One of the challenges in machine learning research is to ensure that presented and published results are sound and reliable. Reproducibility, that is obtaining similar results as presented in a paper or talk, using the same code and data (when available), is a necessary step to verify the reliability of research findings. Reproducibility is also an important step to promote open and accessible research, thereby allowing the scientific community to quickly integrate new findings and convert ideas to practice. Reproducibility also promotes the use of robust experimental workflows, which potentially reduce unintentional errors. In 2019, the Neural Information Processing Systems (NeurIPS) conference, the premier international conference for research in machine learning, introduced a reproducibility program, designed to improve the standards across the community for how we conduct, communicate, and evaluate machine learning research. The program contained three components: a code submission policy, a community-wide reproducibility challenge, and the inclusion of the Machine Learning Reproducibility checklist as part of the paper submission process. In this paper, we describe each of these components, how it was deployed, as well as what we were able to learn from this initiative.

We study the problem of efficient online multiclass linear classification with bandit feedback, where all examples belong to one of $K$ classes and lie in the $d$-dimensional Euclidean space. Previous works have left open the challenge of designing efficient algorithms with finite mistake bounds when the data is linearly separable by a margin $\gamma$. In this work, we take a first step towards this problem. We consider two notions of linear separability: strong and weak. 1. Under the strong linear separability condition, we design an efficient algorithm that achieves a near-optimal mistake bound of $O\left(\frac{K}{\gamma^2} \right)$. 2. Under the more challenging weak linear separability condition, we design an efficient algorithm with a mistake bound of $2^{\widetilde{O}(\min(K \log^2 \frac{1}{\gamma}, \sqrt{\frac{1}{\gamma}} \log K))}$. Our algorithm is based on kernel Perceptron, which is inspired by the work of Klivans & Servedio (2008) on improperly learning intersection of halfspaces.

We design and study a Contextual Memory Tree (CMT), a learning memory controller that inserts new memories into an experience store of unbounded size. It operates online and is designed to efficiently query for memories from that store, supporting logarithmic time insertion and retrieval operations. Hence CMT can be integrated into existing statistical learning algorithms as an augmented memory unit without substantially increasing training and inference computation. Furthermore CMT operates as a reduction to classification, allowing it to benefit from advances in representation or architecture. We demonstrate the efficacy of CMT by augmenting existing multi-class and multi-label classification algorithms with CMT and observe statistical improvement. We also test CMT learning on several image-captioning tasks to demonstrate that it performs computationally better than a simple nearest neighbors memory system while benefitting from reward learning.

We present a systematic approach for achieving fairness in a binary classification setting. While we focus on two well-known quantitative definitions of fairness, our approach encompasses many other previously studied definitions as special cases. The key idea is to reduce fair classification to a sequence of cost-sensitive classification problems, whose solutions yield a randomized classifier with the lowest (empirical) error subject to the desired constraints. We introduce two reductions that work for any representation of the cost-sensitive classifier and compare favorably to prior baselines on a variety of data sets, while overcoming several of their disadvantages.

We investigate active learning with access to two distinct oracles: LABEL (which is standard) and SEARCH (which is not). The SEARCH oracle models the situation where a human searches a database to seed or counterexample an existing solution. SEARCH is stronger than LABEL while being natural to implement in many situations. We show that an algorithm using both oracles can provide exponentially large problem-dependent improvements over LABEL alone.

In this paper, we provide a summary of the mathematical and computational techniques that have enabled learning reductions to effectively address a wide class of tasks, and show that this approach to solving machine learning problems can be broadly useful. Our work is instantiated and tested in a machine learning library, Vowpal Wabbit, to prove that the techniques discussed here are fully viable in practice.

We study online boosting, the task of converting any weak online learner into a strong online learner. Based on a novel and natural definition of weak online learnability, we develop two online boosting algorithms. The first algorithm is an online version of boost-by-majority. By proving a matching lower bound, we show that this algorithm is essentially optimal in terms of the number of weak learners and the sample complexity needed to achieve a specified accuracy. The second algorithm is adaptive and parameter-free, albeit not optimal.

