Missing data is a significant problem impacting all domains. State-of-the-art framework for minimizing missing data bias is multiple imputation, for which the choice of an imputation model remains nontrivial. We propose a multiple imputation model based on overcomplete deep denoising autoencoders. Our proposed model is capable of handling different data types, missingness patterns, missingness proportions and distributions. Evaluation on several real life datasets show our proposed model significantly outperforms current state-of-the-art methods under varying conditions while simultaneously improving end of the line analytics.

The performance of mobile AI accelerators has been evolving rapidly in the past two years, nearly doubling with each new generation of SoCs. The current 4th generation of mobile NPUs is already approaching the results of CUDAcompatible Nvidia graphics cards presented not long ago, which together with the increased capabilities of mobile deep learning frameworks makes it possible to run complex and deep AI models on mobile devices. In this paper, we evaluate the performance and compare the results of all chipsets from Qualcomm, HiSilicon, Samsung, MediaTek and Unisoc that are providing hardware acceleration for AI inference. We also discuss the recent changes in the Android ML pipeline and provide an overview of the deployment of deep learning models on mobile devices. All numerical results provided in this paper can be found and are regularly updated on the official project website 1 . * We also thank Oli Gaymond (ogaymond@google.com), Google Inc., for writing and editing section 3.1 of this paper. 1

Sequential pattern mining, which finds the set of frequent subsequences in sequence databases, is an important datamining task and has broad applications. IJsually, sequence patterns are associated with different circumstances, and such circumstances form a multiple dimensional space. For example, customer purchase sequences are associated with region, time, customer group, and others. It is interesting and useful to mine sequential patterns associated with multi-dimensional information.In this paper, we propose the theme of multi-dimensional sequential pattern mining, which integrates the multidimensional analysis and sequential data mining. We also thoroughly explore efficient methods for multi-dimensional sequential pattern mining. We examine feasible combinations of efficient sequential pattern mining and multidimensional analysis methods, as well as develop uniform methods for high-performance mining. Extensive experiments show the advantages as well as limitations of these methods. Some recommendations on selecting proper method with respect to data set properties are drawn.

Abstract. Many people rely on the recommendations of trusted friends to find restaurants or movies, which match their tastes. But, what if your friends have not sampled the item of interest? Collaborative filtering (CF) seeks to increase the effectiveness of this process by automating the derivation of a recommendation, often from a clique of advisors that we have no prior personal relationship with. CF is a promising tool for dealing with the information overload that we face in the networked world.Prior works in CF have dealt with improving the accuracy of the predictions. However, it is still challenging to scale these methods to large databases. In this study, we develop an efficient collaborative filtering method, called RecTree (which stands for RECommendation Tree) that addresses the scalability problem with a divide-and-conquer approach. The method first performs an efficient k-means-like clustering to group data and creates neighborhood of similar users, and then performs subsequent clustering based on smaller, partitioned databases. Since the progressive partitioning reduces the search space dramatically, the search for an advisory clique will be faster than scanning the entire database of users. In addition, the partitions contain users that are more similar to each other than those in other partitions. This characteristic allows RecTree to avoid the dilution of opinions from good advisors by a multitude of poor advisors and thus yielding a higher overall accuracy.Based on our experiments and performance study, RecTree outperforms the well-known collaborative filter, CorrCF, in both execution time and accuracy. In particular, RecTree's execution time scales by O(nlog 2 (n)) with the dataset size while CorrCF scales quadratically.

Abstract. In this paper, we propose an efficient algorithm, called TD-FPGrowth (the shorthand for Top-Down FP-Growth), to mine frequent patterns. TD-FP-Growth searches the FP-tree in the top-down order, as opposed to the bottom-up order of previously proposed FP-Growth. The advantage of the topdown search is not generating conditional pattern bases and sub-FP-trees, thus, saving substantial amount of time and space. We extend TD-FP-Growth to mine association rules by applying two new pruning strategies: one is to push multiple minimum supports and the other is to push the minimum confidence. Experiments show that these algorithms and strategies are highly effective in reducing the search space.