Guan Cuntai

Erico Tjoa, Hong Jing Khok, Tushar Chouhan et al.

This paper quantifies the quality of heatmap-based eXplainable AI (XAI) methods w.r.t image classification problem. Here, a heatmap is considered desirable if it improves the probability of predicting the correct classes. Different XAI heatmap-based methods are empirically shown to improve classification confidence to different extents depending on the datasets, e.g. Saliency works best on ImageNet and Deconvolution on Chest X-Ray Pneumonia dataset. The novelty includes a new gap distribution that shows a stark difference between correct and wrong predictions. Finally, the generative augmentative explanation is introduced, a method to generate heatmaps capable of improving predictive confidence to a high level.

Erico Tjoa, Guan Cuntai

The black-box nature of deep neural networks (DNN) has brought to attention the issues of transparency and fairness. Deep Reinforcement Learning (Deep RL or DRL), which uses DNN to learn its policy, value functions etc, is thus also subject to similar concerns. This paper proposes a way to circumvent the issues through the bottom-up design of neural networks with detailed interpretability, where each neuron or layer has its own meaning and utility that corresponds to humanly understandable concept. The framework introduced in this paper is called the Self Reward Design (SRD), inspired by the Inverse Reward Design, and this interpretable design can (1) solve the problem by pure design (although imperfectly) and (2) be optimized like a standard DNN. With deliberate human designs, we show that some RL problems such as lavaland and MuJoCo can be solved using a model constructed with standard NN components with few parameters. Furthermore, with our fish sale auction example, we demonstrate how SRD is used to address situations that will not make sense if black-box models are used, where humanly-understandable semantic-based decision is required.

Erico Tjoa, Guan Cuntai

This paper proposes two bottom-up interpretable neural network (NN) constructions for universal approximation, namely Triangularly-constructed NN (TNN) and Semi-Quantized Activation NN (SQANN). Further notable properties are (1) resistance to catastrophic forgetting (2) existence of proof for arbitrarily high accuracies (3) the ability to identify samples that are out-of-distribution through interpretable activation "fingerprints".

Erico Tjoa, Guan Cuntai

Ongoing efforts to understand deep neural networks (DNN) have provided many insights, but DNNs remain incompletely understood. Improving DNN's interpretability has practical benefits, such as more accountable usage, better algorithm maintenance and improvement. The complexity of dataset structure may contribute to the difficulty in solving interpretability problem arising from DNN's black-box mechanism. Thus, we propose to use pattern theory formulated by Ulf Grenander, in which data can be described as configurations of fundamental objects that allow us to investigate convolutional neural network's (CNN) interpretability in a component-wise manner. Specifically, U-Net-like structure is formed by attaching expansion blocks (EB) to ResNet, allowing it to perform semantic segmentation-like tasks at its EB output channels designed to be compatible with pattern theory's configurations. Through these modules, some heatmap-based explainable artificial intelligence (XAI) methods will be shown to extract explanations w.r.t individual generators that make up a single data sample, potentially reducing the impact of dataset's complexity to interpretability problem. The MNIST-equivalent dataset containing pattern theory's elements is designed to facilitate smoother entry into this framework, along which the theory's generative aspect is naturally presented.

Erico Tjoa, Guan Cuntai

The practical application of deep neural networks are still limited by their lack of transparency. One of the efforts to provide explanation for decisions made by artificial intelligence (AI) is the use of saliency or heat maps highlighting relevant regions that contribute significantly to its prediction. A layer-wise amplitude filtering method was previously introduced to improve the quality of heatmaps, performing error corrections by noise-spike suppression. In this study, we generalize the layerwise error correction by considering any identifiable error and assuming there exists a groundtruth interpretable information. The forms of errors propagated through layerwise relevance methods are studied and we propose a filtering technique for interpretability signal rectification taylored to the trend of signal amplitude of the particular neural network used. Finally, we put forth arguments for the use of groundtruth interpretable information.