Search Results for author:
Found 88 papers, 24 papers with code
Disentangling Regional Primitives for Image Generation
We find that the feature component can be represented as an OR relationship between the demands for generating different image regions, which is encoded by the neural network.
Layerwise Change of Knowledge in Neural Networks
This paper aims to explain how a deep neural network (DNN) gradually extracts new knowledge and forgets noisy features through layers in forward propagation.
Towards the Dynamics of a DNN Learning Symbolic Interactions
This study proves the two-phase dynamics of a deep neural network (DNN) learning interactions.
Quantifying In-Context Reasoning Effects and Memorization Effects in LLMs
In this study, we propose an axiomatic system to define and quantify the precise memorization and in-context reasoning effects used by the large language model (LLM) for language generation.
Two-Phase Dynamics of Interactions Explains the Starting Point of a DNN Learning Over-Fitted Features
This paper investigates the dynamics of a deep neural network (DNN) learning interactions.
Identifying Semantic Induction Heads to Understand In-Context Learning
Although large language models (LLMs) have demonstrated remarkable performance, the lack of transparency in their inference logic raises concerns about their trustworthiness.
Defining and Extracting generalizable interaction primitives from DNNs
Faithfully summarizing the knowledge encoded by a deep neural network (DNN) into a few symbolic primitive patterns without losing much information represents a core challenge in explainable AI.
Explaining How a Neural Network Play the Go Game and Let People Learn
The AI model has surpassed human players in the game of Go, and it is widely believed that the AI model has encoded new knowledge about the Go game beyond human players.
Towards Attributions of Input Variables in a Coalition
This paper aims to develop a new attribution method to explain the conflict between individual variables' attributions and their coalition's attribution from a fully new perspective.
Where We Have Arrived in Proving the Emergence of Sparse Symbolic Concepts in AI Models
These conditions are quite common, and we prove that under these conditions, the DNN will only encode a relatively small number of sparse interactions between input variables.
Technical Note: Defining and Quantifying AND-OR Interactions for Faithful and Concise Explanation of DNNs
For faithfulness, we prove the uniqueness of the AND (OR) interaction in quantifying the effect of the AND (OR) relationship between input variables.
HarsanyiNet: Computing Accurate Shapley Values in a Single Forward Propagation
The HarsanyiNet is designed on the theoretical foundation that the Shapley value can be reformulated as the redistribution of Harsanyi interactions encoded by the network.
Can the Inference Logic of Large Language Models be Disentangled into Symbolic Concepts?
In this paper, we explain the inference logic of large language models (LLMs) as a set of symbolic concepts.
Understanding and Unifying Fourteen Attribution Methods with Taylor Interactions
Various attribution methods have been developed to explain deep neural networks (DNNs) by inferring the attribution/importance/contribution score of each input variable to the final output.
Explaining Generalization Power of a DNN Using Interactive Concepts
Although there is no universally accepted definition of the concepts encoded by a DNN, the sparsity of interactions in a DNN has been proved, i. e., the output score of a DNN can be well explained by a small number of interactions between input variables.
Does a Neural Network Really Encode Symbolic Concepts?
Recently, a series of studies have tried to extract interactions between input variables modeled by a DNN and define such interactions as concepts encoded by the DNN.
Bayesian Neural Networks Avoid Encoding Complex and Perturbation-Sensitive Concepts
In this paper, we focus on mean-field variational Bayesian Neural Networks (BNNs) and explore the representation capacity of such BNNs by investigating which types of concepts are less likely to be encoded by the BNN.
Defects of Convolutional Decoder Networks in Frequency Representation
In this paper, we prove the representation defects of a cascaded convolutional decoder network, considering the capacity of representing different frequency components of an input sample.
Quantifying the Knowledge in a DNN to Explain Knowledge Distillation for Classification
This paper provides a new perspective to explain the success of knowledge distillation, i. e., quantifying knowledge points encoded in intermediate layers of a DNN for classification, based on the information theory.
Proving Common Mechanisms Shared by Twelve Methods of Boosting Adversarial Transferability
This paper summarizes the common mechanism shared by twelve previous transferability-boosting methods in a unified view, i. e., these methods all reduce game-theoretic interactions between regional adversarial perturbations.
Batch Normalization Is Blind to the First and Second Derivatives of the Loss
In this paper, we prove the effects of the BN operation on the back-propagation of the first and second derivatives of the loss.
