Unbiased learning to rank (ULTR) studies the problem of mitigating various biases from implicit user feedback data such as clicks, and has been receiving considerable attention recently. A popular ULTR approach for real-world applications uses a two-tower architecture, where click modeling is factorized into a relevance tower with regular input features, and a bias tower with bias-relevant inputs such as the position of a document. A successful factorization will allow the relevance tower to be exempt from biases. In this work, we identify a critical issue that existing ULTR methods ignored - the bias tower can be confounded with the relevance tower via the underlying true relevance. In particular, the positions were determined by the logging policy, i.e., the previous production model, which would possess relevance information. We give both theoretical analysis and empirical results to show the negative effects on relevance tower due to such a correlation. We then propose three methods to mitigate the negative confounding effects by better disentangling relevance and bias. Empirical results on both controlled public datasets and a large-scale industry dataset show the effectiveness of the proposed approaches.
Understanding why a model makes certain predictions is crucial when adapting it for real world decision making. LIME is a popular model-agnostic feature attribution method for the tasks of classification and regression. However, the task of learning to rank in information retrieval is more complex in comparison with either classification or regression. In this work, we extend LIME to propose Rank-LIME, a model-agnostic, local, post-hoc linear feature attribution method for the task of learning to rank that generates explanations for ranked lists. We employ novel correlation-based perturbations, differentiable ranking loss functions and introduce new metrics to evaluate ranking based additive feature attribution models. We compare Rank-LIME with a variety of competing systems, with models trained on the MS MARCO datasets and observe that Rank-LIME outperforms existing explanation algorithms in terms of Model Fidelity and Explain-NDCG. With this we propose one of the first algorithms to generate additive feature attributions for explaining ranked lists.
Learning to rank is an effective recommendation approach since its introduction around 2010. Famous algorithms such as Bayesian Personalized Ranking and Collaborative Less is More Filtering have left deep impact in both academia and industry. However, most learning to rank approaches focus on improving technical accuracy metrics such as AUC, MRR and NDCG. Other evaluation metrics of recommender systems like fairness have been largely overlooked until in recent years. In this paper, we propose a new learning to rank algorithm named Pareto Pairwise Ranking. We are inspired by the idea of Bayesian Personalized Ranking and power law distribution. We show that our algorithm is competitive with other algorithms when evaluated on technical accuracy metrics. What is more important, in our experiment section we demonstrate that Pareto Pairwise Ranking is the most fair algorithm in comparison with 9 other contemporary algorithms.
Inspired by the impressive success of contrastive learning (CL), a variety of graph augmentation strategies have been employed to learn node representations in a self-supervised manner. Existing methods construct the contrastive samples by adding perturbations to the graph structure or node attributes. Although impressive results are achieved, it is rather blind to the wealth of prior information assumed: with the increase of the perturbation degree applied on the original graph, 1) the similarity between the original graph and the generated augmented graph gradually decreases; 2) the discrimination between all nodes within each augmented view gradually increases. In this paper, we argue that both such prior information can be incorporated (differently) into the contrastive learning paradigm following our general ranking framework. In particular, we first interpret CL as a special case of learning to rank (L2R), which inspires us to leverage the ranking order among positive augmented views. Meanwhile, we introduce a self-ranking paradigm to ensure that the discriminative information among different nodes can be maintained and also be less altered to the perturbations of different degrees. Experiment results on various benchmark datasets verify the effectiveness of our algorithm compared with the supervised and unsupervised models.
Many practical applications, such as recommender systems and learning to rank, involve solving multiple similar tasks. One example is learning of recommendation policies for users with similar movie preferences, where the users may still rank the individual movies slightly differently. Such tasks can be organized in a hierarchy, where similar tasks are related through a shared structure. In this work, we formulate this problem as a contextual off-policy optimization in a hierarchical graphical model from logged bandit feedback. To solve the problem, we propose a hierarchical off-policy optimization algorithm (HierOPO), which estimates the parameters of the hierarchical model and then acts pessimistically with respect to them. We instantiate HierOPO in linear Gaussian models, for which we also provide an efficient implementation and analysis. We prove per-task bounds on the suboptimality of the learned policies, which show a clear improvement over not using the hierarchical model. We also evaluate the policies empirically. Our theoretical and empirical results show a clear advantage of using the hierarchy over solving each task independently.
