Multifaceted Feature Visualization: Uncovering the Different Types of Features Learned By Each Neuron in Deep Neural Networks
arXiv:1602.03616
Abstract
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.
23 pages (including SI), 24 figures
References in corpus (4)
Cited by in corpus (65)
- Methods for Interpreting and Understanding Deep Neural Networks
- A Survey on Explainable Artificial Intelligence (XAI): Towards Medical XAI
- Explaining Deep Neural Networks and Beyond: A Review of Methods and Applications
- Explainable Artificial Intelligence: Understanding, Visualizing and Interpreting Deep Learning Models
- One Explanation Does Not Fit All: A Toolkit and Taxonomy of AI Explainability Techniques
- Synthesizing the preferred inputs for neurons in neural networks via deep generator networks
- Do Convolutional Neural Networks Learn Class Hierarchy?
- DeepAID: Interpreting and Improving Deep Learning-based Anomaly Detection in Security Applications
- Explainable Diabetic Retinopathy Detection and Retinal Image Generation
- Neuron Shapley: Discovering the Responsible Neurons
- Explainability Techniques for Graph Convolutional Networks
- Analysis and Optimization of Convolutional Neural Network Architectures
- Beyond saliency: understanding convolutional neural networks from saliency prediction on layer-wise relevance propagation
- How convolutional neural network see the world - A survey of convolutional neural network visualization methods
- Stacked Generative Adversarial Networks
- Understanding trained CNNs by indexing neuron selectivity
- Generative Counterfactual Introspection for Explainable Deep Learning
- Towards falsifiable interpretability research
- Techniques for Interpretable Machine Learning
- Scientific Inference With Interpretable Machine Learning: Analyzing Models to Learn About Real-World Phenomena
- CNNComparator: Comparative Analytics of Convolutional Neural Networks
- Interpreting the Predictions of Complex ML Models by Layer-wise Relevance Propagation
- Precomputed Real-Time Texture Synthesis with Markovian Generative Adversarial Networks
- Improving Interpretability of Deep Neural Networks with Semantic Information
- TIP: Typifying the Interpretability of Procedures
- Representation of linguistic form and function in recurrent neural networks
- Comparing Neural and Attractiveness-based Visual Features for Artwork Recommendation
- Calibrating Healthcare AI: Towards Reliable and Interpretable Deep Predictive Models
- Explaining Bayesian Neural Networks
- Generalizing multistain immunohistochemistry tissue segmentation using one-shot color deconvolution deep neural networks
- Visualization of Convolutional Neural Networks for Monocular Depth Estimation
- Finding and Visualizing Weaknesses of Deep Reinforcement Learning Agents
- Towards Better Analysis of Deep Convolutional Neural Networks
- Enhancing the Extraction of Interpretable Information for Ischemic Stroke Imaging from Deep Neural Networks
- Semantics for Global and Local Interpretation of Deep Neural Networks
- Learning from Higher-Layer Feature Visualizations
- Examining the Benefits of Capsule Neural Networks
- GAN-based Generation and Automatic Selection of Explanations for Neural Networks
- "I know it when I see it". Visualization and Intuitive Interpretability
- A Survey on Understanding, Visualizations, and Explanation of Deep Neural Networks
- Exemplary Natural Images Explain CNN Activations Better than State-of-the-Art Feature Visualization
- Improving image generative models with human interactions
- Revisiting Edge Detection in Convolutional Neural Networks
- TopoAct: Visually Exploring the Shape of Activations in Deep Learning
- Sampling the "Inverse Set" of a Neuron: An Approach to Understanding Neural Nets
- Modeling Latent Attention Within Neural Networks
- Optimising the Input Image to Improve Visual Relationship Detection
- Every Filter Extracts A Specific Texture In Convolutional Neural Networks
- Illuminated Decision Trees with Lucid
- Inverting and Understanding Object Detectors
- Demystifying Brain Tumour Segmentation Networks: Interpretability and Uncertainty Analysis
- Out of the Black Box: Properties of deep neural networks and their applications
- Understanding Convolutional Neural Networks with A Mathematical Model
- Explainability-Aware One Point Attack for Point Cloud Neural Networks
- Efficient Modelling Across Time of Human Actions and Interactions
- Learning Realistic Patterns from Unrealistic Stimuli: Generalization and Data Anonymization
- Explainable Adversarial Attacks in Deep Neural Networks Using Activation Profiles
- Logic and the -Simplicial Transformer
- Are there any 'object detectors' in the hidden layers of CNNs trained to identify objects or scenes?
- Diagnostic Visualization for Deep Neural Networks Using Stochastic Gradient Langevin Dynamics
- Visualizing Classification Structure of Large-Scale Classifiers
- Explainability via Responsibility
- Low-Cost Transfer Learning of Face Tasks
- Hierarchical Neural Representation of Dreamed Objects Revealed by Brain Decoding with Deep Neural Network Features
- IFBiD: Inference-Free Bias Detection