Why Adversarial Training of ReLU Networks Is Difficult?
We also prove that adversarial training tends to strengthen the influence of unconfident input samples with large gradient norms in an exponential manner.
RASAT: Integrating Relational Structures into Pretrained Seq2Seq Model for Text-to-SQL
Our model can incorporate almost all types of existing relations in the literature, and in addition, we propose introducing co-reference relations for the multi-turn scenario.
Ranked #1 on
Dialogue State Tracking
on CoSQL
Towards Theoretical Analysis of Transformation Complexity of ReLU DNNs
Based on the proposed metrics, we analyze two typical phenomena of the change of the transformation complexity during the training process, and explore the ceiling of a DNN's complexity.
A Roadmap for Big Model
With the rapid development of deep learning, training Big Models (BMs) for multiple downstream tasks becomes a popular paradigm.
Trap of Feature Diversity in the Learning of MLPs
In this paper, we focus on a typical two-phase phenomenon in the learning of multi-layer perceptrons (MLPs), and we aim to explain the reason for the decrease of feature diversity in the first phase.
Towards a Unified Game-Theoretic View of Adversarial Perturbations and Robustness
This paper provides a unified view to explain different adversarial attacks and defense methods, i. e. the view of multi-order interactions between input variables of DNNs.
Discovering and Explaining the Representation Bottleneck of DNNs
This paper explores the bottleneck of feature representations of deep neural networks (DNNs), from the perspective of the complexity of interactions between input variables encoded in DNNs.
Defining and Quantifying the Emergence of Sparse Concepts in DNNs
This paper aims to illustrate the concept-emerging phenomenon in a trained DNN.
Visualizing the Emergence of Intermediate Visual Patterns in DNNs
This paper proposes a method to visualize the discrimination power of intermediate-layer visual patterns encoded by a DNN.
Interpreting Representation Quality of DNNs for 3D Point Cloud Processing
In this paper, we evaluate the quality of knowledge representations encoded in deep neural networks (DNNs) for 3D point cloud processing.
A Unified Game-Theoretic Interpretation of Adversarial Robustness
This paper provides a unified view to explain different adversarial attacks and defense methods, \emph{i. e.} the view of multi-order interactions between input variables of DNNs.
A HYPOTHESIS FOR THE COGNITIVE DIFFICULTY OF IMAGES
This paper proposes a hypothesis to analyze the underlying reason for the cognitive difficulty of an image from two perspectives, i. e. a cognitive image usually makes a DNN strongly activated by cognitive concepts; discarding massive non-cognitive concepts may also help the DNN focus on cognitive concepts.
Dissecting Local Properties of Adversarial Examples
Adversarial examples have attracted significant attention over the years, yet a sufficient understanding is in lack, especially when analyzing their performances in combination with adversarial training.
Towards a Game-Theoretic View of Baseline Values in the Shapley Value
In the computation of Shapley values, people usually set an input variable to its baseline value to represent the absence of this variable.
Rapid detection and recognition of whole brain activity in a freely behaving Caenorhabditis elegans
Second, 2-dimensional neuronal regions are fused into 3-dimensional neuron entities.
Interpreting Attributions and Interactions of Adversarial Attacks
This paper aims to explain adversarial attacks in terms of how adversarial perturbations contribute to the attacking task.
A Hypothesis for the Aesthetic Appreciation in Neural Networks
This paper proposes a hypothesis for the aesthetic appreciation that aesthetic images make a neural network strengthen salient concepts and discard inessential concepts.
Proceedings of ICML 2021 Workshop on Theoretic Foundation, Criticism, and Application Trend of Explainable AI
This workshop pays a special interest in theoretic foundations, limitations, and new application trends in the scope of XAI.
Interpretable Compositional Convolutional Neural Networks
The reasonable definition of semantic interpretability presents the core challenge in explainable AI.
A Game-Theoretic Taxonomy of Visual Concepts in DNNs
In this paper, we rethink how a DNN encodes visual concepts of different complexities from a new perspective, i. e. the game-theoretic multi-order interactions between pixels in an image.
Can We Faithfully Represent Masked States to Compute Shapley Values on a DNN?
Masking some input variables of a deep neural network (DNN) and computing output changes on the masked input sample represent a typical way to compute attributions of input variables in the sample.
