| {"6021855|2670121": []} | |
| {"11392154|393948": []} | |
| {"201003|2670121": []} | |
| {"3761015|2670121": ["As a reference, the survey paper by Cook [19] covers transfer learning for the application of activity recognition and the survey papers by Patel [89] and Shao [103] address transfer learning in the domain of image recognition.", "Cook [19] and Feuz [36] provide a different variation where the definition of supervised or unsupervised refers to the presence or absence of labeled data in the source domain and informed or uninformed refers to the presence or absence of labeled data in the target domain."]} | |
| {"10213552|3750204": ["7.0 Batch Java Mahout \u00d7 \u00d7 \u2713 \u00d7 Spark 1.3.1 Batch, streaming Java, Python, R, Scala MLlib, Mahout, H 2 O \u2713 \u2713 \u2713 \u2713 Flink 0.8.1 Batch, streaming Java, Scala Flink-ML, SAMOA \u2713 \u2713 \u2713 \u00d7 Storm 0.9.4 Streaming Any SAMOA \u2713 \u2713 \u2713 \u00d7 H 2 O 3.0.0.12 Batch Java, Python, R, Scala H 2 O, Mahout, MLlib \u2713 \u2713 \u2713 \u2713 6.", "Mahout is one of the more well-known tools for ML.", "On the other hand, a number of companies have reported success using Mahout in production.", "Additionally, while a lot of documentation exists for Mahout, much of it is outdated and irrelevant to people using the current version .", "In April 2015, Mahout 0.9 was updated to 0.10.0, marking something of a shift in the project's goals [92].", "on Spark\u2019s iterative batch and streaming approaches, as well as its use of in-memory computation, enable jobs to run significantly faster than those using Mahout [88, 127].", "One of Mahout's most commonly cited assets is its extensibility and many have achieved good results by building off of the baseline algorithms [87, 99, 100].", "For machine learning tasks, Spark ships with the MLlib [60] and GraphX [61] libraries and the latest version of the Mahout [62] library offers a number of Spark implementations as well.", "Mahout includes several implementations that are not distributed and therefore not included, but they are discussed in the \" Evaluation of machine learning tools \" section.", "The algorithms included in Mahout focus primarily on classification, clustering and collaborative filtering, and have been shown to scale well as the size of the data increases [98].", "The goal of the Mahout-Samsara project is to help users build their own distributed Fig.", "Usability may be considered in terms of initial setup, ongoing maintenance, programming languages available, user interface available, amount Table 2 Overview of machine learning toolkits a Real-time streaming implementation b Single machine, trained using Stochastic gradient descent Mahout MLlib H", "MapReduce is compatible with the Mahout library for ML, and the programming interfaces discussed in the previous section can be used as well.", "One problem with using Mahout in production is that development has moved very slowly; version 0.10.0 was released nearly seven and a half years after the project was initially introduced.", "Among the more commonly cited complaints about Mahout is that it is difficult to set up on an existing Hadoop cluster [93\u201395]."]} | |
| {"7322495|192934": ["We conducted experiments on two types of data, text data and image data, using Amazon product reviews [15] and UCI handwritten digits [11] respectively."]} | |
| {"12935955|7704963": ["Similar techniques can also be found in the research of uncertain data mining [1]."]} | |
| {"11067362|5940452": ["mation literature of using genetic algorithms to automatically learn weights for different parameters [43]."]} | |
| {"15494001|11894197": ["A great deal of work has been conducted on various filter-based approaches to feature selection [7, 10, 27, 21]."]} | |
| {"8500384|14356173": ["The existing literature incorporating statistical or machine learning techniques to detect outliers in healthcare is fairly extensive, using common methods such as decision trees or neural networks [26], [24], [29]."]} | |
| {"8500384|27569632": ["While this paper does not provide a comprehensive discussion on outlier detection methods, we refer the reader to [10], [34]."]} | |
| {"7857907|2670121": ["Figure 1 illustrates the layout and sensor locations for three smart homes from the CASAS datasets [Cook et al., 2013c] used in this paper.", "However, for applications such as sentiment analysis across different languages [Pan, 2010], and activity recognition across different domains [Cook et al., 2013b], the source", "Transfer learning approaches have found success in many applications including activity recognition [Hu et al., 2011; Cook et al., 2013c], sentiment classification [Zhou et al., 2014], document analysis and indoor localization [Pan and Yang, 2010].", "The CASAS dataset [Cook et al., 2013a] is a collection of smart home datasets that are widely used for investigating activity recognition algorithms.", "However, for applications such as sentiment analysis across different languages [Pan, 2010], and activity recognition across different domains [Cook et al., 2013b], the source \u21e4 The author contributed to this work during his internship at IIT Ropar. and target data are represented using\u2026"]} | |
| {"14108784|15888893": ["For example, to predict user behaviors for a new product A where there is not enough training data, but there is useful data from a similar product B, can we use the knowledge learned from B to support predictions about A? Transfer learning has been proposed to address this problem [1,2] and has garnered tremendous attention from researchers [3\u20136]."]} | |