官术网_书友最值得收藏!

  • The Reinforcement Learning Workshop
  • Alessandro Palmas Emanuele Ghelfi Dr. Alexandra Galina Petre Mayur Kulkarni Anand N.S. Quan Nguyen Aritra Sen Anthony So Saikat Basak
  • 456字
  • 2021-06-11 18:37:45

Introduction

In the previous chapter, we covered the theory behind Reinforcement Learning (RL), explaining topics such as Markov chains and Markov Decision Processes (MDPs), Bellman equations, and a number of techniques we can use to solve MDPs. In this chapter, we will be looking at deep learning methods, all of which will play a primary role in building approximate functions for reinforcement learning. Specifically, we will look at different families of deep neural networks: fully connected, convolutional, and recurrent networks. These algorithms have the key capability of encoding knowledge that's been learned through examples in a compact and effective representation. In RL, they are typically used to approximate the so-called policy functions and value functions, which encode how the RL agent chooses its action, given the current state and the value associated with the current state, respectively. We will study the policy and value functions in the upcoming chapters.

Data is the new oil: This famous quote is being heard more and more frequently these days, especially in tech and economic industries. With the great amount of data available today, techniques to leverage such enormous quantities of information, thereby creating value and opportunities, are becoming key competitive factors and skills to have. All products and platforms that are provided to users for free (from social networks to apps related to wearable devices) use data that is provided by the users to generate revenues: think about the huge quantity of information they collect every day relating to our habits, preferences, or even body weight trends. These provide high-value insights that can be leveraged by advertisers, insurance companies, and local businesses to improve their offers so that they fit the market.

Thanks to the relevant increase in computational power availability and theory breakthroughs such as backpropagation-based training, deep learning has seen an explosion in the last 10 years, achieving unprecedented results in many fields, from image processing to speech recognition to natural language processing and understanding. In fact, it is now possible to successfully train large and deep neural networks by leveraging huge amounts of data and overcoming practical roadblocks that impeded their adoption in past decades. These models demonstrated the capability to exceed human performances in terms of both speed and accuracy. This chapter will teach you how to adopt deep learning to solve real-world problems by taking advantage of the top machine learning frameworks. TensorFlow and Keras, are the de facto production standards in the industry. Their success is mainly related to two aspects: TensorFlow's unrivaled performance in production environments in terms of both speed and scalability, and Keras' ease of use, which provides a very powerful, high-level interface that can be used to create deep learning models.

Now, let's take a look at the frameworks.

主站蜘蛛池模板: 潜江市| 青海省| 永靖县| 沧州市| 普定县| 金乡县| 长乐市| 苗栗市| 唐河县| 贵南县| 盐边县| 富阳市| 宁武县| 喀什市| 德惠市| 平南县| 陵川县| 胶南市| 浑源县| 定兴县| 石狮市| 芦溪县| 莱芜市| 靖安县| 禹州市| 明溪县| 阿荣旗| 宁安市| 浙江省| 满洲里市| 长泰县| 中牟县| 沙洋县| 抚宁县| 丰台区| 大渡口区| 剑阁县| 吉安县| 集贤县| 深州市| 九寨沟县|