduces effective and straightforward algorithms of both explicit and implicit model-following designs with state derivative measurement feedback in novel reciprocal state space form (RSS) to handle state derivative related performance output and state related performance output design cases. Applying proposed algorithms, no integrators are required. Consequently, implementation is simple and low-cost. Simulation has also been carried out to verify the proposed algorithms. Since acceleration can only be modeled as state derivative in state space form and micro-accelerometer which is the state derivative sensor is getting more and more attentions in many microelectromechanical and nanoelectromechanical systems (MEMS/NEMS) applications, the proposed algorithms are suitable for MEMS/NEMS systems installed with micro-accelerometers.展开更多
现有推荐系统通常采用评分、评论等显式反馈数据实现个性化推荐.然而,显式反馈数据由于在实际中难以获取或因质量问题而往往变得不可用,从而导致相关推荐算法的应用范围受到很大限制.与此相反,诸如点击行为、浏览记录等隐式反馈数据在...现有推荐系统通常采用评分、评论等显式反馈数据实现个性化推荐.然而,显式反馈数据由于在实际中难以获取或因质量问题而往往变得不可用,从而导致相关推荐算法的应用范围受到很大限制.与此相反,诸如点击行为、浏览记录等隐式反馈数据在现实中大量存在.本文提出了一种面向游戏玩家的基于隐式反馈数据的游戏推荐方法.该方法综合考虑了玩家操作次数、操作时长等隐式反馈数据及其时效性,构建了基于伪评分的玩家对游戏的偏好模型,而后通过改进了的SVD++(Singular Value Decomposition++)算法实现个性化游戏推荐.在大规模真实数据集上的实验结果表明本文提出的方法具有更高的推荐精确率和召回率.展开更多
作为解决信息过载问题的有效方式,推荐系统能够根据用户偏好对海量信息进行过滤,为用户提供个性化的推荐。对如何利用隐式反馈数据进行个性化推荐进行了研究,提出了一种融合上下文信息和用户社交信息的隐式反馈推荐模型(Implicit Feedba...作为解决信息过载问题的有效方式,推荐系统能够根据用户偏好对海量信息进行过滤,为用户提供个性化的推荐。对如何利用隐式反馈数据进行个性化推荐进行了研究,提出了一种融合上下文信息和用户社交信息的隐式反馈推荐模型(Implicit Feedback Recommendation Model Fusing Context-aware and Social Network Process,IFCSP)。首先从数据集中提取与用户兴趣相关的上下文信息的属性集合,并以此作为分裂属性,使用决策树分类算法对"用户-产品-上下文"集合进行分类,从而将历史选择集合分组。对于要推荐的用户,根据其选择产品时的上下文信息,匹配最相似的分组,再使用基于隐式反馈的推荐模型(Implicit Feedback Recommendation Model,IFRM)预测用户对未选择产品的偏好,并结合用户的社交信息,进而对用户进行产品推荐。实验表明,该模型在平均正确率均值(MAP)和平均百分百排序(MPR)评价指标上均优于其他4种算法,可以显著提高系统的预测和推荐质量。展开更多
针对隐式数据单纯利用隐反馈信息往往难以获取较好推荐性能的问题,提出一种融合元数据及隐式反馈信息的多层次深度联合学习(multi-level deep joint learning,MDJL)推荐方法。它利用双深度神经网络共同学习,其中一个网络利用隐式反馈学...针对隐式数据单纯利用隐反馈信息往往难以获取较好推荐性能的问题,提出一种融合元数据及隐式反馈信息的多层次深度联合学习(multi-level deep joint learning,MDJL)推荐方法。它利用双深度神经网络共同学习,其中一个网络利用隐式反馈学习用户及项目个体个性化关系,另一个网络利用元数据学习高层次群体共性化关系,从而有效地表达用户偏好,使MDJL框架在个体及群体因素间达到平衡。最后,MDJL推荐算法在Movie Lens 100K和MovieLens 1M两个公开数据集上进行实验评估。结果表明,该算法比其他基线方法表现出了更为优越的推荐性能。展开更多
基金supported by the National Basic Research Program(973) of China(No.2015CB352400)the National Key Research and Development Program of China(No.2016YFB1200203-03)
文摘Implicit feedback, which indirectly reflects opinion through user behaviors, has gained increasing attention in recommender system communities due to its accessibility and richness in real-world applications. A major way of exploiting implicit feedback is to treat the data as an indication of positive and negative preferences associated with vastly varying confidence levels. Such algorithms assume that the numerical value of implicit feedback, such as time of watching, indicates confidence, rather than degree of preference, and a larger value indicates a higher confidence, although this works only when just one type of implicit feedback is available. However, in real-world applications, there are usually various types of implicit feedback, which can be referred to as heterogeneous implicit feedback. Existing methods cannot efficiently infer confidence levels from heterogeneous implicit feedback. In this paper, we propose a novel confidence estimation approach to infer the confidence level of user preference based on heterogeneous implicit feedback. Then we apply the inferred confidence to both point-wise and pair-wise matrix factorization models, and propose a more generic strategy to select effective training samples for pair-wise methods. Experiments on real-world e-commerce datasets from Tmall.com show that our methods outperform the state-of-the-art approaches, consid- ering several commonly used ranking-oriented evaluation criteria.
