Embedded Files
Novelties:
Novelties:
- Optimal Pipeline
We developed an optimal pipeline combining the strengths of existing techniques
2. Social Relationship Descriptor
2. Social Relationship Descriptor
Since we did not have user study-based data for team member relationship estimation, we came up with our own intuition. We have assumed the social relationship descriptor as the Intersection-over-Union(IOU) of the items watched by all the team members. Our intuition is that there must be some social relationship between the team members because they preferred clicking/opening/watching the common items. They may or may not like the items after watching. However, the very fact that they chose to give it a try forms a direct connection between their preferences.
3. Sparseness Descriptor Experiments
3. Sparseness Descriptor Experiments
We have conducted 4 new experiments as sparseness controller:
Matrix Factorization + Learning Rate Scheduler using Exponential Decay
Matrix Factorization + Learning Rate Scheduler using Exponential Decay with step size
Baseline-included Matrix Factorization + Learning Rate Scheduler using Exponential Decay
Baseline-included Matrix Factorization + Learning Rate Scheduler using Exponential Decay with step size
Wait! How did we achieve this?
Wait! How did we achieve this?
We had a lot of Challenges too!
Page updated
Report abuse