A Unified Game-Theoretic Interpretation of Adversarial Robustness
This paper provides a unified view to explain different adversarial attacks and defense methods, i. e. the view of multi-order interactions between input variables of DNNs.
Towards A Unified Understanding and Improving of Adversarial Transferability
We discover and prove the negative correlation between the adversarial transferability and the interaction inside adversarial perturbations.
Towards Understanding and Improving Dropout in Game Theory
Experimental results on various DNNs and datasets have shown that the interaction loss can effectively improve the utility of dropout and boost the performance of DNNs.
Disentanglement, Visualization and Analysis of Complex Features in DNNs
This paper aims to define, visualize, and analyze the feature complexity that is learned by a DNN.
Understanding, Analyzing, and Optimizing the Complexity of Deep Models
Based on the proposed metrics, we analyze two typical phenomena of the change of the transformation complexity during the training process, and explore the ceiling of a DNN’s complexity.
Technical Note: Game-Theoretic Interactions of Different Orders
In this study, we define interaction components of different orders between two input variables based on game theory.
Interpreting Multivariate Shapley Interactions in DNNs
In this paper, we define and quantify the significance of interactions among multiple input variables of the DNN.
A Unified Approach to Interpreting and Boosting Adversarial Transferability
We discover and prove the negative correlation between the adversarial transferability and the interaction inside adversarial perturbations.
Interpreting and Boosting Dropout from a Game-Theoretic View
This paper aims to understand and improve the utility of the dropout operation from the perspective of game-theoretic interactions.
Achieving Adversarial Robustness via Sparsity
Through experiments on a variety of adversarial pruning methods, we find that weights sparsity will not hurt but improve robustness, where both weights inheritance from the lottery ticket and adversarial training improve model robustness in network pruning.
Building Interpretable Interaction Trees for Deep NLP Models
This paper proposes a method to disentangle and quantify interactions among words that are encoded inside a DNN for natural language processing.
Interpreting and Disentangling Feature Components of Various Complexity from DNNs
This paper aims to define, quantify, and analyze the feature complexity that is learned by a DNN.
Rotation-Equivariant Neural Networks for Privacy Protection
Compared to the traditional neural network, the RENN uses d-ary vectors/tensors as features, in which each element is a d-ary number.
Deep Quaternion Features for Privacy Protection
We propose a method to revise the neural network to construct the quaternion-valued neural network (QNN), in order to prevent intermediate-layer features from leaking input information.
Explaining Knowledge Distillation by Quantifying the Knowledge
Whereas, in the scenario of learning from raw data, the DNN learns visual concepts sequentially.
Learning to Prevent Leakage: Privacy-Preserving Inference in the Mobile Cloud
Powered by machine learning services in the cloud, numerous learning-driven mobile applications are gaining popularity in the market.
Towards a Unified Evaluation of Explanation Methods without Ground Truth
This paper proposes a set of criteria to evaluate the objectiveness of explanation methods of neural networks, which is crucial for the development of explainable AI, but it also presents significant challenges.
Verifiability and Predictability: Interpreting Utilities of Network Architectures for Point Cloud Processing
In this paper, we diagnose deep neural networks for 3D point cloud processing to explore utilities of different intermediate-layer network architectures.
3D-Rotation-Equivariant Quaternion Neural Networks
This paper proposes a set of rules to revise various neural networks for 3D point cloud processing to rotation-equivariant quaternion neural networks (REQNNs).
Knowledge Consistency between Neural Networks and Beyond
As a generic tool, our method can be broadly used for different applications.
Quantification and Analysis of Layer-wise and Pixel-wise Information Discarding
We define two types of entropy-based metrics, i. e. (1) the discarding of pixel-wise information used in the forward propagation, and (2) the uncertainty of the input reconstruction, to measure input information contained by a specific layer from two perspectives.
Interpretable Complex-Valued Neural Networks for Privacy Protection
Previous studies have found that an adversary attacker can often infer unintended input information from intermediate-layer features.
Proceedings of AAAI 2019 Workshop on Network Interpretability for Deep Learning
This is the Proceedings of AAAI 2019 Workshop on Network Interpretability for Deep Learning
Network Transplanting (extended abstract)
This paper focuses on a new task, i. e., transplanting a category-and-task-specific neural network to a generic, modular network without strong supervision.
