Search

With internet delivery of video content surging to an un-precedented level, video recommendation has become an important approach for helping people access interesting videos. In this paper, we propose a novel approach to integrate viewing history for personalized video recommendation. For a given user, our approach calculates a recommendation score for each video candidate, that composes of two parts: the interest degree of this video by the user's friends, and the taste similarities between the user and his friends. We measure the interest degree of each video by considering its textual, visual and popularity information. Meanwhile, we construct tag set for each user based on his/her viewing history to estimate the taste similarities between different users. The final recommended videos are ranked according to the accumulated recommendation scores from different recommenders. We conduct experiments with 45 users and more than 11, 000 videos, and the results demonstrate the feasibility and effectiveness of our approach.

Video recommendation is an important approach for helping people to access interesting videos. In this paper, we propose a scheme to integrate rich information for video recommendation. We regard video recommendation as a ranking problem and generate multiple ranking lists by exploring different information sources. A multi-task rank aggregation approach is proposed to integrate the ranking lists for different users in a joint manner. Our scheme is flexible and can easily incorporate other methods by adding their generated ranking lists into our multi-task learning algorithm. We conduct experiments with 76 users and more than 10,000 videos. The results demonstrate the feasibility and effectiveness of our approach.

Automatic video recommendation is involved in an attempt to tackle the information-overload problem, aiming to present the personalized video list to the user. This paper presents a novel approach to improve the accuracy of the video recommendation by combining the content-based filtering (CBF) method and the collaborative filtering (CF) method. Multimodal information is utilized to calculate the similarity among different videos to overcome the sparseness problem by CF method. We conduct experiments on a dataset of more than 11,000 videos and the results demonstrate the feasibility and effectiveness of our approach.

Data arriving in time order (a data stream) arises in fields including physics, finance, medicine, and music, to name a few. Often the data comes from sensors (in physics and medicine for example) whose data rates continue to improve dramatically as sensor technology improves. Further, the number of sensors is increasing, so correlating data between sensors becomes ever more critical in order to distill knowlege from the data. In many applications such as finance, recent correlations are of far more interest than long-term correlation, so correlation over sliding windows (windowed correlation) is the desired operation. Fast response is desirable in many applications (e.g., to aim a telescope at an activity of interest or to perform a stock trade). These three factors -- data size, windowed correlation, and fast response -- motivate this work.Previous work [10, 14] showed how to compute Pearson correlation using Fast Fourier Transforms and Wavelet transforms, but such techniques don't work for time series in which the energy is spread over many frequency components, thus resembling white noise. For such "uncooperative" time series, this paper shows how to combine several simple techniques -- sketches (random projections), convolution, structured random vectors, grid structures, and combinatorial design -- to achieve high performance windowed Pearson correlation over a variety of data sets.

Financial time series streams are watched closely by millions of traders. What exactly do they look for and how can we help them do it faster? Physicists study the time series emerging from their sensors. The same question holds for them. Musicians produce time series. Consumers may want to compare them. This tutorial presents techniques and case studies for four problems:1. Finding sliding window correlations in financial, physical, and other applications.2. Discovering bursts in large sensor data of gamma rays.3. Matching hums to recorded music, even when people don't hum well.4. Maintaining and manipulating time-ordered data in a database setting.This tutorial draws mostly from the book High Performance Discovery in Time Series: techniques and case studies, Springer-Verlag 2004. You can find the power point slides for this tutorial at http://cs.nyu.edu/cs/faculty/shasha/papers/sigmod04.ppt.The tutorial is aimed at researchers in streams, data mining, and scientific computing. Its applications should interest anyone who works with scientists or financial "quants." The emphasis will be on recent results and open problems. This is a ripe area for further advance.