This paper introduces a new few-shot learning pipeline that casts relevance ranking for image retrieval as binary ranking relation classification. In comparison to image classification, ranking relation classification is sample efficient and domain agnostic. Besides, it provides a new perspective on few-shot learning and is complementary to state-of-the-art methods. The core component of our deep neural network is a simple MLP, which takes as input an image triplet encoded as the difference between two vector-Kronecker products, and outputs a binary relevance ranking order. The proposed RankMLP can be built on top of any state-of-the-art feature extractors, and our entire deep neural network is called the ranking deep neural network, or RankDNN. Meanwhile, RankDNN can be flexibly fused with other post-processing methods. During the meta test, RankDNN ranks support images according to their similarity with the query samples, and each query sample is assigned the class label of its nearest neighbor. Experiments demonstrate that RankDNN can effectively improve the performance of its baselines based on a variety of backbones and it outperforms previous state-of-the-art algorithms on multiple few-shot learning benchmarks, including miniImageNet, tieredImageNet, Caltech-UCSD Birds, and CIFAR-FS. Furthermore, experiments on the cross-domain challenge demonstrate the superior transferability of RankDNN.The code is available at: https://github.com/guoqianyu-alberta/RankDNN.
Persistent Memory (PMEM), also known as Non-Volatile Memory (NVM), can deliver higher density and lower cost per bit when compared with DRAM. Its main drawback is that it is typically slower than DRAM. On the other hand, DRAM has scalability problems due to its cost and energy consumption. Soon, PMEM will likely coexist with DRAM in computer systems but the biggest challenge is to know which data to allocate on each type of memory. This paper describes a methodology for identifying and characterizing application objects that have the most influence on the application's performance using Intel Optane DC Persistent Memory. In the first part of our work, we built a tool that automates the profiling and analysis of application objects. In the second part, we build a machine learning model to predict the most critical object within large-scale graph-based applications. Our results show that using isolated features does not bring the same benefit compared to using a carefully chosen set of features. By performing data placement using our predictive model, we can reduce the execution time degradation by 12\% (average) and 30\% (max) when compared to the baseline's approach based on LLC misses indicator.
As Learning-to-Rank (LTR) approaches primarily seek to improve ranking quality, their output scores are not scale-calibrated by design. This fundamentally limits LTR usage in score-sensitive applications. Though a simple multi-objective approach that combines a regression and a ranking objective can effectively learn scale-calibrated scores, we argue that the two objectives are not necessarily compatible, which makes the trade-off less ideal for either of them. In this paper, we propose a practical regression compatible ranking (RCR) approach that achieves a better trade-off, where the two ranking and regression components are proved to be mutually aligned. Although the same idea applies to ranking with both binary and graded relevance, we mainly focus on binary labels in this paper. We evaluate the proposed approach on several public LTR benchmarks and show that it consistently achieves either best or competitive result in terms of both regression and ranking metrics, and significantly improves the Pareto frontiers in the context of multi-objective optimization. Furthermore, we evaluated the proposed approach on YouTube Search and found that it not only improved the ranking quality of the production pCTR model, but also brought gains to the click prediction accuracy. The proposed approach has been successfully deployed in the YouTube production system.
