The paper introduces submodular information measures for machine learning applications.
problem Generalizing information-theoretic measures to non-random variables.
method Developing combinatorial information measures based on submodular functions.
result Submodular mutual information is submodular in one argument for certain submodular functions.
In this paper, we present an algorithm for minimizing the difference between two submodular functions using a variational framework which is based on (an extension of) the concave-convex procedure [17]. Because several commonly used metrics in machine learning, like mutual information and conditional mutual information…
In the era of big data, learning from categorical features with very large vocabularies (e.g., 28 million for the Criteo click prediction dataset) has become a practical challenge for machine learning researchers and practitioners. We design a highly-scalable vocabulary compression algorithm that seeks to maximize the …
New method selects features via tensor decomposition and submodular optimization.
problem Feature selection for high-dimensional data.
method Low-rank tensor model, submodular optimization, greedy algorithm.
result Proposed method outperforms state-of-the-art feature selection.
The paper tackles model selection for unseen tasks by capturing relationships among checkpoints.
problem Deciding which model combinations are likely to be effective for a new task is difficult.
method The paper models the task space as a Gaussian process and identifies representative checkpoints using mutual information and a greedy algorithm.
result Representative checkpoints generalize to new tasks with superior performance.
We consider the problem of optimal budget allocation for crowdsourcing problems, allocating users to tasks to maximize our final confidence in the crowdsourced answers. Such an optimized worker assignment method allows us to boost the efficacy of any popular crowdsourcing estimation algorithm. We consider a mutual info…
Submodularity is one of the most well-studied properties of problem classes in combinatorial optimization and many applications of machine learning and data mining, with strong implications for guaranteed optimization. In this thesis, we investigate the role of submodularity in provable non-convex optimization and vali…
Study on maximizing submodular functions with limited updates, achieving tight bounds and poly-time algorithms.
problem Online submodular maximization with constant recourse.
method Information-theoretic bounds and poly-time randomized algorithms.
result Achieved tight bounds of 2/3 and 3/4 for general and coverage functions, respectively, with a 0.51 approximation.
The ability to integrate information in the brain is considered to be an essential property for cognition and consciousness. Integrated Information Theory (IIT) hypothesizes that the amount of integrated information (Φ) in the brain is related to the level of consciousness. IIT proposes that to quantify information i…
A new framework uses directed information to efficiently select context chunks.
problem Efficiently selecting relevant context chunks for query understanding.
method Directed Information γ-covering framework, formulated as a γ-cover problem, with a greedy algorithm for context selection. result The γ-covering algorithm provides clear advantages in hard-decision regimes like context compression and single-slot prompt selection. Optimal subset selection for hypothesis testing with penalties.
problem Optimal subset selection of information sources for hypothesis testing with misclassification penalties.
method Proposes a misclassification penalty framework and studies two variants of subset selection problems under centralized Bayesian learning.
result Proves the submodularity of the objective and constraints of the subset selection problems and establishes performance guarantees for greedy algorithms.
In this paper, we propose three online algorithms for submodular maximisation. The first one, Mono-Frank-Wolfe, reduces the number of per-function gradient evaluations from T1/2 [Chen2018Online] and T3/2 [chen2018projection] to 1, and achieves a (1−1/e)-regret bound of O(T4/5). The second one, Bandit-F…
In this paper, we consider the problem of black box continuous submodular maximization where we only have access to the function values and no information about the derivatives is provided. For a monotone and continuous DR-submodular function, and subject to a bounded convex body constraint, we propose Black-box Contin…
Paper proposes DG-ETC for online submodular maximization with stochastic bandit feedback.
problem Online unconstrained submodular maximization with stochastic bandit feedback.
method Double-Greedy - Explore-then-Commit (DG-ETC) approach.
result DG-ETC achieves logarithmic regret O(dlog(dT)) for 1/2-approximate pseudo-regret. This paper studies robust payoff allocation in submodular games, especially against replication.
problem Payoff allocation in submodular games, especially robustness against replication.
method Systematically studied replication manipulation in submodular games, introduced replication robustness metric, and validated with empirical ML data market.
result Conditions characterizing robustness of semivalues in submodular games.
