Deep Reinforcement Learning: An Overview (by Yuxi Li, version 3, July 15, 2017) We give an overview of recent exciting achievements of deep reinforcement learning (RL). We discuss six core elements, six important mechanisms, and twelve applications. We start with background of machine learning, deep learning and reinforcement learning. Next we discuss core RL elements, including value function, in particular, Deep Q-Network (DQN), policy, reward, model, planning, and exploration. After that, we discuss important mechanisms for RL, including attention and memory, in particular, differentiable neural computer (DNC), unsupervised learning, transfer learning, semi-supervised learning, hierarchical RL, and learning to learn. Then we discuss various applications of RL, including games, in particular, AlphaGo, robotics, natural language processing, including dialogue systems (a.k.a. chatbots), machine translation, and text generation, computer vision, neural architecture design, business management, finance, healthcare, Industry 4.0, smart grid, intelligent transportation systems, and computer systems. We mention topics not reviewed yet. After listing a collection of RL resources, we present a brief summary, and close with discussions.
ImageNet Object Localization Challenge This year, Kaggle is thrilled to be the official host of all three ImageNet Challenges for the first time including the other two competitions: Object Detection Challenge Object Detection from Video Challenge Deep Learning for NLP Best Practices This post is a collection of best practices for using neural networks in Natural Language Processing. It will be updated periodically as new insights become available and in order to keep track of our evolving understanding of Deep Learning for NLP.
This is Part 1 of 3 in a series of posts that looks at the landscape of the artificial intelligence industry and the companies and institutes developing products that are moving the needle of knowledge of machine intelligence and consciousness forward for humanity.
This list contains companies working on artificial intelligence and machine learning products primarily for business use, non-specific to any industry. Industry specific AI will be the final part of this series.
There’s been a lot of talks about artificial intelligence (AI). Most notably, Google’s AlphaGodefeated Lee Se-dol last year. The latest one was the announcement of Toronto based Vector Institute for Artificial Intelligence which was backed by Geoffrey Hinton, the Godfather of Deep Learning and funded by government as well as large companies. Executives of most Internet giants believe the next decades will be a golden age for AI.
But how is this relevant to the senior IT professionals? They might think AI is only for those PhDs working for Google Deepmind, Facebook, or professors in the Stanford Lab. If they don’t have an academic background in AI and mathematics, does it mean that the AI trend is not within their reach? Is their job going to be impacted by AI?
To help its 2000+ members better position themselves to participate in rather than being hit by the AI wave, Association of Chinese Senior IT Professionals(ACSIP) partnered with IBM Canada to held the first AI presentation of series of seminars at IBM Amphitheater in Markham on May 14, 2017. The two parties had worked together to promote the Big Data technology as well as Watson Analytics in the past. This event was another successful joint venture and drew more than 150 IT professional in GTA to attend. As honorable guests, Dr. Yuxi Li, the author of recently published paper “Deep Reinforcement Learning: An Overview” and three contributors from Synced, China’s first media entity that focuses on reporting machine intelligence and related technologies, also came to support the event.
IBM has been playing a key role and heavily investing in the AI. Watson, the world's first cognitive system, is the fruit of over 50 years of IBM research in AI. Saeed Aghabozorgi, PhD is a Data Scientist in IBM. He introduced the audience about the Big Data University (BDU), an IBM community initiative, which is also helping promote the AI technologies through free courses such as Deep Learning with TensorFlow. The course content is free, access to tool sets used within the courses is free.
The first speaker Kent Yu is the co-founder of ACSIP. He is.a software architect and certified Scrum Master, working for a world leading software company serving Fortune 500 clients. He is also an entrepreneur who founded and sold 2 startups. Although he possesses a MBA degree and a Computer Science degree, he doesn’t have any PhD degree. He is now a student of Jeremy Howard, former Kaggle president and chief data scientist, and founder of Enlitic. Starting his presentation by Bach Music Test, Kent showed how AI has become very good at imitating human composers. He then uncover the deep learning myth by comparing the human neuron with artificial neuron. Through the demos, he continued to explain how neural training works and how Convolutional Neural Network (CNN) improves computer vision. Regarding AlphaGo, Kent mainly touched the reinforcement learning.
Different from Kent, the second speaker Joseph Santarcangelo use mathematical approach to explain neural networks. Joe has a PhD in Electrical Engineering, his research focused on using machine learning, signal processing and computer vision to determine how videos impact human cognition. Joseph has been working for IBM since he completed his PhD. His main point was “a neural network is a function that can use to approximate ‘something’ using a set of parameters”. Starting from the equation of a line, he focused on classification area to explain how neural network works.
Given the purpose of this event, Kent also shared his own experience on AI learning. He proved that coders without prior AI background could also learn how to build state-of-art deep learning models beating the best academic results. To help those audience who might like moving forward in the AI field, Kent gave his advice - Focus on Your Strengths. In other words, if you are good at programming, you had better start from reading codes rather than reading papers.