In light of the smoothness property brought by skip connections in ResNet, this paper proposed the Skip Logit to introduce the skip connection mechanism that fits arbitrary DNN dimensions and embraces similar properties to ResNet. Meta Tanh Normalization (MTN) is designed to learn variance information and stabilize the training process. With these delicate designs, our Skip Meta Logit (SML) brought incremental boosts to the performance of extensive SOTA ctr prediction models on two real-world datasets. In the meantime, we prove that the optimization landscape of arbitrarily deep skip logit networks has no spurious local optima. Finally, SML can be easily added to building blocks and has delivered offline accuracy and online business metrics gains on app ads learning to rank systems at TikTok.
The page presentation biases in the information retrieval system, especially on the click behavior, is a well-known challenge that hinders improving ranking models' performance with implicit user feedback. Unbiased Learning to Rank~(ULTR) algorithms are then proposed to learn an unbiased ranking model with biased click data. However, most existing algorithms are specifically designed to mitigate position-related bias, e.g., trust bias, without considering biases induced by other features in search result page presentation(SERP), e.g. attractive bias induced by the multimedia. Unfortunately, those biases widely exist in industrial systems and may lead to an unsatisfactory search experience. Therefore, we introduce a new problem, i.e., whole-page Unbiased Learning to Rank(WP-ULTR), aiming to handle biases induced by whole-page SERP features simultaneously. It presents tremendous challenges: (1) a suitable user behavior model (user behavior hypothesis) can be hard to find; and (2) complex biases cannot be handled by existing algorithms. To address the above challenges, we propose a Bias Agnostic whole-page unbiased Learning to rank algorithm, named BAL, to automatically find the user behavior model with causal discovery and mitigate the biases induced by multiple SERP features with no specific design. Experimental results on a real-world dataset verify the effectiveness of the BAL.
A critical need for industrial recommender systems is the ability to evaluate recommendation policies offline, before deploying them to production. Unfortunately, widely used off-policy evaluation methods either make strong assumptions about how users behave that can lead to excessive bias, or they make fewer assumptions and suffer from large variance. We tackle this problem by developing a new estimator that mitigates the problems of the two most popular off-policy estimators for rankings, namely the position-based model and the item-position model. In particular, the new estimator, called INTERPOL, addresses the bias of a potentially misspecified position-based model, while providing an adaptable bias-variance trade-off compared to the item-position model. We provide theoretical arguments as well as empirical results that highlight the performance of our novel estimation approach.
Food recommendation system has proven as an effective technology to provide guidance on dietary choices, and this is especially important for patients suffering from chronic diseases. Unlike other multimedia recommendations, such as books and movies, food recommendation task is highly relied on the context at the moment, since users' food preference can be highly dynamic over time. For example, individuals tend to eat more calories earlier in the day and eat a little less at dinner. However, there are still limited research works trying to incorporate both current context and nutritional knowledge for food recommendation. Thus, a novel restaurant food recommendation system is proposed in this paper to recommend food dishes to users according to their special nutritional needs. Our proposed system utilises Optical Character Recognition (OCR) technology and a transformer-based deep learning model, Learning to Rank (LTR) model, to conduct food recommendation. Given a single RGB image of the menu, the system is then able to rank the food dishes in terms of the input search key (e.g., calorie, protein level). Due to the property of the transformer, our system can also rank unseen food dishes. Comprehensive experiments are conducted to validate our methods on a self-constructed menu dataset, known as MenuRank dataset. The promising results, with accuracy ranging from 77.2% to 99.5%, have demonstrated the great potential of LTR model in addressing food recommendation problems.
Centralized Training with Decentralized Execution (CTDE) has emerged as a widely adopted paradigm in multi-agent reinforcement learning, emphasizing the utilization of global information for learning an enhanced joint $Q$-function or centralized critic. In contrast, our investigation delves into harnessing global information to directly enhance individual $Q$-functions or individual actors. Notably, we discover that applying identical global information universally across all agents proves insufficient for optimal performance. Consequently, we advocate for the customization of global information tailored to each agent, creating agent-personalized global information to bolster overall performance. Furthermore, we introduce a novel paradigm named Personalized Training with Distilled Execution (PTDE), wherein agent-personalized global information is distilled into the agent's local information. This distilled information is then utilized during decentralized execution, resulting in minimal performance degradation. PTDE can be seamlessly integrated with state-of-the-art algorithms, leading to notable performance enhancements across diverse benchmarks, including the SMAC benchmark, Google Research Football (GRF) benchmark, and Learning to Rank (LTR) task.
