2024
Geometric scattering on measure spaces
Chew J, Hirn M, Krishnaswamy S, Needell D, Perlmutter M, Steach H, Viswanath S, Wu H. Geometric scattering on measure spaces. Applied And Computational Harmonic Analysis 2024, 70: 101635. DOI: 10.1016/j.acha.2024.101635.Peer-Reviewed Original ResearchConvolutional neural networkGeometric deep learningDeep learningNeural networkSuccess of convolutional neural networksModel of convolutional neural networkMeasure spaceScattering transformData-driven graphsInvariance propertiesRiemannian manifoldsNon-Euclidean structureUndirected graphWavelet-based transformCompact Riemannian manifoldsData structuresRate of convergenceSpherical imagesNetwork stabilityHigh-dimensional single-cell dataData setsDirected graphDiffusion-mapsSigned graphGraphLearnable Filters for Geometric Scattering Modules
Tong A, Wenkel F, Bhaskar D, Macdonald K, Grady J, Perlmutter M, Krishnaswamy S, Wolf G. Learnable Filters for Geometric Scattering Modules. IEEE Transactions On Signal Processing 2024, 72: 2939-2952. DOI: 10.1109/tsp.2024.3378001.Peer-Reviewed Original ResearchGraph neural networksGraph classification benchmarksEncode graph structureData exploration tasksGeometric scattering transformGraph wavelet filtersClassification benchmarksLearned representationsLearnable filtersLearning parametersGraph structureNeural networkExploration tasksWavelet filtersBand-pass featureBiochemical domainNetworkAdaptive tuningWaveletGraphPredictive performanceScattering modulationScattering transformMathematical propertiesFilterAssessing Neural Network Representations During Training Using Noise-Resilient Diffusion Spectral Entropy
Liao D, Liu C, Christensen B, Tong A, Huguet G, Wolf G, Nickel M, Adelstein I, Krishnaswamy S. Assessing Neural Network Representations During Training Using Noise-Resilient Diffusion Spectral Entropy. 2024, 00: 1-6. DOI: 10.1109/ciss59072.2024.10480166.Peer-Reviewed Original ResearchMutual information neural estimatorMutual informationHigh-dimensional simulation dataHigh-dimensional dataNeural network representationSpectral entropyCIFAR-10Information-theoretic measuresClass labelsSTL-10Classification networkNeural representationSelf-supervisionSupervised learningIntrinsic dimensionalityClassification accuracyNeural networkAmbient dimensionNoise-resilientNeural estimatorNetwork initializationData geometryNetwork representationOverfittingNetwork
2022
Exploring the Geometry and Topology of Neural Network Loss Landscapes
Horoi S, Huang J, Rieck B, Lajoie G, Wolf G, Krishnaswamy S. Exploring the Geometry and Topology of Neural Network Loss Landscapes. Lecture Notes In Computer Science 2022, 13205: 171-184. DOI: 10.1007/978-3-031-01333-1_14.Peer-Reviewed Original ResearchLoss landscapeNon-linear dimensionality reductionLandscape geometryGeneralization performanceLocal minimaDimensionality reduction techniquesLinear natureReduction techniquesDimensionality reductionGeometrySuch visualization methodsCaptures featuresJumpNetwork's abilityNeural networkTrajectoriesRecent workTopologyNetwork
2019
Modeling Global Dynamics from Local Snapshots with Deep Generative Neural Networks
Gigante S, van Dijk D, Moon K, Strzalkowski A, Wolf G, Krishnaswamy S. Modeling Global Dynamics from Local Snapshots with Deep Generative Neural Networks. 2019, 00: 1-4. DOI: 10.1109/sampta45681.2019.9030978.Peer-Reviewed Original ResearchStochastic dynamic systemsDeep generative neural networksProbability distributionDynamic systemsMarkov modelGlobal dynamicsLocal snapshotGenerative neural networksKalman filterSnapshot dataNeural networkNeural network frameworkRecurrent neural networkSuch systemsNext stateModeling networkNetwork frameworkDynamicsShallow modelsLocal transitionsHypothetical trajectoryModelBiological systemsNatural sciencesLongitudinal measurements