While significant progress has been made in specifying neural networks capable of representing uncertainty, deep networks still often suffer from overconfidence and misaligned predictive distributions. Existing approaches for measuring this misalignment are primarily developed under the framework of calibration, with common metrics such as Expected Calibration Error (ECE). However, calibration can only provide a strictly marginal assessment of probabilistic alignment. Consequently, calibration metrics such as ECE are distribution-wise measures and cannot diagnose the point-wise reliability of individual inputs, which is important for real-world decision-making. We propose a stronger condition, which we term conditional congruence, for assessing probabilistic fit. We also introduce a metric, Conditional Congruence Error (CCE), that uses conditional kernel mean embeddings to estimate the distance, at any point, between the learned predictive distribution and the empirical, conditional distribution in a dataset. We perform several high dimensional regression tasks and show that CCE exhibits four critical properties: correctness, monotonicity, reliability, and robustness.

Neural networks capable of accurate, input-conditional uncertainty representation are essential for real-world AI systems. Deep ensembles of Gaussian networks have proven highly effective for continuous regression due to their ability to flexibly represent aleatoric uncertainty via unrestricted heteroscedastic variance, which in turn enables accurate epistemic uncertainty estimation. However, no analogous approach exists for $\textit{count}$ regression, despite many important applications. To address this gap, we propose the Deep Double Poisson Network (DDPN), a novel neural discrete count regression model that outputs the parameters of the Double Poisson distribution, enabling arbitrarily high or low predictive aleatoric uncertainty for count data and improving epistemic uncertainty estimation when ensembled. We formalize and prove that DDPN exhibits robust regression properties similar to heteroscedastic Gaussian models via learnable loss attenuation, and introduce a simple loss modification to control this behavior. Experiments on diverse datasets demonstrate that DDPN outperforms current baselines in accuracy, calibration, and out-of-distribution detection, establishing a new state-of-the-art in deep count regression.

In practice, it is essential to compare and rank candidate policies offline before real-world deployment for safety and reliability. Prior work seeks to solve this offline policy ranking (OPR) problem through value-based methods, such as Off-policy evaluation (OPE). However, they fail to analyze special case performance (e.g., worst or best cases), due to the lack of holistic characterization of policies’ performance. It is even more difficult to estimate precise policy values when the reward is not fully accessible under sparse settings. In this paper, we present Probabilistic Offline Policy Ranking (POPR), a framework to address OPR problems by leveraging expert data to characterize the probability of a candidate policy behaving like experts, and approximating its entire performance posterior distribution to help with ranking. POPR does not rely on value estimation, and the derived performance posterior can be used to distinguish candidates in worst-, best-, and average-cases. To estimate the posterior, we propose POPR-EABC, an Energy-based Approximate Bayesian Computation (ABC) method conducting likelihood-free inference. POPR-EABC reduces the heuristic nature of ABC by a smooth energy function, and improves the sampling efficiency by a pseudo-likelihood. We empirically demonstrate that POPR-EABC is adequate for evaluating policies in both discrete and continuous action spaces across various experiment environments, and facilitates probabilistic comparisons of candidate policies before deployment.

Product assortment selection is a critical challenge facing physical retailers. Effectively aligning inventory with the preferences of shoppers can increase sales and decrease out-of-stocks. However, in real-world settings the problem is challenging due to the combinatorial explosion of product assortment possibilities. Consumer preferences are typically heterogeneous across space and time, making inventory-preference alignment challenging. Additionally, existing strategies rely on syndicated data, which tends to be aggregated, low resolution, and suffer from high latency. To solve these challenges, we introduce a real-time recommendation system, which we call EdgeRec3D. Our system utilizes recent advances in 3D computer vision for perception and automatic, fine grained sales estimation. These perceptual components run on the edge of the network and facilitate real-time reward signals. Additionally, we develop a Bayesian payoff model to account for noisy estimates from 3D LIDAR data. We rely on spatial clustering to allow the system to adapt to heterogeneous consumer preferences, and a graph-based candidate generation algorithm to address the combinatorial search problem. We test our system in real-world stores across two, 6-8 week A/B tests with beverage products and demonstrate a 35% and 27% increase in sales respectively. Finally, we monitor the deployed system for a period of 28 weeks with an observational study and show a 9.4% increase in sales.

3D scene understanding is an important problem that has experienced great progress in recent years, in large part due to the development of state-of-the-art methods for 3D object detection. However, the performance of 3D object detectors can suffer in scenarios where extreme occlusion of objects is present, or the number of object classes is large. In this paper, we study the problem of inferring 3D counts from densely packed scenes with heterogeneous objects. This problem has applications to important tasks such as inventory management or automatic crop yield estimation. We propose a novel regression-based method, CountNet3D, that uses mature 2D object detectors for finegrained classification and localization, and a PointNet backbone for geometric embedding. The network processes fused data from images and point clouds for end-to-end learning of counts. We perform experiments on a novel synthetic dataset for inventory management in retail, which we construct and make publicly available to the community. Our results show that regression-based 3D counting methods systematically outperform detection-based methods, and reveal that directly learning from raw point clouds greatly assists count estimation under extreme occlusion. Finally, we study the effectiveness of CountNet3D on a large dataset of real-world scenes where extreme occlusion is present and achieve an error rate of 11.01% .

