傳統計算的主要特徵是嚴格、確定和精確，但其並不適合處理現實生活中的許多問題，例如駕駛汽車、下棋、家電控制…等。但軟式計算基於其不確定、不精確及不完全真值的容錯特性，可提供低成本的方案解決許多日常生活中的問題。本課程將介紹軟式計算的基本原理、相關計算模式，及其在人工智慧與機器學習領域的應用。本課程將介紹的軟式計算的計算模式主要包括了: Neural networks、fuzzy logic、evolutionary computation、simulated annealing、swarm intelligence…等。此外，在機器學習領域，本課程將著重介紹強化學習 (reinforcement learning)，其靈感發源於心理學的行為主義，有機體在環境給予的獎勵或懲罰的刺激下，逐步形成對刺激的預期，因而產生能獲得最大利益的習慣性行為。強化學習和傳統的監督式學習 (supervised learning) 間的主要區別在於，它在學習時並不需要使用完全正確的輸入/輸出樣本資料，故其需要在未知領域探索和遵從現有知識間找到平衡。強化學習在許多問題上得到應用，包括機器人控制、電梯調度、電信通訊及下棋...等。近年來比較知名的應用包括了: Alpha Go/Alpha Zero、DeepMind 的跑酷機器人、爲 Google 的能源中心節能...等。本課程將搭配相關工具軟體的實際演練，讓學生未來可易於將所學套用在研究與工作上。

The main characteristics of traditional computing are strict, deterministic and precise, but it is not suitable for dealing with many problems in real life, such as driving a car, playing chess, controlling home appliances, etc. However, soft computing can provide low-cost solutions to many problems in daily life based on its fault-tolerant characteristics of uncertainty, imprecision and incomplete truth values. This course will introduce the basic principles of soft computing, related computing models, and its applications in the fields of artificial intelligence and machine learning. The computing models of soft computing that this course will introduce mainly include: neural networks, fuzzy logic, evolutionary computation, simulated annealing, swarm intelligence...etc. In addition, in the field of machine learning, this course will focus on reinforcement learning, which is inspired by behaviorism in psychology. Under the stimulation of rewards or punishments given by the environment, organisms gradually form expectations of stimuli, thus producing habitual behaviors that can obtain the greatest benefit. The main difference between reinforcement learning and traditional supervised learning (supervised learning) is that it does not need to use completely correct input/output sample data when learning, so it needs to find a balance between exploring unknown areas and following existing knowledge. Reinforcement learning has been applied to many problems, including robot control, elevator scheduling, telecommunications, and chess playing...etc. Well-known applications in recent years include: Alpha Go/Alpha Zero, DeepMind’s parkour robot, saving energy for Google’s energy center, etc. This course will be paired with practical exercises on relevant tools and software so that students can easily apply what they have learned to research and work in the future.

1. Sebastian Raschka,‎ Vahid Mirjalili, Python Machine Learning: Machine Learning and Deep Learning with Python, scikit-learn, and TensorFlow, 2nd Edition, Packt Publishing Limited., September 20, 2017.
2. Prateek Joshi, Artificial Intelligence with Python, Packt Publishing Limited., January 2017.
3. Sudharsan Ravichandiran, Hands-On Reinforcement Learning with Python, Packt Publishing Limited., June 2018.

1. Sebastian Raschka,‎ Vahid Mirjalili, Python Machine Learning: Machine Learning and Deep Learning with Python, scikit-learn, and TensorFlow, 2nd Edition, Packt Publishing Limited., September 20, 2017.
2. Prateek Joshi, Artificial Intelligence with Python, Packt Publishing Limited., January 2017.
3. Sudharsan Ravichandiran, Hands-On Reinforcement Learning with Python, Packt Publishing Limited., June 2018.