Unified framework for gradient estimation in combinatorial spaces.
problem Scaling relaxed gradient estimators to large combinatorial distributions.
method Introducing stochastic softmax tricks within the perturbation model framework.
result Stochastic softmax tricks improve model performance and discover more latent structure.
In this paper, we propose an unifying view of several recently proposed structured sparsity-inducing norms. We consider the situation of a model simultaneously (a) penalized by a set- function de ned on the support of the unknown parameter vector which represents prior knowledge on supports, and (b) regularized in Lp-n…
LGS-Net improves NCO performance on combinatorial optimization tasks.
problem NP-hard combinatorial optimization problems in logistics, manufacturing, and drug discovery.
method LGS-Net uses a latent space model that conditions on problem instances and introduces Latent Guided Sampling for efficient inference.
result Empirical results show state-of-the-art performance on benchmark routing tasks.
We characterize the combinatorial structure of conditionally-i.i.d. sequences of negative binomial processes with a common beta process base measure. In Bayesian nonparametric applications, such processes have served as models for latent multisets of features underlying data. Analogously, random subsets arise from cond…
GFlowNet-EM learns complex latent variable models with discrete structures.
problem Challenges in modeling posteriors over discrete compositional latents with expectation-maximization.
method Uses GFlowNets to learn stochastic policies for sampling from complex posterior distributions.
result GFlowNet-EM enables training expressive LVMs with discrete compositional latents.
New method for efficient marginalization of discrete latent variables in neural networks.
problem Computational challenges in training models with discrete latent variables.
method Parameterizing discrete distributions using sparse mappings (sparsemax and structured variants) to reduce support and enable efficient marginalization.
result Achieved good performance in various tasks with efficient and practical training.
We present an integrated approach for structure and parameter estimation in latent tree graphical models. Our overall approach follows a "divide-and-conquer" strategy that learns models over small groups of variables and iteratively merges onto a global solution. The structure learning involves combinatorial operations…
New method uses diffusion models for unsupervised combinatorial optimization.
problem Learning to sample from intractable discrete distributions without training data.
method Lifts the restriction of generative models needing exact sample likelihoods using a loss that bounds reverse KL divergence.
result Achieves new state-of-the-art results in data-free Combinatorial Optimization.
Paper proposes a fast algorithm to recover causal DAGs with latent variables.
problem Discovering causal relationships in the presence of latent variables.
method Cholesky factorization of covariance matrix with optimization for latent variables.
result The algorithm significantly outperforms previous methods in synthetic and real-world datasets.
Study provides selective inference method for latent block models.
problem Challenges in constructing a test on a block structure selected by clustering algorithms.
method Developed a selective inference method for latent block models using squared residue minimization and simulated annealing.
result Proposed tests effectively handle selective bias in block structures compared to naive tests.
Paper introduces models to discover complex structures in large hypergraphs.
problem Understanding dependency structures in complex systems represented as hypergraphs.
method Probabilistic models treating classes of similar units as nodes in a latent hypergraph, using low-rank representations.
result Improves link prediction and discovers interpretable structures in diverse real-world systems.
A new framework for structured bandits using influence diagrams and variational Thompson sampling.
problem Complex statistical dependencies in structured bandit problems.
method Influence diagram framework, variational Thompson sampling, tracking structured posterior distribution.
result Empirically evaluated algorithms perform as well as or better than existing baselines.
Developed a Particle-Gibbs sampler for Bayesian feature allocation models.
problem Intractable exact inference in Bayesian feature allocation models.
method Particle-Gibbs sampler for feature allocation matrix updates.
result PG sampler improves performance of feature allocation models.
The paper introduces combinatorial curvature and flow for polyhedral surfaces, proving rigidity and solving the Yamabe problem.
problem Discrete conformal structures on polyhedral surfaces and their rigidity.
method Parameterized combinatorial curvature, combinatorial α-Ricci flow, and flow extension through singularities.
result Existence and convergence of combinatorial α-Ricci flow for solving the Yamabe problem.
Novel theory combines combinatorial and topological elements.
problem Understanding combinatorial phenomena at the intersection of topology.
method Synthesizes combinatorial and topological approaches with a new framing concept.
result Framed combinatorial spaces exhibit better behavior than classical spaces.
The paper tackles combinatorial pure exploration with various feedback structures and proposes efficient algorithms.
problem Identifying the optimal action in a combinatorial space with limited feedback and nonlinear rewards.
method Designs polynomial-time adaptive algorithms for CPE-BL and CPE-PL, providing sample complexity analyses.
result The proposed algorithms achieve sample complexity close to lower bounds and outperform existing methods.