Mini-batch gradient descent based methods are the de facto algorithms for training neural network architectures today. We introduce a mini-batch selection strategy based on submodular function maximization. Our novel submodular formulation captures the informativeness of each sample and diversity of the whole subset. W…
In this paper, we study the problem of monotone (weakly) DR-submodular continuous maximization. While previous methods require the gradient information of the objective function, we propose a derivative-free algorithm LDGM for the first time. We define β and α to characterize how close a function is to continuous D…
The paper studies continuous submodular functions and their optimization.
problem Maximizing continuous submodular functions in poly. time.
method Characterization of continuous submodularity, operations preserving it, and algorithms for constrained maximization.
result Continuous submodularity is equivalent to a weak DR property, leading to continuous DR-submodular functions with the full DR property.
New neural network approach using mutual information.
problem Training neural networks for imbalanced datasets.
method Converts neural network classifiers to mutual information evaluators.
result New form of softmax leads to better classification accuracy, especially for imbalanced datasets.
Paper describes profiles of multivariate normal distributions and novel estimators for mutual information.
problem Estimating mutual information for complex distributions.
method Analytical description of profiles, introduction of Bend and Mix Models, Monte Carlo estimation.
result Bend and Mix Models accurately estimate mutual information profiles and provide Bayesian estimates.
Improved bounds on learning algorithms' performance using conditional mutual information.
problem Bounding the generalization error of learning algorithms.
method Introducing conditional mutual information and disintegrated mutual information to tighten bounds.
result New bounds are tighter than previous ones, especially for noisy, iterative algorithms.
Paper benchmarks mutual info estimators on diverse distributions.
problem Evaluating mutual information estimators on complex, real-world distributions.
method Constructs a diverse family of known-ground truth distributions, proposes a benchmark platform.
result Highlights differences in classical and neural estimators' performance across various conditions.
A new algorithm ThreeSieves maximizes submodular functions efficiently in streaming data.
problem Maximizing submodular functions in high-dimensional data streams efficiently.
method ThreeSieves algorithm, designed for streaming data, ignoring worst-case guarantees for high probability solutions.
result ThreeSieves outperforms state-of-the-art methods in terms of performance and resource usage.
The paper argues that normalized mutual information is biased in clustering and community detection.
problem Bias in normalized mutual information for clustering and community detection.
method Introducing a modified version of mutual information to correct for information content and spurious dependence.
result The modified mutual information leads to different conclusions about which algorithms are best for community detection.
Submodularity is studied for convex risk measures, including Expected Shortfall.
problem Characterizing submodularity in convex risk measures.
method Analyzing submodularity properties of law-invariant coherent risk measures, including Expected Shortfall and Value-at-Risk.
result AES is submodular only when it reduces to ES, and empirical analysis shows AES violations are less frequent than VaR and ES violations.
New method for probabilistic modeling of integer submodular functions.
problem Lack of probabilistic modeling for integer submodular functions.
method Proposed Generalized Multilinear Extension and block-coordinate ascent algorithm.
result Demonstrated effectiveness and viability on real-world datasets.
Neural estimator improves mutual information estimation in high dimensions.
problem Estimating mutual information in high dimensions is challenging.
method Parametrizing conditional densities with normalizing flows and using block autoregressive structure.
result Improved mutual information estimation on benchmark tasks.
In this paper we develop the first algorithms for online submodular minimization that preserve differential privacy under full information feedback and bandit feedback. A sequence of T submodular functions over a collection of n elements arrive online, and at each timestep the algorithm must choose a subset of $[n]…
New algorithm for online convex minimization over integer lattice.
problem Online decision-making with nonlinear combinatorial objectives.
method Introduces online Latural-convex minimization and proposes efficient algorithms. result Tight regret bound for full information setting algorithm.
We find the maximum mutual information for neural networks and its key determinants.
problem Understanding the maximum mutual information in neural architectures.
method Derived closed-form expression for maximum mutual information across neural network families.
result Maximum mutual information stems from a generalized formula and is influenced by network width and statistical invariances.
Measuring mutual information from finite data is difficult. Recent work has considered variational methods maximizing a lower bound. In this paper, we prove that serious statistical limitations are inherent to any method of measuring mutual information. More specifically, we show that any distribution-free high-confide…
Improved method for encoding contingency tables reduces mutual information bias.
problem Mutual information bias in measuring label similarity.
method Improved method for encoding contingency tables to reduce information cost.
result Better bound on reduced mutual information in typical use cases.
