|
Learning a Self-driving Bicycle Using Deep Deterministic Policy Gradient |
|
Tuyen Pham Le, Quang Dang Nguyen, SeungYoon Choi, TaeChoong Chung(Kyung Hee University, Korea) |
| This paper improves the method for learning a bicycle which can itself balance and go to any specified locations. The bicycle is controlled by a neural network policy which is learned by deep deterministic policy gradient algorithm (DDPG). We propose a procedure which allows the controller can be gradually learned until it can stably balance and lead the bicycle to any specified places. |
|
|
Heuristic approaches to a sequential variant of the p-supplier problem |
|
Kaoru Takehama, Yoshiyuki Karuno(Kyoto Institute of Technology, Japan) |
| A sequential location problem of facilities on a network is considered. An iterative improvement algorithm based on the simulated annealing is desigend for the sequential location problem, where a swapping neighborhood structure in a list of all facility location candidates is explored. Numerical experiments are conducted to demonstrate the improvement from a greedy heuristic algorithm. |
|
|
Synthesis of clustering algorithms based on selection of centroids |
|
Yeskendir Sultanov, Aiman Nickolayevna Moldagulova, Yedilkhan Amirgaliev, Zhaniya Sultanova(International Information Technology University, Kazakhstan) |
| In this paper, we made a review for such clustering algorithms as k-means, maxmin and reduction of centroids, also proposed a new method that makes synthesis of clustering algorithms. The main idea of this method was that the clustering algorithm that we choose can work well for many cases, but in some cases, it can give very poor results. Therefore, we select several algorithms and make a synthesis from their results. This method will always give a good result close to optimal. |
|
|
Spike-inspired Deep Neural Network Design Using Binary Weight |
|
Hyun Myung, Seunghee Lee, Kyukwang Kim, Jinki Kim, Yeeun Kim(KAIST, Korea) |
| Recently, deep learning has achieved great results in many fields. However, most general artificial neural networks require GPU because of hard workload and power consumption. Humans can do many things without consuming a lot of power compared to computers. Human beings or organisms transmit and recognize information through signal transmission between neurons. This study aims to develop a novel deep neural network architecture which simulates the signaling system between biological neurons, unlike conventional neural networks. We propose a novel spike-inspired deep neural network structure. |
|