scholarly journals Explainable AI for ML jet taggers using expert variables and layerwise relevance propagation

2021 ◽  
Vol 2021 (5) ◽  
Author(s):  
Garvita Agarwal ◽  
Lauren Hay ◽  
Ia Iashvili ◽  
Benjamin Mannix ◽  
Christine McLean ◽  
...  
Keyword(s):  
Network Performance ◽  
Deep Neural Network ◽  
Relevant Information ◽  
Jet Substructure ◽  
Discrimination Ability ◽  
Low Level ◽  
Performance Improvements ◽  
Intermediate Layers ◽  
Explainable Ai ◽  
General Method

Abstract A framework is presented to extract and understand decision-making information from a deep neural network (DNN) classifier of jet substructure tagging techniques. The general method studied is to provide expert variables that augment inputs (“eXpert AUGmented” variables, or XAUG variables), then apply layerwise relevance propagation (LRP) to networks both with and without XAUG variables. The XAUG variables are concatenated with the intermediate layers after network-specific operations (such as convolution or recurrence), and used in the final layers of the network. The results of comparing networks with and without the addition of XAUG variables show that XAUG variables can be used to interpret classifier behavior, increase discrimination ability when combined with low-level features, and in some cases capture the behavior of the classifier completely. The LRP technique can be used to find relevant information the network is using, and when combined with the XAUG variables, can be used to rank features, allowing one to find a reduced set of features that capture part of the network performance. In the studies presented, adding XAUG variables to low-level DNNs increased the efficiency of classifiers by as much as 30-40%. In addition to performance improvements, an approach to quantify numerical uncertainties in the training of these DNNs is presented.

scholarly journals Deep Neural Network with Joint Distribution Matching for Cross-Subject Motor Imagery Brain-Computer Interfaces

2020 ◽  
Vol 2020 ◽  
pp. 1-15
Author(s):  
Xianghong Zhao ◽  
Jieyu Zhao ◽  
Cong Liu ◽  
Weiming Cai
Keyword(s):  
Neural Network ◽  
Motor Imagery ◽  
Joint Distribution ◽  
Deep Neural Network ◽  
Brain Computer Interfaces ◽  
Target Domain ◽  
Common Space ◽  
Computer Interfaces ◽  
Intermediate Layers ◽  
Distribution Matching

Motor imagery brain-computer interfaces (BCIs) have demonstrated great potential and attract world-spread attentions. Due to the nonstationary character of the motor imagery signals, costly and boring calibration sessions must be proceeded before use. This prevents them from going into our realistic life. In this paper, the source subject’s data are explored to perform calibration for target subjects. Model trained on source subjects is transferred to work for target subjects, in which the critical problem to handle is the distribution shift. It is found that the performance of classification would be bad when only the marginal distributions of source and target are made closer, since the discriminative directions of the source and target domains may still be much different. In order to solve the problem, our idea comes that joint distribution adaptation is indispensable. It makes the classifier trained in the source domain perform well in the target domain. Specifically, a measure for joint distribution discrepancy (JDD) between the source and target is proposed. Experiments demonstrate that it can align source and target data according to the class they belong to. It has a direct relationship with classification accuracy and works well for transferring. Secondly, a deep neural network with joint distribution matching for zero-training motor imagery BCI is proposed. It explores both marginal and joint distribution adaptation to alleviate distribution discrepancy across subjects and obtain effective and generalized features in an aligned common space. Visualizations of intermediate layers illustrate how and why the network works well. Experiments on the two datasets prove the effectiveness and strength compared to outstanding counterparts.

scholarly journals Deep-learned faces of pain and emotions: Elucidating the differences of facial expressions with the help of explainable AI methods

2019 ◽  
Vol 86 (7-8) ◽  
pp. 404-412 ◽  
Author(s):  
Katharina Weitz ◽  
Teena Hassan ◽  
Ute Schmid ◽  
Jens-Uwe Garbas
Keyword(s):  
Facial Expressions ◽  
Deep Neural Network ◽  
Deep Neural Networks ◽  
Black Box ◽  
Limited Extent ◽  
Linear Relationships ◽  
Learning Procedure ◽  
Explainable Ai ◽  
Interpretable Model ◽  
Pain Recognition

