GUITAR: Gradient Pruning toward Fast Neural Ranking

With the continuous popularity of deep learning and representation learning, fast vector search becomes a vital task in various ranking/retrieval based applications, say recommendation, ads ranking and question answering. Neural network based ranking is widely adopted due to its powerful capacity in modeling complex relationships, such as between users and items, questions and answers. However, it is usually exploited in offline or re-ranking manners for it is time-consuming in computations. Online neural network ranking--so called fast neural ranking--is considered challenging because neural network measures are usually non-convex and asymmetric. Traditional Approximate Nearest Neighbor (ANN) search which usually focuses on metric ranking measures, is not applicable to these advanced measures. In this paper, we introduce a novel graph searching framework to accelerate the searching in the fast neural ranking problem. The proposed graph searching algorithm is bi-level: we first construct a probable candidate set; then we only evaluate the neural network measure over the probable candidate set instead of evaluating the neural network over all neighbors. Specifically, we propose a gradient-based algorithm that approximates the rank of the neural network matching score to construct the probable candidate set; and we present an angle-based heuristic procedure to adaptively identify the proper size of the probable candidate set. Empirical results on public data confirm the effectiveness of our proposed algorithms.

Pb-Hash: Partitioned b-bit Hashing

Many hashing algorithms including minwise hashing (MinHash), one permutation hashing (OPH), and consistent weighted sampling (CWS) generate integers of $B$ bits. With $k$ hashes for each data vector, the storage would be $B\times k$ bits; and when used for large-scale learning, the model size would be $2^B\times k$, which can be expensive. A standard strategy is to use only the lowest $b$ bits out of the $B$ bits and somewhat increase $k$, the number of hashes. In this study, we propose to re-use the hashes by partitioning the $B$ bits into $m$ chunks, e.g., $b\times m =B$. Correspondingly, the model size becomes $m\times 2^b \times k$, which can be substantially smaller than the original $2^B\times k$. Our theoretical analysis reveals that by partitioning the hash values into $m$ chunks, the accuracy would drop. In other words, using $m$ chunks of $B/m$ bits would not be as accurate as directly using $B$ bits. This is due to the correlation from re-using the same hash. On the other hand, our analysis also shows that the accuracy would not drop much for (e.g.,) $m=2\sim 4$. In some regions, Pb-Hash still works well even for $m$ much larger than 4. We expect Pb-Hash would be a good addition to the family of hashing methods/applications and benefit industrial practitioners. We verify the effectiveness of Pb-Hash in machine learning tasks, for linear SVM models as well as deep learning models. Since the hashed data are essentially categorical (ID) features, we follow the standard practice of using embedding tables for each hash. With Pb-Hash, we need to design an effective strategy to combine $m$ embeddings. Our study provides an empirical evaluation on four pooling schemes: concatenation, max pooling, mean pooling, and product pooling. There is no definite answer which pooling would be always better and we leave that for future study.

Blockwise Feature Interaction in Recommendation Systems

Feature interactions can play a crucial role in recommendation systems as they capture complex relationships between user preferences and item characteristics. Existing methods such as Deep & Cross Network (DCNv2) may suffer from high computational requirements due to their cross-layer operations. In this paper, we propose a novel approach called blockwise feature interaction (BFI) to help alleviate this issue. By partitioning the feature interaction process into smaller blocks, we can significantly reduce both the memory footprint and the computational burden. Four variants (denoted by P, Q, T, S, respectively) of BFI have been developed and empirically compared. Our experimental results demonstrate that the proposed algorithms achieves close accuracy compared to the standard DCNv2, while greatly reducing the computational overhead and the number of parameters. This paper contributes to the development of efficient recommendation systems by providing a practical solution for improving feature interaction efficiency.

