基于LSTM的POI个性化推荐框架

近年来,随着基于位置的社会网络（Location-Based Social Network,LBSN）热度的不断增加,为用户推荐下一个POI （Point-Of-Interests）也显得越来越重要.而对应的各种应用搜集到的用户的行为时间、地理、好友以及标签等信息的增多,使得POI推荐变得更加容易.目前针对POI推荐,已经有部分算法提出,但是他们受限于自身的局限性,还都不能很好的解决这个问题,例如,个性化马尔科夫链（Factorizing Personalized Markov Chain,FPMC）、张量分解（Tensor Factorization,TF）、RNN （Recurrent Neural Networks）等.但是,这些模型由于其本身缺陷,都不能完美的糅合POI场景中的所有信息.在这篇文章中,我们扩展了长短时记忆循环神经网络（Long-ShorT Memory recurrent neural networks,LSTM）,提出一种全新的推荐框架POI-LSTM来解决POI推荐问题.POI-LSTM借鉴Embedding的思想,对用户信息、好友关系、POI信息和评论信息进行向量化后,输入到神经网络中,同时利用LSTM捕捉用户的兴趣特征和兴趣的变化趋势,最终能够在不同的输入层拟合社交网络信息和语义信息,同时利用用户历史行为的时间和地理位置信息来为用户推荐下一个兴趣点.     

Diffuse Phase Transition of the Sr0.3−3y/2LayBa0.7Nb2O6 Ceramic System

The effect of Lanthanum cation doping in the diffuse phase transition of the Sr_0.3–3y/2La_yBa_0.7Nb₂O₆ ceramic system is studied for y=0.01, 0.03 and 0.05. For these compositions, the transition temperature T _c shifts towards higher values with increasing frequency and the temperature dependence of the electrical permittivity presents strongly broadened curves which suggest a non Curie-Weiss behavior near the transition temperature for temperatures far from T _c. The diffuse phase transition coefficient () was also determined and its values lead us to conclude that the degree of disorder in the system increases with the presence of the Lanthanum cation. This result is corroborated by calorimetric measurements, where an increase in entropy with the Lanthanum concentration is also found.