Hello everyone, my name is PengXing Cai (蔡 彭杏). I am a Ph.D. in Advanced Mathematics and Human Mechanisms from Artificial Intelligence Lab. My research interests include artificial neural network, image recognition, and machine learning. I hope to design a novel evolutionary wide neural network to solving real-world engineering and intelligent identification problem.

My research is the design and application of evolutionary wide neural network based on dendritic neurons. With the appearance of more and more deep learning neural network models nowadays, the number of layers of deep neural networks is increasing, and the training time and space storage requirements are substantially higher, which will raise the whole threshold of deep learning and make high-performance computers a necessity for experiments, for which I will design a more intelligent, adaptable and faster evolutionary wide neural network. My research is based on a simple dendritic neuron model (DNM) and the way real neurons in the human brain transmit information to design a novel neural network. In detail, I take a single dendritic neuron as the object of study to construct a neural network model with a width structure. In order to obtain less time-space loss, I will make full use of the nonlinear information processing mechanism at the synapse to achieve automatic censoring of the neuron structure, and in subsequent studies, I may add the function of automatic evolution of the neuron in terms of specific function module, hyper-parameters, network structure and learning algorithm based on multi-objective evolutionary algorithms. In comparison with the current mainstream deep learning model, our designed model is based on biologically plausible DNM, with low model complexity, low computational consumption and excellent interpretability. After many experiments, we will apply the proposed model to solve image recognition problem, drug discovery and autonomous vehicles.（旧フェローシップ事業採択学生）