I would add Multi objective optimization as another option to look at while training ML systems. In search ranking, it helps to learn balance between relevance, engagement, business metrics etc.
Multi task DL models has its own set of benefits, specially when data is sparse for one set of objective.
A nice paper on multi objective optimization for search ranking from Amazon
I would add Multi objective optimization as another option to look at while training ML systems. In search ranking, it helps to learn balance between relevance, engagement, business metrics etc.
Multi task DL models has its own set of benefits, specially when data is sparse for one set of objective.
A nice paper on multi objective optimization for search ranking from Amazon
https://assets.amazon.science/4d/9c/69cbef8346408349385c780cac48/scipub-1195.pdf