We extend the theory of boosting for regression problems to the online learning setting. Generalizing from the batch setting for boosting, the notion of a weak learning algorithm is modeled as an online learning algorithm with linear loss functions that competes with a base class of regression functions, while a strong learning algorithm is an online learning algorithm with smooth convex loss functions that competes with a larger class of regression functions. Our main result is an online gradient boosting algorithm which converts a weak online learning algorithm into a strong one where the larger class of functions is the linear span of the base class. We also give a simpler boosting algorithm that converts a weak online learning algorithm into a strong one where the larger class of functions is the convex hull of the base class, and prove its optimality.

We present and analyze an agnostic active learning algorithm that works without keeping a version space. This is unlike all previous approaches where a restricted set of candidate hypotheses is maintained throughout learning, and only hypotheses from this set are ever returned. By avoiding this version space approach, our algorithm sheds the computational burden and brittleness associated with maintaining version spaces, yet still allows for substantial improvements over supervised learning for classification.

We present a practical and statistically consistent scheme for actively learning binary classifiers under general loss functions. Our algorithm uses importance weighting to correct sampling bias, and by controlling the variance, we are able to give rigorous label complexity bounds for the learning process.

No abstract available.

We present an algorithm, called the Offset Tree, for learning to make decisions in situations where the payoff of only one choice is observed, rather than all choices. The algorithm reduces this setting to binary classification, allowing one to reuse any existing, fully supervised binary classification algorithm in this partial information setting. We show that the Offset Tree is an optimal reduction to binary classification. In particular, it has regret at most (k-1) times the regret of the binary classifier it uses (where k is the number of choices), and no reduction to binary classification can do better. This reduction is also computationally optimal, both at training and test time, requiring just O(log2 k) work to train on an example or make a prediction.

Basic properties of wireless ad-hoc networks, such as connectivity, have traditionally been analyzed and evaluated under the assumption that all nodes have homogeneous communication capabilities. While this assumption usually holds for small networks, it is unlikely to be true in most large scale networks. This paper studies how heterogeneous communication capabilities affect basic properties of ad-hoc networks. More specifically, we consider one type of heterogeneity in which nodes are equipped with two different types of antenna technologies, omni-directional and directional, and evaluate them against four properties: network connectivity, energy consumption, interference tolerance and resilience to failures. Our main results show that even under a very simple network model that does not assume any global knowledge of the network, an ad-hoc network where only 10% to 20% of the nodes are equipped with directional antennas, can outperform a comparable homogeneous network (where all nodes have only omni-directional antennas) in all four metrics.

Discovering the topology of a network and detecting link state changes (e.g.: link failures) is an essential element for various network management and monitoring tasks. In this paper, we investigate scalable mechanisms to monitor the topology and link states of networks based on information available in network nodes' routing tables. We first present an algorithm that infers the network topology based on the full or partial information about network distances between nodes, based on which we obtain a scalable network topology discovery solution via a novel use of random walk in graphs. We then present scalable algorithms to detect the state changes of remote links by monitoring the routing tables of a small fraction of the routers, where the routers to be monitored are selected by a greedy approach to an NP-complete Tree Cover problem. We show the efficacy and scalability of our topology monitoring algorithms through experimental evaluation performed both on synthetic topologies and on a large topology data-set from a real enterprise network.

We state and analyze the first active learning algorithm which works in the presence of arbitrary forms of noise. The algorithm, A2 (for Agnostic Active), relies only upon the assumption that the samples are drawn i.i.d. from a fixed distribution. We show that A2 achieves an exponential improvement (i.e., requires only O (ln 1/ε) samples to find an ε-optimal classifier) over the usual sample complexity of supervised learning, for several settings considered before in the realizable case. These include learning threshold classifiers and learning homogeneous linear separators with respect to an input distribution which is uniform over the unit sphere.