Submodular function minimization is well studied, and existing algorithms solve it exactly or up to arbitrary accuracy. However, in many applications, such as structured sparse learning or batch Bayesian optimization, the objective function is not exactly submodular, but close. In this case, no theoretical guarantees e…
Proposes a new bound on generalization error using conditional mutual information.
problem Improving the generalization error bound in machine learning.
method Combines error decomposition and conditional mutual information techniques.
result New bound is order-wise better than previous ones in a simple Gaussian setting.
Differentially private algorithms for submodular maximization under various constraints.
problem Maximizing decomposable submodular functions under constraints while preserving privacy.
method Designing differentially private algorithms for both monotone and non-monotone decomposable submodular maximization under general matroid constraints.
result Improved utility guarantees and competitive performance compared to non-private algorithms.
Proposes a framework to maximize mutual information in VAE models for better latent code representation.
problem Lack of explicit measurement of the quality of learned representations in VAE models.
method Variational Mutual Information Maximization Framework for VAE.
result Maximizes mutual information between latent codes and observations, improving latent code representation.
Online optimization has been a successful framework for solving large-scale problems under computational constraints and partial information. Current methods for online convex optimization require either a projection or exact gradient computation at each step, both of which can be prohibitively expensive for large-scal…
Approximate inference via information projection has been recently introduced as a general-purpose approach for efficient probabilistic inference given sparse variables. This manuscript goes beyond classical sparsity by proposing efficient algorithms for approximate inference via information projection that are applica…
We introduce the Mutual Information Machine (MIM), a novel formulation of representation learning, using a joint distribution over the observations and latent state in an encoder/decoder framework. Our key principles are symmetry and mutual information, where symmetry encourages the encoder and decoder to learn differe…
A new method estimates mutual information using neural classifiers.
problem Estimating mutual information for high-dimensional data is challenging.
method Trains a classifier to estimate joint distribution probability.
result Demonstrates high accuracy and reduces variance compared to variational methods.
We introduce a method to learn a mixture of submodular "shells" in a large-margin setting. A submodular shell is an abstract submodular function that can be instantiated with a ground set and a set of parameters to produce a submodular function. A mixture of such shells can then also be so instantiated to produce a mor…
New estimator improves mutual information estimation.
problem Estimating mutual information in data science and machine learning.
method Proposes a new estimator that uses a preliminary estimate of the data distribution.
result A preliminary estimate helps in estimating mutual information more accurately.
The paper proposes a method to learn structured representations from unlabeled data using mutual information maximization.
problem Learning structured representations from unlabeled data.
method Adversarial maximization of mutual information between a structured latent variable and a target variable.
result The proposed method outperforms current baselines in document hashing and yields highly compressed interpretable representations.
Paper proposes mutual information learning for deep learning classifiers.
problem Overfitting in deep learning models.
method Mutual information learning framework to train classifiers.
result MILCs achieve better generalization than conditional entropy classifiers.
We are motivated by large scale submodular optimization problems, where standard algorithms that treat the submodular functions in the \emph{value oracle model} do not scale. In this paper, we present a model called the \emph{precomputational complexity model}, along with a unifying memoization based framework, which l…
New algorithm maximizes non-monotone adaptive submodular functions in linear time.
problem Maximizing non-monotone adaptive submodular functions subject to a cardinality constraint.
method Developed a linear-time algorithm for non-monotone adaptive submodular maximization.
result Achieved a 1/e−ε approximation ratio with O(nε−2logε−1) value oracle queries. TCMI assesses mutual dependence of continuous variables without parametric assumptions.
problem Estimating mutual information from continuous distributions.
method TCMI extends mutual information to continuous variables using cumulative distributions.
result TCMI facilitates feature selection and ranking of variable sets.
A new kernel for ranked data tackles computational challenges.
problem Complex geometric structure and partial rankings make existing algorithms infeasible for real-world applications.
method Derives a graph cut kernel that combines submodular optimization and kernel-based methods.
result The graph cut kernel efficiently handles large-scale ranked data.