New combinatorial structure for hierarchically hyperbolic spaces.
problem Constructing new hierarchically hyperbolic spaces.
method Combinatorial hierarchical hyperbolicity criterion to construct and clarify HHS structures.
result HHSs admit a combinatorial structure, clarifying the application of the combinatorial HHS criterion.
The paper studies deformation of discrete conformal structures on surfaces using combinatorial curvature flows.
problem Finding piecewise constant curvature metrics on surfaces with prescribed combinatorial curvatures.
method Combinatorial curvature flows, including Ricci flow and Calabi flow, are applied to deform Glickenstein's discrete conformal structures.
result The solution of the combinatorial Ricci flow can be uniquely extended and converges exponentially fast for any initial value under certain conditions.
Topic models have emerged as fundamental tools in unsupervised machine learning. Most modern topic modeling algorithms take a probabilistic view and derive inference algorithms based on Latent Dirichlet Allocation (LDA) or its variants. In contrast, we study topic modeling as a combinatorial optimization problem, and p…
Fractional combinatorial flow improves surface conformal structures.
problem Improving discrete conformal structures on surfaces.
method Introducing a fractional combinatorial Calabi flow for discrete conformal structures on surfaces.
result Longtime existence and global convergence of the fractional combinatorial Calabi flow for various surface types.
New discrete conformal structures on surfaces with boundary, proving global rigidity and constructing hyperbolic metrics.
problem Creating new discrete conformal structures on surfaces with boundary.
method Introducing new discrete conformal structures, proving global rigidity using variational principles, and introducing combinatorial curvature flows.
result Global rigidity of new discrete conformal structures and effective algorithms for constructing hyperbolic metrics.
Study on deforming discrete conformal structures on surfaces with boundaries.
problem Deforming discrete conformal structures on surfaces with boundaries.
method Introduce combinatorial Ricci flow and combinatorial Calabi flow, establish longtime existence and global convergence of solutions.
result Effective algorithms for finding discrete hyperbolic metrics on surfaces with totally geodesic boundaries of prescribed lengths.
Study explores properties of bipartite knots.
problem None explicitly stated; focuses on properties of bipartite knots.
method Exploration of combinatorial structure.
result Rich combinatorial structure of bipartite knots.
The optimization of expensive-to-evaluate black-box functions over combinatorial structures is an ubiquitous task in machine learning, engineering and the natural sciences. The combinatorial explosion of the search space and costly evaluations pose challenges for current techniques in discrete optimization and machine …
SAMS-VAE models cellular perturbations using sparse additive mechanisms.
problem Modeling effects of diverse interventions on cells.
method Sparse Additive Mechanism Shift Variational Autoencoder (SAMS-VAE).
result SAMS-VAE identifies disentangled, perturbation-specific latent subspaces.
We propose a new family of combinatorial inference problems for graphical models. Unlike classical statistical inference where the main interest is point estimation or parameter testing, combinatorial inference aims at testing the global structure of the underlying graph. Examples include testing the graph connectivity…
SRL embeds combinatorial optimization into RL for better decision-making.
problem Challenges of standard RL in complex, structured decision-making problems.
method Structured Reinforcement Learning (SRL) with combinatorial optimization layers in actor neural network.
result SRL outperforms unstructured RL and imitation learning by up to 92% on dynamic problems.
We investigate the combinatorial analogues, in the context of normal surfaces, of taut and transversely measured (codimension 1) foliations of 3-manifolds. We establish that the existence of certain combinatorial structures, a priori weaker than the existence of the corresponding foliation, is sufficient to guarantee t…
The paper applies combinatorial Ricci flows to prove hyperbolic structures on 3-manifolds.
problem Proving the existence of hyperbolic structures on 3-manifolds with cusps.
method Combinatorial Ricci curvature flow methods to study pseudo 3-manifolds and ideal triangulations.
result The extended Ricci flow converges to a decorated hyperbolic polyhedral metric if and only if there exists a zero Ricci curvature metric.
Combinatorial definition of bordered Floer theory for torus-boundary manifolds.
problem Defining Floer homology for manifolds with torus boundaries.
method Combinatorial definition with Z coefficients. result Recovery of combinatorial Heegaard Floer homology.
Metric learning enhances combinatorial coverage metrics' ability to predict classification errors.
problem Dataset dependence of combinatorial coverage metrics in anticipating classification errors.
method Metric learning to improve latent space separation of data classes.
result Metric learning increases SDCCMs' ability to distinguish between correctly and incorrectly classified data.
New methods learn DAGs from noisy data, adapting to noise levels.
problem Inferring causal relationships from observational data with noise and confounding.
method Reformulate DAG learning as a continuous optimization problem over adjacency matrices, jointly inferring structure and noise levels.
result Improved robustness to heteroscedasticity and distribution shifts.
