China 
Africa 
Explore chapters and articles related to this topic
An Overview of Deep Learning in Industry
Published in Jay Liebowitz, Data Analytics and AI, 2020
Quan Le, Luis Miralles-Pechuán, Shridhar Kulkarni, Jing Su, Oisín Boydell
The most widely known example of a deep learning approach to play board games is probably that of DeepMind’s AlphaGo, which is an autonomous agent for the game of Go. AlphaGo defeated the world Go champion, Lee Sedol, 4-1 in March 2016 (Chouard, 2016) and continues to beat world-class human players. The AlphaGo model uses deep convolutional neural networks and a general tree search algorithm (Silver et al. 2017). The architecture of AlphaGo is especially interesting as it is a hybrid system incorporating Monte Carlo tree search algorithms, supervised learning (SL) policy networks, reinforcement learning policy networks, and value networks (Silver et al. 2016). The first of these components, Monte Carlo tree search, has been a mainstay of automated Go playing systems since the 1990s (Brugmann, 1993). The latter three components are implemented as CNN models with slightly different objectives. This makes AlphaGo an interesting mix of traditional and cutting edge AI techniques.
How to Untangle Complex Systems?
Published in Pier Luigi Gentili, Untangling Complex Systems, 2018
There are two main strategies to develop artificial intelligence: one is writing human-like intelligent programs running on computers or special-purpose hardware, and the other is neuromorphic engineering. For the first strategy, computer scientists are writing algorithms that can learn, analyze extensive data, and recognize patterns. At the same time, psychologists, biologists, and social scientists are giving information on human sensations, emotions, and intuitions. The merger of the two contributions provides algorithms that can easily communicate with us. Among the most promising algorithms, there are the artificial neural networks (remember Chapter 10, when we learned these algorithms for predicting chaotic time series) (Castelvecchi 2016). Recently, a program called AlphaGo Zero, based on an artificial neural network that learns through trial and error, has mastered the game of Go without any human data or guidance, and it has outperformed the skills of the best human players (Silver et al., 2017).
A Brief History of Artificial Intelligence
Published in Ron Fulbright, Democratization of Expertise, 2020
In 2016, Google’s AlphaGo, developed by DeepMind Technologies, defeated the reigning world champion in Go, Lee Sedol, a game vastly more complex than Chess (Silver et al., 2016; DeepMind, 2018a). Go defies brute-force attempts which rely on calculating and scoring millions of potential moves. Instead, AlphaGo used a Monte Carlo tree search algorithm to find its moves based on knowledge “learned” by machine learning. AlphaGo used an artificial neural network trained extensively both by human and computer play. Thus the tree search gets better which each iteration. An even stronger version called AlphaGo Master won 60 online games against professional human players over a one-week period.
Development of a real-time security management system for restricted access areas using computer vision and deep learning
Published in Journal of Transportation Safety & Security, 2022
Surveillance images are a collection of different pixels. CV performs tasks such as object recognition, tracking, and 3D modeling from image and video data through algorithms that analyze pixels and improve visual understanding of mechanical systems (Maire, 2009). Deep Learning is a machine learning technology that uses a number of double-layer neural networks to solve problems given to the system through pattern recognition (Seongeun, Subeom, Jeonghyeok, & Jongseok, 2016). AlphaGo, developed by Google DeepMind, is an artificial intelligence system based on deep learning. It has shown excellent problem-solving skills by winning four matches against world champion Baduk player (Borowiec, 2016). Artificial intelligence using deep learning is used to optimize the flow of traffic on the road (Behbahani, Mohammadian Amiri, Nadimi, & Ragland, 2020).
Performance Study of Minimax and Reinforcement Learning Agents Playing the Turn-based Game Iwoki
Published in Applied Artificial Intelligence, 2021
Santiago Videgaín, Pablo García Sánchez
Other technique is Deep Reinforcement Learning (van Hasselt, Guez, and Silver 2016). This method uses a deep neural network to approximate the values of . The state is the input of the neural network, the output is a Q-value for each possible action and are the parameters of the network. Deep Reinforcement Learning has led to great achievements in the recent history of artificial intelligence, as is the case of AlphaGo, created by DeepMind (Google) (Silver et al. 2016). In 2016, AlphaGo beat 18-time world Go champion Lee Sedol in 4 of the 5 games they played. Until then it was thought that this was a goal which artificial intelligence would still take many years to achieve. Another successful example of Deep Reinforcement Learning implementation is the one applied to classic Atari 2600 computer games, such as Space Invaders, Seaquest, Breakout, Beam Rider, etc., in the Arcade Learning Environment (Mnih et al. 2013). In this case, a Convolutional Neural Network is trained with a Deep Reinforcement Learning algorithm that uses as input the raw pixels and returns as output a value function that estimates future rewards.
Understanding the mechanism of human–computer game: a distributed reinforcement learning perspective
Published in International Journal of Systems Science, 2020
Zhinan Peng, Jiangping Hu, Yiyi Zhao, Bijoy K. Ghosh
In 2016, a computer program called AlphaGo beat the world champion Lee Sedol in a five-game match, which inspires the research enthusiasm for artificial intelligence (AI) in many academical communities (Gibney, 2016). Even though there have been many references devoted to understanding the mechanism of AlphaGo with reinforcement learning, what can not be ignored is that the computer agent AlphaGo improves its skills by constantly learning the experience of human players. On the other hand, computer agent can improve the ability of human players' decision-making. How to facilitate the evolution of interactions among multiple agents becomes an interesting question in the filed of imitation learning (Bao et al., 2018). From the game theory point of view (Camerer, 2011), the interaction process of computer agent and player can be regarded as a human-computer game/interaction process, where each side can improve their decision making through interaction with their opponent (Haima et al., 2017; Hassabis, 2017). The human-computer game has the characteristics that one side wins and the other side loses with the same modulus but opposite sign. This phenomenon also exists in nature and society, such as predator-prey chain (Lima, 2002), economic competition (Neuman & Morgenstern, 2007), politics election (Ware, 2009), and so on.