Practice with Graph-based ANN Algorithms on Sparse Data: Chi-square Two-tower model, HNSW, Sign Cauchy Projections

Sparse data are common. The traditional ``handcrafted'' features are often sparse. Embedding vectors from trained models can also be very sparse, for example, embeddings trained via the ``ReLu'' activation function. In this paper, we report our exploration of efficient search in sparse data with graph-based ANN algorithms (e.g., HNSW, or SONG which is the GPU version of HNSW), which are popular in industrial practice, e.g., search and ads (advertising). We experiment with the proprietary ads targeting application, as well as benchmark public datasets. For ads targeting, we train embeddings with the standard ``cosine two-tower'' model and we also develop the ``chi-square two-tower'' model. Both models produce (highly) sparse embeddings when they are integrated with the ``ReLu'' activation function. In EBR (embedding-based retrieval) applications, after we the embeddings are trained, the next crucial task is the approximate near neighbor (ANN) search for serving. While there are many ANN algorithms we can choose from, in this study, we focus on the graph-based ANN algorithm (e.g., HNSW-type). Sparse embeddings should help improve the efficiency of EBR. One benefit is the reduced memory cost for the embeddings. The other obvious benefit is the reduced computational time for evaluating similarities, because, for graph-based ANN algorithms such as HNSW, computing similarities is often the dominating cost. In addition to the effort on leveraging data sparsity for storage and computation, we also integrate ``sign cauchy random projections'' (SignCRP) to hash vectors to bits, to further reduce the memory cost and speed up the ANN search. In NIPS'13, SignCRP was proposed to hash the chi-square similarity, which is a well-adopted nonlinear kernel in NLP and computer vision. Therefore, the chi-square two-tower model, SignCRP, and HNSW are now tightly integrated.

Building K-Anonymous User Cohorts with\\ Consecutive Consistent Weighted Sampling (CCWS)

To retrieve personalized campaigns and creatives while protecting user privacy, digital advertising is shifting from member-based identity to cohort-based identity. Under such identity regime, an accurate and efficient cohort building algorithm is desired to group users with similar characteristics. In this paper, we propose a scalable $K$-anonymous cohort building algorithm called {\em consecutive consistent weighted sampling} (CCWS). The proposed method combines the spirit of the ($p$-powered) consistent weighted sampling and hierarchical clustering, so that the $K$-anonymity is ensured by enforcing a lower bound on the size of cohorts. Evaluations on a LinkedIn dataset consisting of $>70$M users and ads campaigns demonstrate that CCWS achieves substantial improvements over several hashing-based methods including sign random projections (SignRP), minwise hashing (MinHash), as well as the vanilla CWS.

Constrained Approximate Similarity Search on Proximity Graph

Search engines and recommendation systems are built to efficiently display relevant information from those massive amounts of candidates. Typically a three-stage mechanism is employed in those systems: (i) a small collection of items are first retrieved by (e.g.,) approximate near neighbor search algorithms; (ii) then a collection of constraints are applied on the retrieved items; (iii) a fine-grained ranking neural network is employed to determine the final recommendation. We observe a major defect of the original three-stage pipeline: Although we only target to retrieve $k$ vectors in the final recommendation, we have to preset a sufficiently large $s$ ($s > k$) for each query, and ``hope'' the number of survived vectors after the filtering is not smaller than $k$. That is, at least $k$ vectors in the $s$ similar candidates satisfy the query constraints. In this paper, we investigate this constrained similarity search problem and attempt to merge the similarity search stage and the filtering stage into one single search operation. We introduce AIRSHIP, a system that integrates a user-defined function filtering into the similarity search framework. The proposed system does not need to build extra indices nor require prior knowledge of the query constraints. We propose three optimization strategies: (1) starting point selection, (2) multi-direction search, and (3) biased priority queue selection. Experimental evaluations on both synthetic and real data confirm the effectiveness of the proposed AIRSHIP algorithm. We focus on constrained graph-based approximate near neighbor (ANN) search in this study, in part because graph-based ANN is known to achieve excellent performance. We believe it is also possible to develop constrained hashing-based ANN or constrained quantization-based ANN.

