Multifaceted Feature Visualization: Uncovering the Different Types of Features Learned By Each Neuron in Deep Neural Networks

We can better understand deep neural networks by identifying which features each of their neurons have learned to detect. To do so, researchers have created Deep Visualization techniques including activation maximization, which synthetically generates inputs (e.g. images) that maximally activate each neuron. A limitation of current techniques is that they assume each neuron detects only one type of feature, but we know that neurons can be multifaceted, in that they fire in response to many different types of features: for example, a grocery store class neuron must activate either for rows of produce or for a storefront. Previous activation maximization techniques constructed images without regard for the multiple different facets of a neuron, creating inappropriate mixes of colors, parts of objects, scales, orientations, etc. Here, we introduce an algorithm that explicitly uncovers the multiple facets of each neuron by producing a synthetic visualization of each of the types of images that activate a neuron. We also introduce regularization methods that produce state-of-the-art results in terms of the interpretability of images obtained by activation maximization. By separately synthesizing each type of image a neuron fires in response to, the visualizations have more appropriate colors and coherent global structure. Multifaceted feature visualization thus provides a clearer and more comprehensive description of the role of each neuron.

Unsupervised Transductive Domain Adaptation

Homonym Population Protocols

On maximal tail probability of sums of nonnegative, independent and identically distributed random variables

Ising Spin Glasses and Renormalization Group Theory: the Binder cumulant

Enumeration of colored Dyck paths via partial Bell polynomials

Knowledge Transfer with Medical Language Embeddings

Learning Privately from Multiparty Data

The independence number of non-uniform uncrowded hypergraphs and an anti-Ramsey type result

High Dimensional Inference with Random Maximum A-Posteriori Perturbations

Effective Sample Size for Importance Sampling based on the discrepancy measures

A knockoff filter for high-dimensional selective inference

Guessing Numbers of Odd Cycles

The knockoff filter for FDR control in group-sparse and multitask regression

Spectral characterization of matchings in graphs

Reversible Communicating Processes

Distributed Programming via Safe Closure Passing

Data-Driven Online Decision Making with Costly Information Acquisition

Variations of the Similarity Function of TextRank for Automated Summarization

Attentive Pooling Networks

Asymptotics of the number of involutions in finite classical groups

Optimality of Belief Propagation for Crowdsourced Classification

Singular behavior of the leading Lyapunov exponent of a product of random $2 \times 2$ matrices

High performance Python for direct numerical simulations of turbulent flows

Parallel Vertex Approximate Gradient discretization of hybrid dimensional Darcy flow and transport in discrete fracture networks

On the Difficulty of Selecting Ising Models with Approximate Recovery

Bayesian Filtering of Smooth Signals: Application to Altimetry

Symmetries and martingales in a stochastic model for the Navier-Stokes equation

A randomized maximum a posterior method for posterior sampling of high dimensional nonlinear Bayesian inverse problems

On the emergence of syntactic structures: quantifying and modelling duality of patterning

Extremal results for odd cycles in sparse pseudorandom graphs

Online Low-Rank Subspace Learning from Incomplete Data: A Bayesian View

Package equivalence in complex software network

A Universal Approximation Theorem for Mixture of Experts Models

Medical Concept Representation Learning from Electronic Health Records and its Application on Heart Failure Prediction

Maximum Likelihood Estimation of Triangular and Polygonal Distributions

Linear Mixed Models with Marginally Symmetric Nonparametric Random Effects

The variation of the Randic index with regard to minimum and maximum degree

A Distributed $(2+ε)$-Approximation for Vertex Cover in $O(\logΔ/ε\log\logΔ)$ Rounds

Fast Distributed Algorithms for Testing Graph Properties

Sequences with small correlation

Resampling-based inference methods for comparing two coefficient alpha

The List Distinguishing Number of Kneser Graphs

Looking for a Needle in a Haystack? Look Elsewhere! A statistical comparison of approximate global p-values

Integrative Dynamic Reconfiguration in a Parallel Stream Processing Engine

A sequence of triangle-free pseudorandom graphs

Planar growth generates scale free networks

Network of Bandits

Optimal designs for regression models with autoregressive errors structure

Cellular automaton for chimera states

On 2K2-free graphs – Structural and Combinatorial View

Bayesian Sparsity for Intractable Distributions

Semi-supervised Learning with Explicit Relationship Regularization

Optimal quantitative estimates in stochastic homogenization for elliptic equations in nondivergence form

Enabling Basic Normative HRI in a Cognitive Robotic Architecture

Neural Network Support Vector Detection via a Soft-Label, Hybrid K-Means Classifier

Community Recovery in Graphs with Locality

Intertwinings and Generalized Brascamp-Lieb Inequalities

Central limit theorem and law of the iterated logarithm for the linear random walk on the torus

A study of large fringe and non-fringe subtrees in conditional Galton-Watson trees

A nonlinear Kolmogorov equation for stochastic functional delay differential equations with jumps

A Statistical Framework for Single Subject Design with an Application in Post-stroke Rehabilitation

Statistical Foundation of Spectral Graph Theory

Parallel Shortest-Paths Using Radius Stepping

CLE Percolations