arXiv subjects

Denali Molitor

Publications and source records attributed to Denali Molitor.

17 recordsLinked to original sources

arXiv2023

Massively Scalable Inverse Reinforcement Learning in Google Maps

Inverse reinforcement learning (IRL) offers a powerful and general framework for learning humans' latent preferences in route recommendation, yet no approach has successfully addressed planetary-scale problems with hundreds of millions of states and demonstration trajectories. In this paper, we introduce scaling techniques based on graph compression, spatial parallelization, and improved initialization conditions inspired by a connection to eigenvector algorithms. We revisit classic IRL methods in the routing context, and make the key observation that there exists a trade-off between the use of cheap, deterministic planners and expensive yet robust stochastic policies. This insight is leveraged in Receding Horizon Inverse Planning (RHIP), a new generalization of classic IRL algorithms that provides fine-grained control over performance trade-offs via its planning horizon. Our contributions culminate in a policy that achieves a 16-24% improvement in route quality at a global scale, and to the best of our knowledge, represents the largest published study of IRL algorithms in a real-world setting to date. We conclude by conducting an ablation study of key components, presenting negative results from alternative eigenvalue solvers, and identifying opportunities to further improve scalability via IRL-specific batching strategies.

cs.LG↗

arXiv2023

Neural Nonnegative Matrix Factorization for Hierarchical Multilayer Topic Modeling

We introduce a new method based on nonnegative matrix factorization, Neural NMF, for detecting latent hierarchical structure in data. Datasets with hierarchical structure arise in a wide variety of fields, such as document classification, image processing, and bioinformatics. Neural NMF recursively applies NMF in layers to discover overarching topics encompassing the lower-level features. We derive a backpropagation optimization scheme that allows us to frame hierarchical NMF as a neural network. We test Neural NMF on a synthetic hierarchical dataset, the 20 Newsgroups dataset, and the MyLymeData symptoms dataset. Numerical results demonstrate that Neural NMF outperforms other hierarchical NMF methods on these data sets and offers better learned hierarchical structure and interpretability of topics.

cs.LG↗

arXiv2023

Structural Equivalence in Subgraph Matching

Symmetry plays a major role in subgraph matching both in the description of the graphs in question and in how it confounds the search process. This work addresses how to quantify these effects and how to use symmetries to increase the efficiency of subgraph isomorphism algorithms. We introduce rigorous definitions of structural equivalence and establish conditions for when it can be safely used to generate more solutions. We illustrate how to adapt standard search routines to utilize these symmetries to accelerate search and compactly describe the solution space. We then adapt a state-of-the-art solver and perform a comprehensive series of tests to demonstrate these methods' efficacy on a standard benchmark set. We extend these methods to multiplex graphs and present results on large multiplex networks drawn from transportation systems, social media, adversarial attacks, and knowledge graphs.

cs.DS↗

arXiv2022

Inference of Media Bias and Content Quality Using Natural-Language Processing

Media bias can significantly impact the formation and development of opinions and sentiments in a population. It is thus important to study the emergence and development of partisan media and political polarization. However, it is challenging to quantitatively infer the ideological positions of media outlets. In this paper, we present a quantitative framework to infer both political bias and content quality of media outlets from text, and we illustrate this framework with empirical experiments with real-world data. We apply a bidirectional long short-term memory (LSTM) neural network to a data set of more than 1 million tweets to generate a two-dimensional ideological-bias and content-quality measurement for each tweet. We then infer a ``media-bias chart'' of (bias, quality) coordinates for the media outlets by integrating the (bias, quality) measurements of the tweets of the media outlets. We also apply a variety of baseline machine-learning methods, such as a naive-Bayes method and a support-vector machine (SVM), to infer the bias and quality values for each tweet. All of these baseline approaches are based on a bag-of-words approach. We find that the LSTM-network approach has the best performance of the examined methods. Our results illustrate the importance of leveraging word order into machine-learning methods in text analysis.

physics.soc-ph↗

arXiv2021

Analysis of Legal Documents via Non-negative Matrix Factorization Methods

The California Innocence Project (CIP), a clinical law school program aiming to free wrongfully convicted prisoners, evaluates thousands of mails containing new requests for assistance and corresponding case files. Processing and interpreting this large amount of information presents a significant challenge for CIP officials, which can be successfully aided by topic modeling techniques.In this paper, we apply Non-negative Matrix Factorization (NMF) method and implement various offshoots of it to the important and previously unstudied data set compiled by CIP. We identify underlying topics of existing case files and classify request files by crime type and case status (decision type). The results uncover the semantic structure of current case files and can provide CIP officials with a general understanding of newly received case files before further examinations. We also provide an exposition of popular variants of NMF with their experimental results and discuss the benefits and drawbacks of each variant through the real-world application.