Product allocation in retail is the process of placing products throughout a store to connect consumers with relevant products. Discovering a good allocation strategy is challenging due to the scarcity of data and the high cost of experimentation in the physical world. Some work explores Reinforcement learning (RL) as a solution, but these approaches are often limited because of the sim2real problem. Learning policies from logged trajectories of a system is a key step forward for RL in physical systems. Recent work has shown that model-based offline RL can improve the effectiveness of offline policy estimation through uncertainty-penalized exploration. However, existing work assumes a continuous state space and access to a covariance matrix of the environment dynamics, which is not possible in the discrete case. To solve this problem, we propose a Bayesian model-based technique that naturally produces probabilistic estimates of the environment dynamics via the posterior predictive distribution, which we use for uncertainty-penalized exploration. We call our approach Posterior Penalized Offline Policy Optimization (PPOPO). We show that our world model better fits historical data due to informative priors, and that PPOPO outperforms other offline techniques in simulation and against real-world data.

Current neural network architectures have no mechanism for explicitly reasoning about item trade-offs. Such trade-offs are important for popular tasks such as recommendation. The main idea of this work is to give neural networks inductive biases that are inspired by economic theories. To this end, we propose Neural Utility Functions, which directly optimize the gradients of a neural network so that they are more consistent with utility theory, a mathematical framework for modeling choice among items. We demonstrate that Neural Utility Functions can recover theoretical item relationships better than vanilla neural networks, analytically show existing neural networks are not quasi-concave and do not inherently reason about trade-offs, and that augmenting existing models with a utility loss function improves recommendation results. The Neural Utility Functions we propose are theoretically motivated, and yield strong empirical results.

Small data has been a barrier for many machine learning tasks, especially when applied in scientific domains. Fortunately, we can utilize domain knowledge to make up the lack of data. Hence, in this paper, we propose a hybrid model KRL that treats domain knowledge model as a weak learner and uses another neural net model to boost it. We prove that KRL is guaranteed to improve over pure domain knowledge model and pure neural net model under certain loss functions. Extensive experiments have shown the superior performance of KRL over baselines. In addition, several case studies have explained how the domain knowledge can assist the prediction.

The 1st Workshop on City Brain Research examines the current challenges and recent breakthroughs related to intelligent urban transportation. The workshop will be organized in a novel form --- offering debates on three main components involved in the transportation policy development cycle: data collection, policy learning, and the effects on human behavior. The organizers intend to invite speakers and attendees from different backgrounds, ranging from computer science, transportation, to urban planning. The final outcomes include live discussions of the three consistent topics, a comprehensive annual report summarizing current practices and future directions, and a detailed tutorial on the workshop day.

Understanding content at scale is a difficult but important problem for many platforms. Many previous studies focus on content understanding to optimize engagement with existing users. However, little work studies how to leverage better content understanding to attract new users. In this work, we build a framework for generating natural language content annotations and show how they can be used for search engine optimization. The proposed framework relies on an XGBoost model that labels “pins” with high probability phrases, and a logistic regression layer that learns to rank aggregated annotations for groups of content. The pipeline identifies keywords that are descriptive and contextually meaningful. We perform a large-scale production experiment deployed on the Pinterest platform and show that natural language annotations cause a 1-2% increase in traffic from leading search engines. This increase is statistically significant. Finally, we explore and interpret the characteristics of our annotations framework.

Increasing urbanization across the globe has coincided with greater access to urban data; this enables researchers and city administrators with better tools to understand urban dynamics, such as crime, traffic, and living standards. In this paper, we study the Learning an Embedding Space for Regions (LESR) problem, wherein we aim to produce vector representations of discrete regions. Recent studies have shown that embedding geospatial regions in a latent vector space can be useful in a variety of urban computing tasks. However, previous studies do not consider regions across multiple modalities in an end-to-end framework. We argue that doing so facilitates the learning of greater semantic relationships among regions. We propose a novel method, RegionEncoder, that jointly learns region representations from satellite image, point-of-interest, human mobility, and spatial graph data. We demonstrate that these region embeddings are useful as features in two regression tasks and across two distinct urban environments. Additionally, we perform an ablation study that evaluates each major architectural component. Finally, we qualitatively explore the learned embedding space, and show that semantic relationships are discovered across modalities

The proliferation of publicly accessible urban data provide new insights on various urban tasks. A frequently used approach is to treat each region as a data sample and build a model over all the regions to observe the correlations between urban features (e.g., demographics) and the target variable (e.g., crime count). To define regions, most existing studies use fixed grids or pre-defined administrative boundaries (e.g., census tracts or community areas). In reality, however, definitions of regions should be different depending on tasks (e.g., regional crime count prediction vs. real estate prices estimation). In this paper, we propose a new problem of task-specific city region partitioning, aiming to find the best partition in a city w.r.t. a given task. We prove this is an NP-hard search problem with no trivial solution. To learn the partition, we first study two variants of Markov Chain Monte Carlo (MCMC). We further propose a reinforcement learning scheme for effective sampling the search space. We conduct experiments on two real datasets in Chicago (i.e., crime count and real estate price) to demonstrate the effectiveness of our proposed method.