Asymmetric Hashing for Fast Ranking via Neural Network Measures

Fast item ranking is an important task in recommender systems. In previous works, graph-based Approximate Nearest Neighbor (ANN) approaches have demonstrated good performance on item ranking tasks with generic searching/matching measures (including complex measures such as neural network measures). However, since these ANN approaches must go through the neural measures several times during ranking, the computation is not practical if the neural measure is a large network. On the other hand, fast item ranking using existing hashing-based approaches, such as Locality Sensitive Hashing (LSH), only works with a limited set of measures. Previous learning-to-hash approaches are also not suitable to solve the fast item ranking problem since they can take a significant amount of time and computation to train the hash functions. Hashing approaches, however, are attractive because they provide a principle and efficient way to retrieve candidate items. In this paper, we propose a simple and effective learning-to-hash approach for the fast item ranking problem that can be used for any type of measure, including neural network measures. Specifically, we solve this problem with an asymmetric hashing framework based on discrete inner product fitting. We learn a pair of related hash functions that map heterogeneous objects (e.g., users and items) into a common discrete space where the inner product of their binary codes reveals their true similarity defined via the original searching measure. The fast ranking problem is reduced to an ANN search via this asymmetric hashing scheme. Then, we propose a sampling strategy to efficiently select relevant and contrastive samples to train the hashing model. We empirically validate the proposed method against the existing state-of-the-art fast item ranking methods in several combinations of non-linear searching functions and prominent datasets.

Package for Fast ABC-Boost

This report presents the open-source package which implements the series of our boosting works in the past years. In particular, the package includes mainly three lines of techniques, among which the following two are already the standard implementations in popular boosted tree platforms: (i) The histogram-based (feature-binning) approach makes the tree implementation convenient and efficient. In Li et al (2007), a simple fixed-length adaptive binning algorithm was developed. In this report, we demonstrate that such a simple algorithm is still surprisingly effective compared to more sophisticated variants in popular tree platforms. (ii) The explicit gain formula in Li (20010) for tree splitting based on second-order derivatives of the loss function typically improves, often considerably, over the first-order methods. Although the gain formula in Li (2010) was derived for logistic regression loss, it is a generic formula for loss functions with second-derivatives. For example, the open-source package also includes $L_p$ regression for $p\geq 1$. The main contribution of this package is the ABC-Boost (adaptive base class boosting) for multi-class classification. The initial work in Li (2008) derived a new set of derivatives of the classical multi-class logistic regression by specifying a "base class". The accuracy can be substantially improved if the base class is chosen properly. The major technical challenge is to design a search strategy to select the base class. The prior published works implemented an exhaustive search procedure to find the base class which is computationally too expensive. Recently, a new report (Li and Zhao, 20022) presents a unified framework of "Fast ABC-Boost" which allows users to efficiently choose the proper search space for the base class. The package provides interfaces for linux, windows, mac, matlab, R, python.

pGMM Kernel Regression and Comparisons with Boosted Trees

In this work, we demonstrate the advantage of the pGMM (``powered generalized min-max'') kernel in the context of (ridge) regression. In recent prior studies, the pGMM kernel has been extensively evaluated for classification tasks, for logistic regression, support vector machines, as well as deep neural networks. In this paper, we provide an experimental study on ridge regression, to compare the pGMM kernel regression with the ordinary ridge linear regression as well as the RBF kernel ridge regression. Perhaps surprisingly, even without a tuning parameter (i.e., $p=1$ for the power parameter of the pGMM kernel), the pGMM kernel already performs well. Furthermore, by tuning the parameter $p$, this (deceptively simple) pGMM kernel even performs quite comparably to boosted trees. Boosting and boosted trees are very popular in machine learning practice. For regression tasks, typically, practitioners use $L_2$ boost, i.e., for minimizing the $L_2$ loss. Sometimes for the purpose of robustness, the $L_1$ boost might be a choice. In this study, we implement $L_p$ boost for $p\geq 1$ and include it in the package of ``Fast ABC-Boost''. Perhaps also surprisingly, the best performance (in terms of $L_2$ regression loss) is often attained at $p>2$, in some cases at $p\gg 2$. This phenomenon has already been demonstrated by Li et al (UAI 2010) in the context of k-nearest neighbor classification using $L_p$ distances. In summary, the implementation of $L_p$ boost provides practitioners the additional flexibility of tuning boosting algorithms for potentially achieving better accuracy in regression applications.