cs.LG↗

arXiv2020

Sparseness-constrained Nonnegative Tensor Factorization for Detecting Topics at Different Time Scales

Temporal data (such as news articles or Twitter feeds) often consists of a mixture of long-lasting trends and popular but short-lasting topics of interest. A truly successful topic modeling strategy should be able to detect both types of topics and clearly locate them in time. In this paper, we first show that nonnegative CANDECOMP/PARAFAC decomposition (NCPD) is able to discover topics of variable persistence automatically. Then, we propose sparseness-constrained NCPD (S-NCPD) and its online variant in order to actively control the length of the learned topics effectively and efficiently. Further, we propose quantitative ways to measure the topic length and demonstrate the ability of S-NCPD (as well as its online variant) to discover short and long-lasting temporal topics in a controlled manner in semi-synthetic and real-world data including news headlines. We also demonstrate that the online variant of S-NCPD reduces the reconstruction error more rapidly than S-NCPD.

cs.IR↗

arXiv2020

Randomized Kaczmarz for Tensor Linear Systems

Solving linear systems of equations is a fundamental problem in mathematics. When the linear system is so large that it cannot be loaded into memory at once, iterative methods such as the randomized Kaczmarz method excel. Here, we extend the randomized Kaczmarz method to solve multi-linear (tensor) systems under the tensor-tensor t-product. We provide convergence guarantees for the proposed tensor randomized Kaczmarz that are analogous to those of the randomized Kaczmarz method for matrix linear systems. We demonstrate experimentally that the tensor randomized Kaczmarz method converges faster than traditional randomized Kaczmarz applied to a naively matricized version of the linear system. In addition, we draw connections between the proposed algorithm and a previously known extension of the randomized Kaczmarz algorithm for matrix linear systems.

math.NA↗

arXiv2020

Randomized Kaczmarz with Averaging

The randomized Kaczmarz (RK) method is an iterative method for approximating the least-squares solution of large linear systems of equations. The standard RK method uses sequential updates, making parallel computation difficult. Here, we study a parallel version of RK where a weighted average of independent updates is used. We analyze the convergence of RK with averaging and demonstrate its performance empirically. We show that as the number of threads increases, the rate of convergence improves and the convergence horizon for inconsistent systems decreases.

math.NA↗

arXiv2019

Adaptive Sketch-and-Project Methods for Solving Linear Systems

We present new adaptive sampling rules for the sketch-and-project method for solving linear systems. To deduce our new sampling rules, we first show how the progress of one step of the sketch-and-project method depends directly on a sketched residual. Based on this insight, we derive a 1) max-distance sampling rule, by sampling the sketch with the largest sketched residual 2) a proportional sampling rule, by sampling proportional to the sketched residual, and finally 3) a capped sampling rule. The capped sampling rule is a generalization of the recently introduced adaptive sampling rules for the Kaczmarz method. We provide a global linear convergence theorem for each sampling rule and show that the max-distance rule enjoys the fastest convergence. This finding is also verified in extensive numerical experiments that lead us to conclude that the max-distance sampling rule is superior both experimentally and theoretically to the capped sampling rule. We also provide numerical insights into implementing the adaptive strategies so that the per iteration cost is of the same order as using a fixed sampling strategy when the number of sketches times the sketch size is not significantly larger than the number of columns.

math.NA↗

arXiv2019

Bias of Homotopic Gradient Descent for the Hinge Loss

Gradient descent is a simple and widely used optimization method for machine learning. For homogeneous linear classifiers applied to separable data, gradient descent has been shown to converge to the maximal margin (or equivalently, the minimal norm) solution for various smooth loss functions. The previous theory does not, however, apply to non-smooth functions such as the hinge loss which is widely used in practice. Here, we study the convergence of a homotopic variant of gradient descent applied to the hinge loss and provide explicit convergence rates to the max-margin solution for linearly separable data.

stat.ML↗

arXiv2019

On Inferences from Completed Data

Matrix completion has become an extremely important technique as data scientists are routinely faced with large, incomplete datasets on which they wish to perform statistical inferences. We investigate how error introduced via matrix completion affects statistical inference. Furthermore, we prove recovery error bounds which depend upon the matrix recovery error for several common statistical inferences. We consider matrix recovery via nuclear norm minimization and a variant, $\ell_1$-regularized nuclear norm minimization for data with a structured sampling pattern. Finally, we run a series of numerical experiments on synthetic data and real patient surveys from MyLymeData, which illustrate the relationship between inference recovery error and matrix recovery error. These results indicate that exact matrix recovery is often not necessary to achieve small inference recovery error.

