Table 3.
Overview of XAI techniques.
| References | Local/ | Model specific/ | Intrinsic/ | Dataset | Models examined |
|---|---|---|---|---|---|
| global | model agnostic | post-hoc | |||
| Che et al. (2015) | G | MA | PH | Khemani et al. (2009) | 1. Deep feed-forward neural network |
| 2. Stack denoising autoencoder | |||||
| 3. Long Short-Term Memory | |||||
| Ribeiro et al. (2016) | L | MA | PH | Blitzer et al. (2007) | 1. InceptionV3 |
| 2. Word2vec | |||||
| Ribeiro et al. (2018) | L | MA | PH | Ribeiro et al. (2018) | InceptionV3 |
| Lundberg and Lee (2017) | L, G | MA | PH | Deng (2012) | CNN: 2 Conv layer, 1 FCN layer |
| Lapuschkin et al. (2016) | L | MS | PH | Everingham et al. (2015) | 1. BVLC reference classifier |
| 2. VGG16 | |||||
| 3. GoogleNet | |||||
| Hendricks et al. (2016) | L | MS | PH | Wah et al. (2011) | Proposed model combining VGG16 and LSTM |
| Zhou et al. (2016) | L | MS | PH | Russakovsky et al. (2015) | 1. Network in network |
| 2. GoogleNet | |||||
| 3. VGG16 | |||||
| Selvaraju et al. (2017) | L | MS | PH | Russakovsky et al. (2015) | 1. VGG16 |
| 2. AlexNet | |||||
| 3. Neuraktalk2 | |||||
| Chattopadhay et al. (2018) | L | MS | PH | 1. Russakovsky et al. (2015) | VGG16 |
| 2. Everingham et al. (2015) | |||||
| Simonyan et al. (2013) | L | MS | PH | Berg et al. (2010) | ImageNet Classification with deep sonvolutional neural networks |
| Li et al. (2018) | G | MA | PH | Krizhevsky and Hinton (2009b) | 1. ResNet |
| 2. DenseNet | |||||
| 3. VGG16 | |||||
| Bach et al. (2015) | L | MA | PH | 1. Everingham et al. (2015) | 1. Shallow CNN |
| 2. Deng (2012) | 2. Caffe open source pack-age | ||||
| Ghorbani et al. (2019b) | G | MA | PH | Russakovsky et al. (2015) | InceptionV3 |
| Ibrahim et al. (2019) | G | MA | PH | 1. Synthetic data | Shallow CNN |
| 2. Dua et al. (2017) | |||||
| Agarwal et al. (2021) | G | MS | PH | 1. Saeed et al. (2011) | Shallow CNN |
| 2. ProPublica (2016) | |||||
| Zeiler and Fergus (2014) | L | MA | PH | 1. Fei-Fei et al. (2006) | ImageNet Classification with deep convolutional neural networks |
| 2. Griffin et al. (2007) | |||||
| 3. Everingham and Winn (2011) | |||||
| 4. Deng et al. (2009) | |||||
| Zintgraf et al. (2017) | L | MA | PH | Deng et al. (2009) | 1. AlexNet |
| 2. GoogleNet | |||||
| 3. VGG16 | |||||
| L | MA | PH | 1. Krizhevsky and Hinton (2009b) | 1. ImageNet Classification with deep convolutional neural networks | |
| 2. Krizhevsky and Hinton (2009a) | 2. Network in network | ||||
| 3. Russakovsky et al. (2015) | |||||
| Burns et al. (2020) | L | MA | PH | Deng et al. (2009) | 1. InceptionV3 |
| 2. Bidirectional encoder representations from transformers | |||||
| Soares et al. (2020) | G | MS | PH | Nageshrao et al. (2019) | - |
| Angelov and Soares (2020) | G | - | I | 1. Rezaei and Terauchi (2013) | Proposed the model |
| 2. Griffin et al. (2007) | |||||
| 3. Fei-Fei et al. (2006) | |||||
| 4. Yang et al. (2020) | |||||
| Lee et al. (2019) | L | MA | PH | Wang et al. (2017) | 1. VGG16 |
| 2. ResNet50 | |||||
| 3. InceptionV3 | |||||
| 4. Inception-RecNet-v2 | |||||
| Brunese et al. (2020) | L | MA | PH | Cohen et al. (2020) | VGG16 |
| Assaf and Schumann (2019) | L | MS | PH | Energy consumption of photovoltaic power plant | Proposed the model |
| Nigri et al. (2020) | G | MS | PH | 1. Weiner et al. (2013) | 1. AlexNet |
| 2. Ellis et al. (2009) | 2. VGG16 | ||||
| 3. ResNet50 | |||||
| Erion et al. (2019) | G | - | I | Krizhevsky and Hinton (2009b) | VGG16 |
L, Local; G, Global; MA, Model Agnostic; MS, Model Specific; PH, Post-Hoc; I, Intrinsic; and Dataset refers to where the dataset used in the study was first proposed.