PDF Publication Title:
Text from PDF Page: 009
Type of Test Same category and up- loader, unweighted Same category, different uploader, unweighted Different category, weighted Figure 3: Seed set for prediction Figure 4: Recommended videos Seed Set 5 Sports videos from NBA uploader 5 Music videos from differ- ent uploaders 5 Music videos, highly rated; 2 Sports videos, low rated Result Set 5 Sports videos from basketball-related channels 5 videos from Music and Entertainment 4 Music videos, 1 Sports video Same category, negative weights 3 positive rated Music videos, 3 negative rated Music videos from uploader UMG 5 Music and Entertainment videos, not from uploader UMG Table 3: Results of Video Recommendation Using LSH closest neighbors in the LSH space include videos in the sports category, mostly being uploaded by NBA, but also other basketball-related channels. The thumbnails of the corresponding seed set videos can be viewed in 3 and our recommendations in 4, you can zoom in to be view the video titles and other details. Next, we relax the condition that the seed set contains videos by the same uploader. We start with 5 videos in the Music category but all different uploaders, and our 5 closest neighbors in LSH space are all in the Music or Entertainment category, with different uploaders. Further, we relax the condition that all ratings are the same, by starting with a seed set of 7 Music and Sports videos, with the Music videos having higher weights. Our 5 closest neighbors in LSH space contain 4 Music videos and 1 Sports video. Finally, we explore the case where some videos have negative ratings. We start with a seed set con- taining positively weighted Music videos of different uploaders, and negatively weighted Music videos by the Universal Music Group uploader. Our results contain videos in the Music and Enter- tainment category, none of which are from the Universal Music Group uploader. This uploader is one of the most highly present uploaders in the Music category, with over 900 videos, and hence not including any of their videos in our closest neighbors indicates that our approach is effective. In all our experiments, we obtain qualitative results that are in line with what we would expect from a YouTube video recommender system. Hence, we believe that our method shows great promise and solves our problem effectively. 6 Conclusion We explore the problem of YouTube video recommendations using graph information. After a thorough analysis of our dataset and its properties, we identify key structure in our problem. We 9PDF Image | Recommending Related YouTube Videos
PDF Search Title:
Recommending Related YouTube VideosOriginal File Name Searched:
cs224w-66-final.pdfDIY PDF Search: Google It | Yahoo | Bing
Cruise Ship Reviews | Luxury Resort | Jet | Yacht | and Travel Tech More Info
Cruising Review Topics and Articles More Info
Software based on Filemaker for the travel industry More Info
The Burgenstock Resort: Reviews on CruisingReview website... More Info
Resort Reviews: World Class resorts... More Info
The Riffelalp Resort: Reviews on CruisingReview website... More Info
| CONTACT TEL: 608-238-6001 Email: greg@cruisingreview.com | RSS | AMP |