Polynomial-time method solves complex combinatorial semi-bandits.
problem Optimal strategies for combinatorial semi-bandits with uncorrelated Gaussian rewards.
method Proposes a polynomial-time method to solve the Graves-Lai optimization problem for various combinatorial structures.
result First known approach to implement asymptotically optimal algorithms in polynomial time for combinatorial semi-bandits.
A deep probabilistic model analyzes DNA-encoded library data for efficient screening.
problem Complex data from DNA-encoded library experiments mask underlying signals.
method Compositional deep probabilistic model of DEL data, modeling latent reactions between synthons.
result DEL-Compose model demonstrates strong performance and valuable insights.
LP-SparseMAP relaxes SparseMAP for more complex structures.
problem SparseMAP's tractable MAP inference oracle limits its applicability.
method Local polytope relaxation for factor graphs.
result LP-SparseMAP outperforms SparseMAP and Structured SVM in structured prediction tasks.
New method for mixed-variable GSA improves material design efficiency.
problem Designing materials with both quantitative and qualitative variables.
method Integrates LVGP with Sobol' analysis for mixed-variable GSA.
result Accelerates exploration of novel MOF candidates in combinatorial design spaces.
Modeling correlated mutations in cancer for personalized treatment.
problem Identifying mutations for personalized cancer therapy in heterogeneous profiles.
method Proposed correlated zero-inflated negative binomial process with mixed beta-Bernoulli and variational inference.
result Identified biologically relevant correlations between somatic mutations.
Paper shows identifiability of causal models with unobserved variables.
problem Identify latent variables in causal models with unobserved variables.
method Developed an autoencoding variational Bayes algorithm.
result Identifiability achieved with generalized faithfulness assumptions.
We study the moduli space of euclidean structures with cone points on a surface, and describe a decomposition into cells each of which corresponds to a given combinatorial type of Delaunay tessellation. We use some of the ideas to study hyperbolic structures on three-dimensional manifolds
In this paper we develop several algebraic structures on the simplicial cochains of a triangulated manifold that are analogues of objects in differential geometry. We study a cochain product and prove several statements about its convergence to the wedge product on differential forms. Also, for cochains with an inner p…
The study generalizes origamis to flat surfaces, exploring their combinatorial and geometric properties.
problem Understanding the geometric and combinatorial properties of flat surfaces.
method Developing a system of linear equations to represent flat surfaces and studying their Veech groups.
result Veech groups of certain flat surfaces are included under a specific covering relation.
The paper proves a combinatorial Ricci flow converges to hyperbolic structures on certain 3-manifolds.
problem Proving convergence of combinatorial Ricci flow to hyperbolic structures.
method Combinatorial Ricci flow on closed pseudo 3-manifolds with specific edge valences.
result Existence and uniqueness of a complete hyperbolic metric with totally geodesic boundary.
We introduce Deep Reasoning Networks (DRNets), an end-to-end framework that combines deep learning with reasoning for solving complex tasks, typically in an unsupervised or weakly-supervised setting. DRNets exploit problem structure and prior knowledge by tightly combining logic and constraint reasoning with stochastic…
Permutations and matchings are core building blocks in a variety of latent variable models, as they allow us to align, canonicalize, and sort data. Learning in such models is difficult, however, because exact marginalization over these combinatorial objects is intractable. In response, this paper introduces a collectio…
This paper gives a combinatorial description of spin and spin^c-structures on triangulated PL-manifolds of arbitrary dimension. These formulations of spin and spin^c-structures are established primarily for the purpose of aiding in computations. The novelty of the approach is we rely heavily on the naturality of binary…
New combinatorial structures for Teichmüller spaces with Thurston's metric are explored.
problem Understanding the combinatorial structures of Teichmüller spaces with Thurston's metric.
method Analyzing the unit tangent and cotangent spheres of Teichmüller space, proving formulas for dimensions and codimensions of faces.
result The combinatorial structure of unit spheres in Teichmüller spaces is independent of the underlying point and is isomorphic to the extended mapping class group.
Quantizes latent space to improve disentanglement in models.
problem Learning disentangled representations from unlabeled data.
method Quantizes latent space into discrete code vectors with a learnable scalar codebook and applies high weight decay regularization.
result Quantized-latent autoencoder (QLAE) outperforms prior work in disentanglement without sacrificing data reconstruction.
Study inert and ambiguous classes in modular group using combinatorial methods.
problem Counting inert and ambiguous conjugacy classes in modular group.
method Purely combinatorial approach using word length in free product representation.
result Exact counting formulas and asymptotic growth rates for inert and ambiguous classes.