Airbnb is a two-sided marketplace, bringing together hosts who own listings for rent, with prospective guests from around the globe. Applying neural network-based learning to rank techniques has led to significant improvements in matching guests with hosts. These improvements in ranking were driven by a core strategy: order the listings by their estimated booking probabilities, then iterate on techniques to make these booking probability estimates more and more accurate. Embedded implicitly in this strategy was an assumption that the booking probability of a listing could be determined independently of other listings in search results. In this paper we discuss how this assumption, pervasive throughout the commonly-used learning to rank frameworks, is false. We provide a theoretical foundation correcting this assumption, followed by efficient neural network architectures based on the theory. Explicitly accounting for possible similarities between listings, and reducing them to diversify the search results generated strong positive impact. We discuss these metric wins as part of the online A/B tests of the theory. Our method provides a practical way to diversify search results for large-scale production ranking systems.
There is extensive interest in metric learning methods for image retrieval. Many metric learning loss functions focus on learning a correct ranking of training samples, but strongly overfit semantically inconsistent labels and require a large amount of data. To address these shortcomings, we propose a new metric learning method, called contextual loss, which optimizes contextual similarity in addition to cosine similarity. Our contextual loss implicitly enforces semantic consistency among neighbors while converging to the correct ranking. We empirically show that the proposed loss is more robust to label noise, and is less prone to overfitting even when a large portion of train data is withheld. Extensive experiments demonstrate that our method achieves a new state-of-the-art across four image retrieval benchmarks and multiple different evaluation settings. Code is available at: https://github.com/Chris210634/metric-learning-using-contextual-similarity
In learning-to-rank problems, a privileged feature is one that is available during model training, but not available at test time. Such features naturally arise in merchandised recommendation systems; for instance, "user clicked this item" as a feature is predictive of "user purchased this item" in the offline data, but is clearly not available during online serving. Another source of privileged features is those that are too expensive to compute online but feasible to be added offline. Privileged features distillation (PFD) refers to a natural idea: train a "teacher" model using all features (including privileged ones) and then use it to train a "student" model that does not use the privileged features. In this paper, we first study PFD empirically on three public ranking datasets and an industrial-scale ranking problem derived from Amazon's logs. We show that PFD outperforms several baselines (no-distillation, pretraining-finetuning, self-distillation, and generalized distillation) on all these datasets. Next, we analyze why and when PFD performs well via both empirical ablation studies and theoretical analysis for linear models. Both investigations uncover an interesting non-monotone behavior: as the predictive power of a privileged feature increases, the performance of the resulting student model initially increases but then decreases. We show the reason for the later decreasing performance is that a very predictive privileged teacher produces predictions with high variance, which lead to high variance student estimates and inferior testing performance.
In this paper, we present a novel perspective towards IR evaluation by proposing a new family of evaluation metrics where the existing popular metrics (e.g., nDCG, MAP) are customized by introducing a query-specific lower-bound (LB) normalization term. While original nDCG, MAP etc. metrics are normalized in terms of their upper bounds based on an ideal ranked list, a corresponding LB normalization for them has not yet been studied. Specifically, we introduce two different variants of the proposed LB normalization, where the lower bound is estimated from a randomized ranking of the corresponding documents present in the evaluation set. We next conducted two case-studies by instantiating the new framework for two popular IR evaluation metric (with two variants, e.g., DCG_UL_V1,2 and MSP_UL_V1,2 ) and then comparing against the traditional metric without the proposed LB normalization. Experiments on two different data-sets with eight Learning-to-Rank (LETOR) methods demonstrate the following properties of the new LB normalized metric: 1) Statistically significant differences (between two methods) in terms of original metric no longer remain statistically significant in terms of Upper Lower (UL) Bound normalized version and vice-versa, especially for uninformative query-sets. 2) When compared against the original metric, our proposed UL normalized metrics demonstrate higher Discriminatory Power and better Consistency across different data-sets. These findings suggest that the IR community should consider UL normalization seriously when computing nDCG and MAP and more in-depth study of UL normalization for general IR evaluation is warranted.
