Relevance-Based Feature Masking: Improving Neural Network Based Whale Classification Through Explainable Artificial Intelligence

Schiller, Dominik; Huber, Tobias B.; Lingenfelser, Florian; Dietz, Michael; Seiderer, Andreas; André, Elisabeth

doi:10.21437/interspeech.2019-2707

Cited by 12 publications

(9 citation statements)

References 18 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Initially, we intended to use the negative masking algorithm to zero out relevant information and therefore train a network that focuses on different areas than the original Trainer network. This newly created network could then be fused on the decision level with the explanatory network to improve results-an approach that has been successfully used before in a different domain (Schiller et al, 2019). Yet we found that the negative masking approach performed better than the mean masking method.…”

Section: Resultsmentioning

confidence: 99%

“…In this section we present a generic method to assess the relevance of specific areas in the input and integrate this knowledge into a new model during training. Based on previous work from Schiller et al (2019) our approach utilizes saliency maps generated by XAI algorithms like deep Taylor decomposition. These maps aim to identify the parts of the input that were relevant for a specific decision of a neural network.…”

Section: Overview Of Our Approachmentioning

confidence: 99%

“…To this end we implemented and tested three different masking algorithms that zero out parts of an input picture based on the saliency map, for this picture, of the Trainer network. Two of those algorithms build up on previous work in Schiller et al (2019) and the third is inspired by the human perception we investigated in section 5.3. Each of those algorithms takes an image I and generates a binary mask M with same shape as the input image.…”

Section: Masking the Inputmentioning

confidence: 99%

See 2 more Smart Citations

Relevance-Based Data Masking: A Model-Agnostic Transfer Learning Approach for Facial Expression Recognition

Schiller

Huber

Dietz

et al. 2020

Front. Comput. Sci.

Self Cite

View full text Add to dashboard Cite

Deep learning approaches are now a popular choice in the field of automatic emotion recognition (AER) across various modalities. Due to the high costs of manually labeling human emotions however, the amount of available training data is relatively scarce in comparison to other tasks. To facilitate the learning process and reduce the necessary amount of training-data, modern approaches therefore often rely on leveraging knowledge from models that have already been trained on related tasks where data is available abundantly. In this work we introduce a novel approach to transfer learning, which addresses two shortcomings of traditional methods: The (partial) inheritance of the original models structure and the restriction to other neural network models as an input source. To this end we identify the parts in the input that have been relevant for the decision of the model we want to transfer knowledge from, and directly encode those relevant regions in the data on which we train our new model. To validate our approach we performed experiments on well-established datasets for the task of automatic facial expression recognition. The results of those experiments are suggesting that our approach helps to accelerate the learning process.

show abstract

Section: Resultsmentioning

confidence: 99%

Section: Overview Of Our Approachmentioning

confidence: 99%

Section: Masking the Inputmentioning

confidence: 99%

See 1 more Smart Citation

Relevance-Based Data Masking: A Model-Agnostic Transfer Learning Approach for Facial Expression Recognition

Schiller

Huber

Dietz

et al. 2020

Front. Comput. Sci.

Self Cite

View full text Add to dashboard Cite

show abstract

“…While a binary map that exactly pinpoints the location of classdiscriminative features in the input is the most desirable and has been found to improve classification performance [28], such accurate maps require human-in-the-loop processing and cannot be readily available for most datasets. Our strategy also differs from feature muting [29] that sets a fixed number of features to zero, which can drastically change sample statistics and risk muting useful features through the fixed threshold. In DFDG, we use SmoothGrad saliency maps [15] to rank the relevance of each observation.…”

Section: Masking Superficial Observationsmentioning

confidence: 99%

Robust Domain-Free Domain Generalization with Class-Aware Alignment

Zhang

Ragab

Sagarna

2021

ICASSP 2021 - 2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

View full text Add to dashboard Cite

While deep neural networks demonstrate state-of-the-art performance on a variety of learning tasks, their performance relies on the assumption that train and test distributions are the same, which may not hold in real-world applications. Domain generalization addresses this issue by employing multiple source domains to build robust models that can generalize to unseen target domains subject to shifts in data distribution. In this paper, we propose Domain-Free Domain Generalization (DFDG), a model-agnostic method to achieve better generalization performance on the unseen test domain without the need for source domain labels. DFDG uses novel strategies to learn domain-invariant class-discriminative features. It aligns class relationships of samples through class-conditional soft labels, and uses saliency maps, traditionally developed for post-hoc analysis of image classification networks, to remove superficial observations from training inputs. DFDG obtains competitive performance on both time series sensor and image classification public datasets.

show abstract

“…While such saliency map algorithms are already well established and were even used to improve classification models [15], they are usually developed with experienced machine learning practitioners in mind. This can make the generated explanations difficult to interpret for beginners or users who are unrelated to the field of machine learning.…”

Section: Introductionmentioning

confidence: 99%

Enhancing Explainability of Deep Reinforcement Learning Through Selective Layer-Wise Relevance Propagation

Huber

Schiller

André

2019

Lecture Notes in Computer Science

Self Cite

View full text Add to dashboard Cite

Modern deep reinforcement learning agents are capable of achieving superhuman performance in tasks like playing Atari games, solely based on visual input. However, due to their use of neural networks the trained models are lacking transparency which makes their inner workings incomprehensible for humans. A promising approach to gain insights into the opaque reasoning process of neural networks is the layer-wise relevance propagation (LRP) concept. This visualization technique creates saliency maps that highlight the areas in the input which were relevant for the agents' decision-making process. Since such saliency maps cover every possible cause for a prediction, they are often accentuating very diverse parts of the input. This makes the results difficult to understand for people without a machine-learning background. In this work, we introduce an adjustment to the LRP concept that utilizes only the most relevant neurons of each convolutional layer and thus generates more selective saliency maps. We test our approach with a dueling Deep Q-Network (DQN) agent which we trained on three different Atari games of varying complexity. Since the dueling DQN approach considerably alters the neural network architecture of the original DQN algorithm, it requires its own LRP variant which will be presented in this paper.

show abstract

Relevance-Based Feature Masking: Improving Neural Network Based Whale Classification Through Explainable Artificial Intelligence

Cited by 12 publications

References 18 publications

Relevance-Based Data Masking: A Model-Agnostic Transfer Learning Approach for Facial Expression Recognition

Relevance-Based Data Masking: A Model-Agnostic Transfer Learning Approach for Facial Expression Recognition

Robust Domain-Free Domain Generalization with Class-Aware Alignment

Enhancing Explainability of Deep Reinforcement Learning Through Selective Layer-Wise Relevance Propagation

Contact Info

Product

Resources

About