AbstractDeep neural networks are successfully used for object and face recognition in images and videos. In order to be able to apply such networks in practice, for example in hospitals as a pain recognition tool, the current procedures are only suitable to a limited extent. The advantage of deep neural methods is that they can learn complex non-linear relationships between raw data and target classes without limiting themselves to a set of hand-crafted features provided by humans. However, the disadvantage is that due to the complexity of these networks, it is not possible to interpret the knowledge that is stored inside the network. It is a black-box learning procedure. Explainable Artificial Intelligence (AI) approaches mitigate this problem by extracting explanations for decisions and representing them in a human-interpretable form. The aim of this paper is to investigate the explainable AI methods Layer-wise Relevance Propagation (LRP) and Local Interpretable Model-agnostic Explanations (LIME). These approaches are applied to explain how a deep neural network distinguishes facial expressions of pain from facial expressions of emotions such as happiness and disgust.

scholarly journals Handwritten Bangla Numerical Digit Recognition Using Fine Regulated Deep Neural Network

2021 ◽  
Vol 9 (2) ◽  
pp. 73-84
Author(s):  
Md. Shahadat Hossain ◽  
Md. Anwar Hossain ◽  
AFM Zainul Abadin ◽  
Md. Manik Ahmed
Keyword(s):  
Neural Network ◽  
Character Recognition ◽  
Optical Character Recognition ◽  
Network Performance ◽  
Deep Neural Network ◽  
Recognition Accuracy ◽  
Digit Recognition ◽  
Optical Character ◽  
Regularization Parameters ◽  
Handwritten Recognition

The recognition of handwritten Bangla digit is providing significant progress on optical character recognition (OCR). It is a very critical task due to the similar pattern and alignment of handwriting digits. With the progress of modern research on optical character recognition, it is reducing the complexity of the classification task by several methods, a few problems encounter during recognition and wait to be solved with simpler methods. The modern emerging field of artificial intelligence is the Deep Neural Network, which promises a solid solution to these few handwritten recognition problems. This paper proposed a fine regulated deep neural network (FRDNN) for the handwritten numeric character recognition problem that uses convolutional neural network (CNN) models with regularization parameters which makes the model generalized by preventing the overfitting. This paper applied Traditional Deep Neural Network (TDNN) and Fine regulated deep neural network (FRDNN) models with a similar layer experienced on BanglaLekha-Isolated databases and the classification accuracies for the two models were 96.25% and 96.99%, respectively over 100 epochs. The network performance of the FRDNN model on the BanglaLekha-Isolated digit dataset was more robust and accurate than the TDNN model and depend on experimentation. Our proposed method is obtained a good recognition accuracy compared with other existing available methods.

scholarly journals Attention Control and Process Overlap Theory: Searching for Cognitive Processes Underpinning the Positive Manifold

2021 ◽  
Author(s):  
Alexander P. Burgoyne ◽  
Cody Anthony Mashburn ◽  
Jason S. Tsukahara ◽  
Randall W Engle
Keyword(s):  
Cognitive Ability ◽  
Cognitive Processes ◽  
Latent Variables ◽  
Fluid Intelligence ◽  
Strong Relationship ◽  
Relevant Information ◽  
Attention Control ◽  
Future Research ◽  
Discrimination Ability ◽  
Positive Correlations

Process overlap theory provides a contemporary explanation for the positive correlations observed among cognitive ability measures, a phenomenon which intelligence researchers refer to as the positive manifold. According to process overlap theory, cognitive tasks tap domain-general executive processes as well as domain-specific processes, and correlations between measures reflect the degree of overlap in the cognitive processes that are engaged when performing the tasks. In this article, we discuss points of agreement and disagreement between the executive attention framework and process overlap theory, with a focus on attention control: the domain-general ability to maintain focus on task-relevant information and disengage from irrelevant and no-longer relevant information. After describing the steps our lab has taken to improve the measurement of attention control, we review evidence suggesting that attention control can explain many of the positive correlations between broad cognitive abilities, such as fluid intelligence, working memory capacity, and sensory discrimination ability. Furthermore, when these latent variables are modeled under a higher-order g factor, attention control has the highest loading on g, indicating a strong relationship between attention control and domain-general cognitive ability. In closing, we reflect on the challenge of directly measuring cognitive processes and provide suggestions for future research.

scholarly journals Transfer Learning Using Rotated Image Data to Improve Deep Neural Network Performance

2014 ◽  
pp. 290-300 ◽  
Author(s):  
Telmo Amaral ◽  
Luís M. Silva ◽  
Luís A. Alexandre ◽  
Chetak Kandaswamy ◽  
Joaquim Marques de Sá ◽  
...  
Keyword(s):  
Neural Network ◽  
Transfer Learning ◽  
Network Performance ◽  
Deep Neural Network ◽  
Image Data
Sign in / Sign up

Export Citation Format

Share Document