Abstract: Machine Learning based ranking models are ubiquitous in powering recommendation engines at internet companies. These models typically use a combination of real-valued numerical and categorical features to generate predictions. Feature selection may be a widely encountered problem in this setting, that entails picking the optimal set of features as inputs to these models from a large pool of candidate real-valued and categorical features. A novel feature selection algorithm for categorical features building on stochastic neural networks is provided. It is shown empirically through results, the superiority of this algorithm over existing approaches. Study and proposal of best practices are also provided to practitioners to extract maximum value out of the new feature selection approach.

Abstract: Methods and apparatus related to determining a transition probability related to transition from one or more past activity indications to one or more subsequent activity indications. Some implementations of the specification are directed to methods and apparatus related to identifying a set of one or more activity indications of a user, identifying at least first and second transition probabilities from the set to a subsequent set of one or more activity indications, and determining a user transition probability from the set to the subsequent set based on the first and second transition probabilities.