Abstract
Convolutional neural networks utilize a hierarchy of neural network layers. The statistical aspects of information concentration in successive layers can bring an insight into the feature abstraction process. We analyze the saliency maps of these layers from the perspective of semiotics, also known as the study of signs and sign-using behavior. In computational semiotics, this aggregation operation (known as superization) is accompanied by a decrease of spatial entropy: signs are aggregated into supersign. Using spatial entropy, we compute the information content of the saliency maps and study the superization processes which take place between successive layers of the network. In our experiments, we visualize the superization process and show how the obtained knowledge can be used to explain the neural decision model. In addition, we attempt to optimize the architecture of the neural model employing a semiotic greedy technique. To the extent of our knowledge, this is the first application of computational semiotics in the analysis and interpretation of deep neural networks.
Highlights
Convolutional neural networks (CNNs) were first made popular by Lecun et al [1] with their seminal work on handwritten character recognition, where they introduced the currently popular LeNet-5 architecture
We focus on the statistical aspects of the information concentration processes which appear in the saliency maps of successive CNN layers
We study the possibility of applying semiotic tools to optimize the architecture of deep learning neural networks
Summary
Convolutional neural networks (CNNs) were first made popular by Lecun et al [1] with their seminal work on handwritten character recognition, where they introduced the currently popular LeNet-5 architecture. Computational power constraints and lack of data prohibited those CNNs from achieving their true potential in terms of computer vision capabilities. Marked the start of the current deep learning revolution, when, during the ILSVRC 2012 competition, their CNN, entitled AlexNet, overrun its competitor from the previous year by a margin of almost 10%. Despite the ability of generating human-alike predictions, CNNs still lack a major component: interpretability. Neural networks in general are known for their black-box type of behavior, being capable of capturing semantic information using numerical computations and gradient-based learning, but hiding the inner working mechanisms of reasoning. There is a trade-off between accuracy and interpretability
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