文摘Humans are able to overcome sensory perturbations imposed on their movements through motor learning. One of the key mechanisms to accomplish this is sensorimotor adaptation, an implicit, error-driven learning mechanism. Past work on sensorimotor adaptation focused mainly on adaptation to rotated visual feedback—A paradigm known as visuomotor rotation. Recent studies have shown that sensorimotor adaptation can also occur under mirror-reversed visual feedback. In visuomotor rotation, sensorimotor adaptation can be driven by both endpoint and online feedback [1] [2]. However, it’s not been clear whether both kinds of feedback can similarly drive adaptation under a mirror reversed perturbation. We performed a study to establish what kinds of feedback can drive adaptation under mirror reversal. In the first two conditions, the participants were asked to ignore visual feedback. In the first condition, we provided mirror reversed online feedback and endpoint feedback. We reproduced previous findings showing that online feedback elicited adaptation under mirror reversal. In a second condition, we provided mirror reversed endpoint feedback. However, in the second condition, we found that endpoint feedback alone failed to elicit adaptation. In a third condition, we provided both types of feedback at the same time, but in a conflicting way: endpoint feedback was non-reversed while online feedback was mirror reversed. The participants were asked to ignore online visual feedback and try to hit the target with help from veridical endpoint feedback. In the third condition, in which veridical endpoint feedback and mirror reversed online feedback were provided, adaptation still occurred. Our results showed that endpoint feedback did not elicit adaptation under mirror reversal but online feedback did. This dissociation between effects of endpoint feedback and online feedback on adaptation under mirror reversal suggests that adaptation under these different kinds of feedback might in fact operate via distinct mechanisms.
文摘The paper introduces effective and straightforward algorithms of both explicit and implicit model-following designs with state derivative measurement feedback in novel reciprocal state space form (RSS) to handle state derivative related performance output and state related performance output design cases. Applying proposed algorithms, no integrators are required. Consequently, implementation is simple and low-cost. Simulation has also been carried out to verify the proposed algorithms. Since acceleration can only be modeled as state derivative in state space form and micro-accelerometer which is the state derivative sensor is getting more and more attentions in many microelectromechanical and nanoelectromechanical systems (MEMS/NEMS) applications, the proposed algorithms are suitable for MEMS/NEMS systems installed with micro-accelerometers.
文摘现有推荐系统通常采用评分、评论等显式反馈数据实现个性化推荐.然而,显式反馈数据由于在实际中难以获取或因质量问题而往往变得不可用,从而导致相关推荐算法的应用范围受到很大限制.与此相反,诸如点击行为、浏览记录等隐式反馈数据在现实中大量存在.本文提出了一种面向游戏玩家的基于隐式反馈数据的游戏推荐方法.该方法综合考虑了玩家操作次数、操作时长等隐式反馈数据及其时效性,构建了基于伪评分的玩家对游戏的偏好模型,而后通过改进了的SVD++(Singular Value Decomposition++)算法实现个性化游戏推荐.在大规模真实数据集上的实验结果表明本文提出的方法具有更高的推荐精确率和召回率.
文摘作为解决信息过载问题的有效方式,推荐系统能够根据用户偏好对海量信息进行过滤,为用户提供个性化的推荐。对如何利用隐式反馈数据进行个性化推荐进行了研究,提出了一种融合上下文信息和用户社交信息的隐式反馈推荐模型(Implicit Feedback Recommendation Model Fusing Context-aware and Social Network Process,IFCSP)。首先从数据集中提取与用户兴趣相关的上下文信息的属性集合,并以此作为分裂属性,使用决策树分类算法对"用户-产品-上下文"集合进行分类,从而将历史选择集合分组。对于要推荐的用户,根据其选择产品时的上下文信息,匹配最相似的分组,再使用基于隐式反馈的推荐模型(Implicit Feedback Recommendation Model,IFRM)预测用户对未选择产品的偏好,并结合用户的社交信息,进而对用户进行产品推荐。实验表明,该模型在平均正确率均值(MAP)和平均百分百排序(MPR)评价指标上均优于其他4种算法,可以显著提高系统的预测和推荐质量。
