趋势比率模型在卷烟预测中的应用

中国烟草行业具有行政上的垄断性和生产上的计划性，有别于其它行业。做好卷烟销量的预测，是当前烟草行业工业生产环节与商业环节协同平滑进展的前提。鉴于此，提出基于趋势比率的卷烟预测模型，对该预测模型设计了相应的算法，最后以实际数据为例验证了趋势比率模型预测方法的有效性和实用性。  

基于混合选择和动态变异的基因表达式编程

基因表达式编程（GEP）是一种基于基因型和表现型的新型遗传算法，目前被广泛应用在函数发现、时间序列预测和分类等领域。传统GEP算法采用轮盘赌方式来选择种群个体，其择优强度过大，易导致个体多样性减弱，产生“近亲繁殖”；种群个体的变异概率固定，变异幅度不能动态地适应每代的进化结果，影响进化效率。针对上述两个缺陷，本文对传统GEP做出两点改进：作者采用混合选择策略，以维持进化过程中个体的多样性，避免“近亲繁殖”；引入动态变异思想，使种群在进化过程中能根据自身适应性的高低来动态调整个体的变异概率，以最大限度地保留高适应度基因片段，消除低适应度基因片段。通过实验，本文验证了两项改进的有效性。  

The weak aggregating algorithm and weak mixability

This paper resolves the problem of predicting as well as the best expert up to an additive term of the order o(n), where n is the length of a sequence of letters from a finite alphabet. We call the games that permit this weakly mixable and give a geometrical characterisation of the class of weakly mixable games. Weak mixability turns out to be equivalent to convexity of the finite part of the set of superpredictions. For bounded games we introduce the Weak Aggregating Algorithm that allows us to obtain additive terms of the form .  

On semimeasures predicting Martin-Löf random sequences

Solomonoff’s central result on induction is that the prediction of a universal semimeasure M$M$ converges rapidly and with probability 1 to the true sequence generating predictor μ$\mu$, if the latter is computable. Hence, M$M$ is eligible as a universal sequence predictor in the case of unknown μ$\mu$. Despite some nearby results and proofs in the literature, the stronger result of convergence for all (Martin-Löf) random sequences remained open. Such a convergence result would be particularly interesting and natural, since randomness can be defined in terms of M$M$ itself. We show that there are universal semimeasures M$M$ which do not converge to μ$\mu$ on all μ$\mu$-random sequences, i.e. we give a partial negative answer to the open problem. We also provide a positive answer for some non-universal semimeasures. We define the incomputable measure D$D$ as a mixture over all computable measures and the enumerable semimeasure W$W$ as a mixture over all enumerable nearly measures. We show that W$W$ converges to D$D$ and D$D$ to μ$\mu$ on all random sequences. The Hellinger distance measuring closeness of two distributions plays a central role.  

On universal prediction and Bayesian confirmation

The Bayesian framework is a well-studied and successful framework for inductive reasoning, which includes hypothesis testing and confirmation, parameter estimation, sequence prediction, classification, and regression. But standard statistical guidelines for choosing the model class and prior are not always available or can fail, in particular in complex situations. Solomonoff completed the Bayesian framework by providing a rigorous, unique, formal, and universal choice for the model class and the prior. I discuss in breadth how and in which sense universal (non-i.i.d.) sequence prediction solves various (philosophical) problems of traditional Bayesian sequence prediction. I show that Solomonoff’s model possesses many desirable properties: strong total and future bounds, and weak instantaneous bounds, and, in contrast to most classical continuous prior densities, it has no zero p(oste)rior problem, i.e. it can confirm universal hypotheses, is reparametrization and regrouping invariant, and avoids the old-evidence and updating problem. It even performs well (actually better) in non-computable environments.  

Improved second-order bounds for prediction with expert advice

This work studies external regret in sequential prediction games with both positive and negative payoffs. External regret measures the difference between the payoff obtained by the forecasting strategy and the payoff of the best action. In this setting, we derive new and sharper regret bounds for the well-known exponentially weighted average forecaster and for a second forecaster with a different multiplicative update rule. Our analysis has two main advantages: first, no preliminary knowledge about the payoff sequence is needed, not even its range; second, our bounds are expressed in terms of sums of squared payoffs, replacing larger first-order quantities appearing in previous bounds. In addition, our most refined bounds have the natural and desirable property of being stable under rescalings and general translations of the payoff sequence. Editor: Avrim Blum An extended abstract appeared in the Proceedings of the 18th Annual Conference on Learning Theory, Springer, 2005. The work of all authors was supported in part by the IST Programme of the European Community, under the PASCAL Network of Excellence, IST-2002-506778. The work was done while Yishay Mansour was a fellow in the Institute of Advance studies, Hebrew University. His work was also supported by a grant no. 1079/04 from the Israel Science Foundation and an IBM faculty award.  

