Visualizing the Hidden Activity of Artificial Neural Networks

Publikation: Beiträge in ZeitschriftenZeitschriftenaufsätzeForschungbegutachtet

Authors

In machine learning, pattern classification assigns high-dimensional vectors (observations) to classes based on generalization from examples. Artificial neural networks currently achieve state-of-the-art results in this task. Although such networks are typically used as black-boxes, they are also widely believed to learn (high-dimensional) higher-level representations of the original observations. In this paper, we propose using dimensionality reduction for two tasks: visualizing the relationships between learned representations of observations, and visualizing the relationships between artificial neurons. Through experiments conducted in three traditional image classification benchmark datasets, we show how visualization can provide highly valuable feedback for network designers. For instance, our discoveries in one of these datasets (SVHN) include the presence of interpretable clusters of learned representations, and the partitioning of artificial neurons into groups with apparently related discriminative roles.

OriginalspracheEnglisch
Aufsatznummer7539329
ZeitschriftIEEE Transactions on Visualization and Computer Graphics
Jahrgang23
Ausgabenummer1
Seiten (von - bis)101-110
Anzahl der Seiten10
ISSN1077-2626
DOIs
PublikationsstatusErschienen - 01.2017
Extern publiziertJa

DOI

Zuletzt angesehen

Projekte

  1. XR2ESILIENCE

Publikationen

  1. Q-Adaptive Control of the nonlinear dynamics of the cantilever-sample system of an Atomic Force Microscope
  2. How, when and why do negotiators use reference points?
  3. An observer for sensorless variable valve control in camless internal combustion engines
  4. Using conditional inference trees and random forests to predict the bioaccumulation potential of organic chemicals
  5. Visualization of the Plasma Frequency by means of a Particle Simulation using a Normalized Periodic Model
  6. Assembly Modes of General Planar 3-RPR Parallel Mechanisms when Using the Linear Actuators’ Orientations
  7. Use of Machine-Learning Algorithms Based on Text, Audio and Video Data in the Prediction of Anxiety and Post-Traumatic Stress in General and Clinical Populations
  8. Hypertext
  9. Special Issue in Acquisitional Pragmatics in Foreign Language Learning
  10. Distributed robust Gaussian Process regression
  11. Assessment of cognitive load in multimedia learning using dual-task methodology
  12. Diffusion patterns in small vs. large capital markets-the case of value-based management
  13. A MODEL FOR QUANTIFICATION OF SOFTWARE COMPLEXITY
  14. Mathematical relation between extended connectivity and eigenvector coefficients.
  15. Artificial Intelligence in Foreign Language Learning and Teaching
  16. Assembly Theory for Restoring Ecosystem Structure and Functioning
  17. A tutorial introduction to adaptive fractal analysis
  18. Early Detection of Faillure in Conveyor Chain Systems by Wireless Sensor Node
  19. Temperature control in Peltier cells comparing sliding mode control and PID controllers
  20. Assessment of university students’ understanding of abstract binary operations
  21. An Adaptive and Optimized Switching Observer for Sensorless Control of an Electromagnetic Valve Actuator in Camless Internal Combustion Engines
  22. What´s in a net? or: The end of the average
  23. Obstacle Coordinates Transformation from TVS Body-Frame to AGV Navigation-Frame
  24. Top-down contingent feature-specific orienting with and without awareness of the visual input
  25. A Sliding Mode Control with a Bang-Bang Observer for Detection of Particle Pollution
  26. Kalman Filter for Predictive Maintenance and Anomaly Detection
  27. Applications of the Simultaneous Modular Approach in the Field of Material Flow Analysis
  28. Active learning for network intrusion detection
  29. Explaining and controlling for the psychometric properties of computer-generated figural matrix items