Unsupervised Learning of Neural Networks to Explain Neural Networks (extended abstract)
This paper presents an unsupervised method to learn a neural network, namely an explainer, to interpret a pre-trained convolutional neural network (CNN), i. e., the explainer uses interpretable visual concepts to explain features in middle conv-layers of a CNN.
Explaining AlphaGo: Interpreting Contextual Effects in Neural Networks
In this paper, we propose to disentangle and interpret contextual effects that are encoded in a pre-trained deep neural network.
Interpretable CNNs for Object Classification
This paper proposes a generic method to learn interpretable convolutional filters in a deep convolutional neural network (CNN) for object classification, where each interpretable filter encodes features of a specific object part.
Explanatory Graphs for CNNs
This paper introduces a graphical model, namely an explanatory graph, which reveals the knowledge hierarchy hidden inside conv-layers of a pre-trained CNN.
Explaining Neural Networks Semantically and Quantitatively
This paper presents a method to explain the knowledge encoded in a convolutional neural network (CNN) quantitatively and semantically.
Mining Interpretable AOG Representations from Convolutional Networks via Active Question Answering
The AOG associates each object part with certain neural units in feature maps of conv-layers.
Unsupervised Learning of Neural Networks to Explain Neural Networks
Given feature maps of a certain conv-layer of the CNN, the explainer performs like an auto-encoder, which first disentangles the feature maps into object-part features and then inverts object-part features back to features of higher conv-layers of the CNN.
Network Transplanting
This paper focuses on a new task, i. e., transplanting a category-and-task-specific neural network to a generic, modular network without strong supervision.
Visual Interpretability for Deep Learning: a Survey
This paper reviews recent studies in understanding neural-network representations and learning neural networks with interpretable/disentangled middle-layer representations.
Interpreting CNNs via Decision Trees
We propose to learn a decision tree, which clarifies the specific reason for each prediction made by the CNN at the semantic level.
Examining CNN Representations with respect to Dataset Bias
We aim to discover representation flaws caused by potential dataset bias.
Interpretable Convolutional Neural Networks
Instead, the interpretable CNN automatically assigns each filter in a high conv-layer with an object part during the learning process.
Ranked #1 on
single catogory classification
on ILSVRC Part
Mining Deep And-Or Object Structures via Cost-Sensitive Question-Answer-Based Active Annotations
The loss is defined for nodes in all layers of the AOG, including the generative loss (measuring the likelihood of the images) and the discriminative loss (measuring the fitness to human answers).
Visual Graph Mining
In this study, we formulate the concept of "mining maximal-size frequent subgraphs" in the challenging domain of visual data (images and videos).
Interpreting CNN Knowledge via an Explanatory Graph
Considering that each filter in a conv-layer of a pre-trained CNN usually represents a mixture of object parts, we propose a simple yet efficient method to automatically disentangles different part patterns from each filter, and construct an explanatory graph.
Interactively Transferring CNN Patterns for Part Localization
In the scenario of one/multi-shot learning, conventional end-to-end learning strategies without sufficient supervision are usually not powerful enough to learn correct patterns from noisy signals.
Mining Object Parts from CNNs via Active Question-Answering
We use an active human-computer communication to incrementally grow such an AOG on the pre-trained CNN as follows.
Growing Interpretable Part Graphs on ConvNets via Multi-Shot Learning
This paper proposes a learning strategy that extracts object-part concepts from a pre-trained convolutional neural network (CNN), in an attempt to 1) explore explicit semantics hidden in CNN units and 2) gradually grow a semantically interpretable graphical model on the pre-trained CNN for hierarchical object understanding.
Mining And-Or Graphs for Graph Matching and Object Discovery
This paper reformulates the theory of graph mining on the technical basis of graph matching, and extends its scope of applications to computer vision.
When 3D Reconstruction Meets Ubiquitous RGB-D Images
3D reconstruction from a single image is a classical problem in computer vision.
Attributed Graph Mining and Matching: An Attempt to Define and Extract Soft Attributed Patterns
Graph matching and graph mining are two typical areas in artificial intelligence.
Category Modeling from Just a Single Labeling: Use Depth Information to Guide the Learning of 2D Models
We design a graphical model that uses object edges to represent object structures, and this paper aims to incrementally learn this category model from one labeled object and a number of casually captured scenes.