RCGNN：一种基于实数编码的遗传神经网络预测方法

将遗传算法与神经网络相结合，提出一种实数编码、自适应选择、算术交叉、高斯变异、爬山操作的改进遗传BP神经网络RCGNN，利用遗传算法对神经网络权值和阈值进行优化。以时间序列预测的实例进行编程计算表明，用遗传算法进行网络训练，其收敛速度快，最终总误差最小，预测准确率高。对算法中参数进行的相应研究表明，增加爬山操作次数能很好地提高网络训练的速度，同时使误差下降快；爬山操作越多，收敛速度越快，最终误差越小，但计算运行时间也会增加。  

基于深度学习的时空序列预测方法综述

随着数据采集技术的蓬勃发展，各个领域的时空数据不断累积，迫切需要探索高效的时空数据预测方法。深度学习是一种基于人工神经网络的机器学习方法，能有效地处理大规模的复杂数据，因而研究基于深度学习的时空序列预测方法具有十分重要的意义。在这一背景下，针对已有的预测方法进行归纳和总结，首先回顾了深度学习在时空序列预测中的应用背景和发展历程，介绍了时空序列的相关定义、特点及分类；然后按照时空序列数据的类别介绍了基于网格数据的预测方法、基于图数据的预测方法和基于轨迹数据的预测方法；最后总结了上述预测方法，并对当前面临的一些问题及可能的解决方案进行了探讨。  

面向电子商务平台用户意图预测的时间感知分层自注意力网络

在推荐系统领域，了解电商平台中在线用户的行为意图至关重要。目前的一些方法通常将用户与商品之间的交互历史数据视为有序的序列，却忽视了不同交互行为之间的时间间隔信息。另外，一个用户的在线行为可能不仅仅包含一种意图，而是包含多种意图。例如，当一位用户在浏览运动品类下的商品时，其可能同时有购买足球和运动衫这两种商品的意图。但是现有的一些电商平台用户意图预测方法很难有效对用户-商品交互对时间间隔信息进行建模，也难以捕捉用户多方面的购物意图。为了解决上述问题，我们提出了一种时间感知分层自注意力网络模型THSNet，以更有效对电商平台的用户意图进行预测。具体而言，THSNet模型采用一种分层注意力机制来有效地捕获用户-商品交互历史中的时间跨度信息以更有效建模用户的多种意图。THSNet模型的注意力层分为两层，底层的注意力层用于建模每个会话内部的用户-商品交互，上层的注意力层学习不同会话之间的长期依赖关系。另外，为了提高预测结果的鲁棒性和准确度，我们采用BERT预训练的方法，通过随机遮盖部分会话的特征表示，构造了一个完形填空任务，并将该任务与用户意图预测任务耦合成为多任务学习模型，这种多任务预测方法有助于模型学到一个具有鲁棒性和双向性的会话特征表示。我们在两个真实数据集上对所提方法对有效性进行了验证。实验结果表明，我们所提出的THSNet模型要明显优于目前最先进的方法。  

On generalized computable universal priors and their convergence

Solomonoff unified Occam's razor and Epicurus’ principle of multiple explanations to one elegant, formal, universal theory of inductive inference, which initiated the field of algorithmic information theory. His central result is that the posterior of the universal semimeasure M   converges rapidly to the true sequence generating posterior μ$\mu$, if the latter is computable. Hence, M   is eligible as a universal predictor in case of unknown μ$\mu$. The first part of the paper investigates the existence and convergence of computable universal (semi) measures for a hierarchy of computability classes: recursive, estimable, enumerable, and approximable. For instance, M is known to be enumerable, but not estimable, and to dominate all enumerable semimeasures. We present proofs for discrete and continuous semimeasures. The second part investigates more closely the types of convergence, possibly implied by universality: in difference and in ratio, with probability 1, in mean sum, and for Martin-Löf random sequences. We introduce a generalized concept of randomness for individual sequences and use it to exhibit difficulties regarding these issues. In particular, we show that convergence fails (holds) on generalized-random sequences in gappy (dense) Bernoulli classes.