We present a tree data structure for fast nearest neighbor operations in general n-point metric spaces (where the data set consists of n points). The data structure requires O(n) space regardless of the metric's structure yet maintains all performance properties of a navigating net (Krauthgamer & Lee, 2004b). If the point set has a bounded expansion constant c, which is a measure of the intrinsic dimensionality, as defined in (Karger & Ruhl, 2002), the cover tree data structure can be constructed in O (c6n log n) time. Furthermore, nearest neighbor queries require time only logarithmic in n, in particular O (c12 log n) time. Our experimental results show speedups over the brute force search varying between one and several orders of magnitude on natural machine learning datasets.

We introduce a reduction-based model for analyzing supervised learning tasks. We use this model to devise a new reduction from multi-class cost-sensitive classification to binary classification with the following guarantee: If the learned binary classifier has error rate at most ε then the cost-sensitive classifier has cost at most 2ε times the expected sum of costs of all possible lables. Since cost-sensitive classification can embed any bounded loss finite choice supervised learning task, this result shows that any such task can be solved using a binary classification oracle. Finally, we present experimental results showing that our new reduction outperforms existing algorithms for multi-class cost-sensitive learning.

A common problem encountered in many application scenarios is how to represent some prior knowledge about a system in order to determine its true state as efficiently as possible. The information is typically in the form of tests, or questions about the system. Each test can potentially reduce our uncertainty about the system's state. The problem is to represent the information capturing the dependence between tests, their outcomes, and possible states in an efficiently navigable way to aid diagnosis. The most common such representation is a flowchart with leaf nodes corresponding to possible states, and non-leaf nodes corresponding to tests about the state. The problem with flowcharts is that they are notoriously difficult to maintain. Additional knowledge often has to be manually integrated as the system changes, making it impossible to keep track of all possible decision paths, let alone optimize the flow to maximize performance. We propose an efficient method for optimizing an existing flowchart based on a conversion to an auxiliary matrix representation. The main goal of the paper is show a synergy between the two representations in the hope that this will help practitioners choose a better strategy for their applications. We show that such a conversion suggests ways to improve both representations - ways that were not envisioned when using each representation alone. Finally, we show that the two representations are informationally equivalent in the sense that one can be transformed into the other so that if both are used as black-boxes, one would not be able to tell them apart, regardless of which state the system is in.

Irregularly observed time series and their analysis are fundamental for any application in which data are collected in a distributed or asynchronous manor. We propose a theoretical framework for analyzing both stationary and non-stationary irregularly spaced time series. Our models can be viewed as extensions of the well known autoregression (AR) model. We provide experiments suggesting that, in practice, the proposed approach performs well in computing the basic statistics and doing prediction. We also develop a resampling strategy that uses the proposed models to reduce irregular time series to regular time series. This enables us to take advantage of the vast number of approaches developed for analyzing regular time series.

Real-time problem diagnosis in large distributed computer systems and networks is a challenging task that requires fast and accurate inferences from potentially huge data volumes. In this paper, we propose a cost-efficient, adaptive diagnostic technique called active probing. Probes are end-to-end test transactions that collect information about the performance of a distributed system. Active probing uses probabilistic reasoning techniques combined with information-theoretic approach, and allows a fast online inference about the current system state via active selection of only a small number of most-informative tests. We demonstrate empirically that the active probing scheme greatly reduces both the number of probes (from 60% to 75% in most of our real-life applications), and the time needed for localizing the problem when compared with nonadaptive (preplanned) probing schemes. We also provide some theoretical results on the complexity of probe selection, and the effect of "noisy" probes on the accuracy of diagnosis. Finally, we discuss how to model the system's dynamics using dynamic Bayesian networks (DBNs), and an efficient approximate approach called sequential multifault; empirical results demonstrate clear advantage of such approaches over "static" techniques that do not handle system's changes.

We consider the question of how well a given distribution can be approx- imated with probabilistic graphical models. We introduce a new param- eter, effective treewidth, that captures the degree of approximability as a tradeoff between the accuracy and the complexity of approximation. We present a simple approach to analyzing achievable tradeoffs that ex- ploits the threshold behavior of monotone graph properties, and provide experimental results that support the approach.

An important issue in visualizing categorical data is how to order categorical values. The focus of this paper is on constructing such orderings efficiently without compromising their visual quality.