math.ST↗

arXiv2019

Model Agnostic Supervised Local Explanations

Model interpretability is an increasingly important component of practical machine learning. Some of the most common forms of interpretability systems are example-based, local, and global explanations. One of the main challenges in interpretability is designing explanation systems that can capture aspects of each of these explanation types, in order to develop a more thorough understanding of the model. We address this challenge in a novel model called MAPLE that uses local linear modeling techniques along with a dual interpretation of random forests (both as a supervised neighborhood approach and as a feature selection method). MAPLE has two fundamental advantages over existing interpretability systems. First, while it is effective as a black-box explanation system, MAPLE itself is a highly accurate predictive model that provides faithful self explanations, and thus sidesteps the typical accuracy-interpretability trade-off. Specifically, we demonstrate, on several UCI datasets, that MAPLE is at least as accurate as random forests and that it produces more faithful local explanations than LIME, a popular interpretability system. Second, MAPLE provides both example-based and local explanations and can detect global patterns, which allows it to diagnose limitations in its local explanations.

cs.LG↗

arXiv2018

An iterative method for classification of binary data

In today's data driven world, storing, processing, and gleaning insights from large-scale data are major challenges. Data compression is often required in order to store large amounts of high-dimensional data, and thus, efficient inference methods for analyzing compressed data are necessary. Building on a recently designed simple framework for classification using binary data, we demonstrate that one can improve classification accuracy of this approach through iterative applications whose output serves as input to the next application. As a side consequence, we show that the original framework can be used as a data preprocessing step to improve the performance of other methods, such as support vector machines. For several simple settings, we showcase the ability to obtain theoretical guarantees for the accuracy of the iterative classification method. The simplicity of the underlying classification framework makes it amenable to theoretical analysis and studying this approach will hopefully serve as a step toward developing theory for more sophisticated deep learning technologies.

stat.ML↗

arXiv2018

Hierarchical Classification using Binary Data

In classification problems, especially those that categorize data into a large number of classes, the classes often naturally follow a hierarchical structure. That is, some classes are likely to share similar structures and features. Those characteristics can be captured by considering a hierarchical relationship among the class labels. Here, we extend a recent simple classification approach on binary data in order to efficiently classify hierarchical data. In certain settings, specifically, when some classes are significantly easier to identify than others, we showcase computational and accuracy advantages.

cs.LG↗

arXiv2018

Matrix Completion for Structured Observations

The need to predict or fill-in missing data, often referred to as matrix completion, is a common challenge in today's data-driven world. Previous strategies typically assume that no structural difference between observed and missing entries exists. Unfortunately, this assumption is woefully unrealistic in many applications. For example, in the classic Netflix challenge, in which one hopes to predict user-movie ratings for unseen films, the fact that the viewer has not watched a given movie may indicate a lack of interest in that movie, thus suggesting a lower rating than otherwise expected. We propose adjusting the standard nuclear norm minimization strategy for matrix completion to account for such structural differences between observed and unobserved entries by regularizing the values of the unobserved entries. We show that the proposed method outperforms nuclear norm minimization in certain settings.

math.NA↗

arXiv2014

The Generalized Symmetric Tequila Problem: Influence and Independence in N-Player Games

This paper extends results from Mike Steel and Amelia Taylor's paper The Structure of Symmetric N-Player Games when Influence and Independence Collide. These games include n causes, which are dichotomous random variables whose values determine the probabilities of the values of n dichotomous effects. We denote the probability spaces that exhibit independence and influence among n players as Ind_n and Inf_n respectively. We define the solution space of the "generalized symmetric tequila problem," GST_n, as the set of probabilities for a set of given effects such that the causes and effects are independent and each cause influences the effects, that is GST_n is the intersection of Ind_n and Inf_n. Steel and Taylor showed that GST_n is connected for n greater than or equal to 8 and disconnected for n = 3, 4. We prove that for n = 5, 6, 7, GST_n is connected and determine the number of connected components of GST_4.

math.PR↗

arXiv2014

Using Peano Curves to Construct Laplacians on Fractals

We describe a new method to construct Laplacians on fractals using a Peano curve from the circle onto the fractal, extending an idea that has been used in the case of certain Julia sets. The Peano curve allows us to visualize eigenfunctions of the Laplacian by graphing the pullback to the circle. We study in detail three fractals: the pentagasket, the octagasket and the magic carpet. We also use the method for two nonfractal self-similar sets, the torus and the equilateral triangle, obtaining appealing new visualizations of eigenfunctions on the triangle. In contrast to the many familiar pictures of approximations to standard Peano curves, that do no show self-intersections, our descriptions of approximations to the Peano curves have self-intersections that play a vital role in constructing graph approximations to the fractal with explicit graph Laplacians that give the fractal Laplacian in the limit.

math.FA↗