This work demonstrates that deep neural networks (DNNs) can solve a combinatorial problem merely through self-supervised learning. While researchers have employed explicit logic, heuristics, and reinforcement learning to tackle combinatorial problems, such methods are often complex and costly to implement, requiring lots of knowledge, coding, and adjustments. Hence, in the present study, I propose a robust and straightforward method of self-supervised learning to solve a combinatorial problem. Specifically, taking Rubik's Cube as an example, this work shows that a DNN can implicitly learn convoluted probability distributions of optimal choices from randomly generated combinations. Tested on $1,000$ Rubik's Cube instances, a DNN successfully solved all of them near-optimally. Although the proposed method is validated only on Rubik's Cube, it is potentially useful for other problems and real-world applications with its simplicity, stability, and robustness.
Over the last several years, research on facial recognition based on Deep Neural Network has evolved with approaches like task-specific loss functions, image normalization and augmentation, network architectures, etc. However, there have been few approaches with attention to how human faces differ from person to person. Premising that inter-personal differences are found both generally and locally on the human face, I propose FusiformNet, a novel framework for feature extraction that leverages the nature of discriminative facial features. Tested on Image-Unrestricted setting of Labeled Face in the Wild benchmark, this method achieved a state-of-the-art accuracy of 96.67% without labeled outside data, image augmentation, normalization, or special loss functions. Likewise, the method also performed on par with previous state-of-the-arts when pre-trained on CASIA-WebFace dataset. Considering its ability to extract both general and local facial features, the utility of FusiformNet may not be limited to facial recognition but also extend to other DNN-based tasks.