We assume you already know the basic knowledge of deep learning, and here we will present the implementation details (tricks or tips) in Deep Neural Networks, especially CNN for image-related tasks, mainly in eight aspects: 1) data augmentation; 2) pre-processing on images; 3) initializations of Networks; 4) some tips during training; 5) selections of activation functions; 6) diverse regularizations; 7)some insights found from figures and finally 8) methods of ensemble multiple deep networks. Additionally, the corresponding slides are available at [slide]. If there are any problems/mistakes in these materials and slides, or there are something important/interesting you consider that should be added, just feel free to contact me.

Artificial intelligence has a new target in its cross hairs for 2017: StarCraft, a real-time war strategy game series. Like other gamers before them, StarCraft fans may soon be forced to bow down to their machine overlords, as some of the biggest AI research groups set out to beat the best human players. Demis Hassabis, cofounder of Google-owned firm DeepMind, and Jeff Dean, who leads the Google Brain project, have both hinted that StarCraft will be their next target, while Facebook researchers have just released an open-source platform designed to help people develop AI to play the game. Succeeding in StarCraft would be a show of strategic strength. AI's gaming prowess reached new heights in March when DeepMind's AlphaGo system defeated one of the world's best Go players, Lee Sedol.

When we come into this world, it takes us three years or less to build up in our brain a model box with which to reconstruct in our mind real-time representations of the reality immediately surrounding us. We are then able to speak and be conscious about that reality, act purposefully and start establishing in our memory a personal history. The raw amount of sensory information we have taken up by that time is less than 200 Million sensory patterns -- counting a handful per waking second. A decent nursery could probably be recreated by a compact virtual reality system not bigger than the human genome. Compare this to the flood of information needed to train a deep learning system on a single task like sorting objects on photographs, each into one of a thousand classes.

Machines may soon be trying to master just about anything you can do on a computer. Open AI, a nonprofit dedicated to pursuing big advances in AI and making that progress freely available to anyone, has released Universe, a platform that will let AI programs learn, through experimentation and positive reward, how to do all sorts of things on a computer. Universe will include more than a thousand games, but also desktop programs such as Web browsers. It will make it possible for AI researchers to train programs to do all sorts of new tricks, including potentially useful tasks like filling out online forms, responding to e-mails, and updating spreadsheets. But Ilya Sutskevar, cofounder and research director at OpenAI, says the motivation for developing and releasing Universe is a lot bigger.

Has Deep Learning Made Traditional Machine Learning Irrelevant?

Autonomous machines shining lasers at ants -- and spraying water at bewildered cats -- for the amusement of cackling grandchildren. Hobbyists are just getting started with deep-learning technologies that give them cheap, off-the-shelf capabilities that put Ronald Reagan's Star Wars program to shame. In the latest edition of the AI Podcast, NVIDIA engineer Bob Bond and Make: Magazine Executive Editor Mike Senese explain to host Michael Copeland how they've taken the once esoteric technology of deep learning and put it to work on offbeat projects that can be tackled on budgets of a few hundred bucks. "One of the big things that's happening -- and it's happening in real time right now -- is the technology is finally hitting a point where we, as consumers, have access to this type of capability," Senese says. Bond, a veteran engineer, is no technical novice.

The race is on between the big tech giants to develop the best artificially intelligent assistant on almost human parity levels and Zo is next in line. It seems 2016 is the year of the Artificial Intelligence (AI) assistant or indeed, chatbot. Their success depends on the machine's "IQ and EQ [Emotional Quotient -- ability to understand the emotions of others]," Harry Shum executive VP of Microsoft's AI research group told a conference in San Francisco. Creating #AI for all: Microsoft Ventures supports startups focused on inclusive growth & societal good. IQ can been developed by using deep learning techniques and speech recognition software and is essential if the bot is going to complete specific tasks.

Since 2006, deep learning (DL) has become a rapidly growing research direction, redefining state-of-the-art performances in a wide range of areas such as object recognition, image segmentation, speech recognition and machine translation. In modern manufacturing systems, data-driven machine health monitoring is gaining in popularity due to the widespread deployment of low-cost sensors and their connection to the Internet. Meanwhile, deep learning provides useful tools for processing and analyzing these big machinery data. The main purpose of this paper is to review and summarize the emerging research work of deep learning on machine health monitoring. After the brief introduction of deep learning techniques, the applications of deep learning in machine health monitoring systems are reviewed mainly from the following aspects: Auto-encoder (AE) and its variants, Restricted Boltzmann Machines and its variants including Deep Belief Network (DBN) and Deep Boltzmann Machines (DBM), Convolutional Neural Networks (CNN) and Recurrent Neural Networks (RNN). Finally, some new trends of DL-based machine health monitoring methods are discussed.

Existing models based on artificial neural networks (ANNs) for sentence classification often do not incorporate the context in which sentences appear, and classify sentences individually. However, traditional sentence classification approaches have been shown to greatly benefit from jointly classifying subsequent sentences, such as with conditional random fields. In this work, we present an ANN architecture that combines the effectiveness of typical ANN models to classify sentences in isolation, with the strength of structured prediction. Our model achieves state-of-the-art results on two different datasets for sequential sentence classification in medical abstracts.

Engineers.SG 32,435 views Sebastian Raschka Learning scikit learn - An Introduction to Machine Learning in Python - Duration: 1:26:40.