Learning sophisticated feature interactions behind user behaviors is critical in maximizing CTR for recommender systems. Despite great progress, existing methods seem to have a strong bias towards low- or high-order interactions, or require expertise feature engineering. In this paper, we show that it is possible to derive an end-to-end learning model that emphasizes both low- and high-order feature interactions. The proposed model, DeepFM, combines the power of factorization machines for recommendation and deep learning for feature learning in a new neural network architecture. Compared to the latest Wide & Deep model from Google, DeepFM has a shared input to its "wide" and "deep" parts, with no need of feature engineering besides raw features. Comprehensive experiments are conducted to demonstrate the effectiveness and efficiency of DeepFM over the existing models for CTR prediction, on both benchmark data and commercial data.

/pdf/deepfm-a-factorization-machine-based-neural-network-for-ctr-xde2g2fq1v.pdf

DeepFM: a factorization-machine based neural network for CTR prediction

Learning algorithms related to artificial neural networks and in particular for Deep Learning may seem to involve many bells and whistles, called hyper-parameters. This chapter is meant as a practical guide with recommendations for some of the most commonly used hyper-parameters, in particular in the context of learning algorithms based on back-propagated gradient and gradient-based optimization. It also discusses how to deal with the fact that more interesting results can be obtained when allowing one to adjust many hyper-parameters. Overall, it describes elements of the practice used to successfully and efficiently train and debug large-scale and often deep multi-layer neural networks. It closes with open questions about the training difficulties observed with deeper architectures.

Practical recommendations for gradient-based training of deep architectures

Thompson sampling is one of oldest heuristic to address the exploration / exploitation trade-off, but it is surprisingly unpopular in the literature. We present here some empirical results using Thompson sampling on simulated and real data, and show that it is highly competitive. And since this heuristic is very easy to implement, we argue that it should be part of the standard baselines to compare against.

/pdf/an-empirical-evaluation-of-thompson-sampling-cmerwqpc37.pdf

An Empirical Evaluation of Thompson Sampling

Learning algorithms related to artificial neural networks and in particular for Deep Learning may seem to involve many bells and whistles, called hyper-parameters. This chapter is meant as a practical guide with recommendations for some of the most commonly used hyperparameters, in particular in the context of learning algorithms based on back-propagated gradient and gradient-based optimization. It also discusses how to deal with the fact that more interesting results can be obtained when allowing one to adjust many hyper-parameters. Overall, it describes elements of the practice used to successfully and efficiently train and debug large-scale and often deep multi-layer neural networks. It closes with open questions about the training difficulties observed with deeper architectures.

The twenty last years have been marked by an increase in available data and computing power In parallel to this trend, the focus of neural network research and the practice of training neural networks has undergone a number of important changes, for example, use of deep learning machines The second edition of the book augments the first edition with more tricks, which have resulted from 14 years of theory and experimentation by some of the world's most prominent neural network researchers These tricks can make a substantial difference (in terms of speed, ease of implementation, and accuracy) when it comes to putting algorithms to work on real problems

Neural Networks: Tricks of the Trade

We describe a new Bayesian click-through rate (CTR) prediction algorithm used for Sponsored Search in Microsoft's Bing search engine. The algorithm is based on a probit regression model that maps discrete or real-valued input features to probabilities. It maintains Gaussian beliefs over weights of the model and performs Gaussian online updates derived from approximate message passing. Scalability of the algorithm is ensured through a principled weight pruning procedure and an approximate parallel implementation. We discuss the challenges arising from evaluating and tuning the predictor as part of the complex system of sponsored search where the predictions made by the algorithm decide about future training sample composition. Finally, we show experimental results from the production system and compare to a calibrated Naive Bayes algorithm.

https://www.microsoft.com/en-us/research/wp-content/uploads/2010/06/AdPredictor-ICML-2010-final.pdf

Web-Scale Bayesian Click-Through rate Prediction for Sponsored Search Advertising in Microsoft's Bing Search Engine

Event prediction in dynamic environments is described. In an embodiment a prediction engine may use the learnt information to predict events in order to control a system such as for internet advertising, email filtering, fraud detection or other applications. In an example one or more variables exists for pre-specified features describing or associated with events and each variable is considered to have an associated weight and time stamp. For example, belief about each weight is represented using a probability distribution and a dynamics process is used to modify the probability distribution in a manner dependent on the time stamp for that weight. For example, the uncertainty about the associated variable's influence on prediction of future events is increased. Examples of different schedules for applying the dynamics process are given.

Event prediction in dynamic environments

Methods, systems, and media are provided for a dynamic batch strategy utilized in parallelization of online learning algorithms. The dynamic batch strategy provides a merge function on the basis of a threshold level difference between the original model state and an updated model state, rather than according to a constant or pre-determined batch size. The merging includes reading a batch of incoming streaming data, retrieving any missing model beliefs from partner processors, and training on the batch of incoming streaming data. The steps of reading, retrieving, and training are repeated until the measured difference in states exceeds a set threshold level. The measured differences which exceed the threshold level are merged for each of the plurality of processors according to attributes. The merged differences which exceed the threshold level are combined with the original partial model states to obtain an updated global model state.

Parallelization of online learning algorithms

We present an efficient Bayesian online learning algorithm for clustering vectors of binary values based on a well known model, the mixture of Bernoulli profiles. The model includes conjugate Beta priors over the success probabilities and maintains discrete probability distributions for cluster assignments. Clustering is then formulated as inference in a factor graph which is solved efficiently using online approximate message passing. The resulting algorithm has three key features: a) it requires only a single pass across the data and can hence be used on data streams, b) it maintains the uncertainty of parameters and cluster assignments, and c) it implements an automatic step size adaptation based on the current model uncertainty. The model is tested on an artificially generated toy dataset and applied to a large scale real-world data set from online advertising, the data being online ads characterized by the set of keywords to which they have been subscribed. The proposed approach scales well for large datasets, and compares favorably to other clustering algorithms on the ads dataset. As a concrete application to online advertising we show how the learnt model can be used to recommend new keywords for given ads.

/pdf/scalable-clustering-and-keyword-suggestion-for-online-353j8q00jc.pdf

Scalable clustering and keyword suggestion for online advertisements

Methods, computer systems, and computer-storage media are provided for ranking ads. A reserve price is included in a calculation of a score to rank one or more advertisements for display. The calculation may further rely on a bid submitted by an advertiser for an advertisement, a click probability associated with the advertisement, a relevance of the advertisement to a search query and/or user, and the like. Once the reserve price is used to establish a score for one or more advertisements, a price is calculated for each of the one or more advertisements. The one or more advertisements may be displayed in an order indicated by the scores of each of the advertisements.

Thomas Borchert

Papers

Web-Scale Bayesian Click-Through rate Prediction for Sponsored Search Advertising in Microsoft's Bing Search Engine

Event prediction in dynamic environments

Parallelization of online learning algorithms

Scalable clustering and keyword suggestion for online advertisements

Utilizing a reserve price for ranking