I carried out few experiments by combining 7 CNN based feature extractors and 7 machine learning classifier. In total there were 49 pairs.
For CNNs i used VGG16, Mobile net v2, Densenet121, Inception V3, ResNet 101, ResNet 152 and XceptionNet as feature extractors and passed the generated feature vector to ML classifiers to perform binary classification.
The ML classifiers i have used are support vector machine, k nearest neighbour, gaussian naive Bayes, decision tree, random forest, extra tress and Multi layer perceptron.
For all the evaluation metrics like accuracy, precision, recall and F1 score i achieved best results with the combination of ResNet 101 and Multi layer perceptron.
I'm not able to understand that why it is performing the best. Resnet152 has a deeper network and support vector machine generally perform well. In my case Resnet101 and multi layer perceptron is giving the best results.
Please help me to understand the reason behind it.
Dear Akhil Kumar ,
Traditional deep learning methods are sub-optimal in classifying ambiguity features, which often arise in noisy and hard to predict categories, especially, to distinguish semantic scoring. Semantic scoring, depending on semantic logic to implement evaluation, inevitably contains fuzzy description and misses some concepts, for example, the ambiguous relationship between normal and probably normal always presents unclear boundaries (normal − more likely normal - probably normal). Thus, human error is common when annotating images. Differing from existing methods that focus on modifying kernel structure of neural networks, this study proposes a dominant fuzzy fully connected layer (FFCL) for Breast Imaging Reporting and Data System (BI-RADS) scoring and validates the universality of this proposed structure. This proposed model aims to develop complementary properties of scoring for semantic paradigms, while constructing fuzzy rules based on analyzing human thought patterns, and to particularly reduce the influence of semantic conglutination.
Article A Heuristic Neural Network Structure Relying on Fuzzy Logic ...
Regards,
Shafagat
The mathematical model of this technique is described and supported by experimental results on different datasets with complex analysis of different heuristics.
ResNet is a way to handle the vanishing gradient problem in very deep CNNs. They work by skipping some layers assuming the fact that very deep networks should not produce a training error higher than its shallower counterparts. In an overall perspective they can be thought of as a model similar to LSTM in RNNs
In my opinion, if you will use pre-trained CNN based on ImageNet then better results can give Inception V3 or Xception.
Article Architectural and mathematical fundamentals of improvement n...
I agree with @Arturo Geigel that it will depend on a dataset.
- Badges
- Science topic
- Similar topics
- Correction