A Space Alignment Method for Cold-Start TV Show Recommendations
Authors: Shiyu Chang, Jiayu Zhou, Pirooz Chubak, Junling Hu, Thomas Huang
IJCAI 2015 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We present experimental results on a real TV watch behavior data set to demonstrate the significant performance improvement over other state-of-the-art algorithms. |
| Researcher Affiliation | Collaboration | Beckman Institute, University of Illinois at Urbana-Champaign, IL 61801. Samsung Research America, San Jose, CA 95134. |
| Pseudocode | No | The paper contains mathematical formulations and descriptions of the methods but does not include any clearly labeled 'Pseudocode' or 'Algorithm' blocks. |
| Open Source Code | No | A detailed tutorial on item-item similarity computing within the Hadoop framework can be found in [Apache, ]. Once the similarity matrix S is obtained, training the offline model only involves top SVD, which can be readily solved by the Apache Mahout framework [Apache, ]. |
| Open Datasets | No | The data we report upon in this paper comes from the server logs at Samsung where the data is anonymized and encrypted for privacy reasons. |
| Dataset Splits | Yes | To evaluate the quality of the recommender systems, we split these 15 days of data into eight folds, each containing eight consecutive days of data. Then we use the first seven days as the training set and the last day as a cold-start testing set." and "The parameters of all competing methods are determined by a 10% validation set separated from the training data. |
| Hardware Specification | No | The paper mentions using a 'Map Reduce framework' and 'Hadoop' but does not provide specific details on the hardware (e.g., GPU/CPU models, memory) used for the experiments. |
| Software Dependencies | No | A detailed tutorial on item-item similarity computing within the Hadoop framework can be found in [Apache, ]. Once the similarity matrix S is obtained, training the offline model only involves top SVD, which can be readily solved by the Apache Mahout framework [Apache, ]. |
| Experiment Setup | Yes | We compute the neighborhood coverage on the testing day and vary the rank from five to 50 for the purposed method." and "Throughout this paper, we set the rank of M in the range of 40-50." and "We vary N from 5 to 100 and present the performance in terms of top-k MAP and MAR for different k values." and "We use a neighborhood size N = 40 through all our experiments." Also "We present each TV show to a 200-dimensional vector by using its title, description and genre (after tokenizing and stemming). In addition, we concatenate it with channel, program starting time and ending time as three additional features. The final dimensionality of each TV program in the content space is 203. We further normalize the content vector to make the the ℓ2-norm to one. |