Searching for internal symbols underlying deep learning

Deep learning (DL) enables deep neural networks (DNNs) to automatically learn complex tasks or rules from given examples without instructions or guiding principles. As we do not engineer DNNs' functions, it is extremely difficult to diagnose their decisions, and multiple lines of studies proposed to explain principles of DNNs/DL operations. Notably, one line of studies suggests that DNNs may learn concepts, the high level features recognizable to humans. Thus, we hypothesized that DNNs develop abstract codes, not necessarily recognizable to humans, which can be used to augment DNNs' decision-making. To address this hypothesis, we combined foundation segmentation models and unsupervised learning to extract internal codes and identify potential use of abstract codes to make DL's decision-making more reliable and safer.

Having Second Thoughts? Let's hear it

Deep learning models loosely mimic bottom-up signal pathways from low-order sensory areas to high-order cognitive areas. After training, DL models can outperform humans on some domain-specific tasks, but their decision-making process has been known to be easily disrupted. Since the human brain consists of multiple functional areas highly connected to one another and relies on intricate interplays between bottom-up and top-down (from high-order to low-order areas) processing, we hypothesize that incorporating top-down signal processing may make DL models more robust. To address this hypothesis, we propose a certification process mimicking selective attention and test if it could make DL models more robust. Our empirical evaluations suggest that this newly proposed certification can improve DL models' accuracy and help us build safety measures to alleviate their vulnerabilities with both artificial and natural adversarial examples.

Analysis of functional neural codes of deep learning models

Deep neural networks (DNNs), the agents of deep learning (DL), require a massive number of parallel/sequential operations. This makes it extremely challenging to comprehend DNNs' operations and hinders proper diagnosis. Consequently, DNNs cannot be readily used in high-stakes domains, in which incorrect decisions can lead to catastrophic failures. Therefore, to build more reliable DNNs/DL to be deployed in high-stakes real-world problems, it is imperative that we develop proper analysis tools that will allow us to better understand DNNs' internal operations underlying their decision-making. Here, we used the self-organizing map (SOM) to analyze internal codes of DL models associated with their decision-making. Our analyses suggest that hidden layer activation patterns can be mapped onto a finite number of patterns and are correlated with DL predictions, raising the possibility that they could serve as functional codes of DL models. Encouraged by this observation, we further used SOM to estimate input features coded in hidden layers, analyzed the effects of adversarial inputs to better understand characterized internal representations' evolution and adversarial perturbations' propagation in DL models.