Active Learning (AL) is a well-known standard method for efficiently obtaining annotated data by first labeling the samples that contain the most information based on a query strategy. In the past, a large variety of such query strategies has been proposed, with each generation of new strategies increasing the runtime and adding more complexity. However, to the best of our our knowledge, none of these strategies excels consistently over a large number of datasets from different application domains. Basically, most of the the existing AL strategies are a combination of the two simple heuristics informativeness and representativeness, and the big differences lie in the combination of the often conflicting heuristics. Within this paper, we propose ImitAL, a domain-independent novel query strategy, which encodes AL as a learning-to-rank problem and learns an optimal combination between both heuristics. We train ImitAL on large-scale simulated AL runs on purely synthetic datasets. To show that ImitAL was successfully trained, we perform an extensive evaluation comparing our strategy on 13 different datasets, from a wide range of domains, with 7 other query strategies.
A common way to avoid overfitting in supervised learning is early stopping, where a held-out set is used for iterative evaluation during training to find a sweet spot in the number of training steps that gives maximum generalization. However, such a method requires a disjoint validation set, thus part of the labeled data from the training set is usually left out for this purpose, which is not ideal when training data is scarce. Furthermore, when the training labels are noisy, the performance of the model over a validation set may not be an accurate proxy for generalization. In this paper, we propose a method to spot an early stopping point in the training iterations without the need for a validation set. We first show that in the overparameterized regime the randomly initialized weights of a linear model converge to the same direction during training. Using this result, we propose to train two parallel instances of a linear model, initialized with different random seeds, and use their intersection as a signal to detect overfitting. In order to detect intersection, we use the cosine distance between the weights of the parallel models during training iterations. Noticing that the final layer of a NN is a linear map of pre-last layer activations to output logits, we build on our criterion for linear models and propose an extension to multi-layer networks, using the new notion of counterfactual weights. We conduct experiments on two areas that early stopping has noticeable impact on preventing overfitting of a NN: (i) learning from noisy labels; and (ii) learning to rank in IR. Our experiments on four widely used datasets confirm the effectiveness of our method for generalization. For a wide range of learning rates, our method, called Cosine-Distance Criterion (CDC), leads to better generalization on average than all the methods that we compare against in almost all of the tested cases.
Ranking lies at the core of many Information Retrieval (IR) tasks. While existing research on Learning to Rank (LTR) using Deep Neural Network (DNN) has achieved great success, it is somewhat limited because of its dependence on fine-grained labels. In practice, fine-grained labels are often expensive to acquire, i.e. explicit relevance judgements, or suffer from biases, i.e. click logs. Compared to fine-grained labels, coarse-grained labels are easier and cheaper to collect. Some recent works propose utilizing only coarse-grained labels for LTR tasks. A most representative line of work introduces Reinforcement Learning (RL) algorithms. RL can help train the LTR model with little reliance on fine-grained labels compared to Supervised Learning. To study the effectiveness of the RL-based LTR algorithm on coarse-grained labels, in this paper, we implement four different RL paradigms and conduct extensive experiments on two well-established LTR datasets. The results on simulated coarse-grained labeled dataset show that while using coarse-grained labels to train an RL model for LTR tasks still can not outperform traditional approaches using fine-grained labels, it still achieve somewhat promising results and is potentially helpful for future research in LTR. Our code implementations will be released after this work is accepted.