This series of blog posts aims to provide an intuitive and gentle introduction to deep learning that does not rely heavily on math or theoretical constructs. The first part in this series provided an overview over the field of deep learning, covering fundamental and core concepts. The third part of the series covers sequence learning topics such as recurrent neural networks and LSTM. I wrote this series in a glossary style so it can also be used as a reference for deep learning concepts. The earliest deep-learning-like algorithms that had multiple layers of non-linear features can be traced back to Ivakhnenko and Lapa in 1965 (Figure 1), who used thin but deep models with polynomial activation functions which they analyzed with statistical methods. In each layer, they selected the best features through statistical methods and forwarded them to the next layer. They did not use backpropagation to train their network end-to-end but used layer-by-layer least squares fitting where previous layers were independently fitted from later layers.

Description: This tutorial will teach you the main ideas of Unsupervised Feature Learning and Deep Learning. By working through it, you will also get to implement several feature learning/deep learning algorithms, get to see them work for yourself, and learn how to apply/adapt these ideas to new problems. This tutorial assumes a basic knowledge of machine learning (specifically, familiarity with the ideas of supervised learning, logistic regression, gradient descent). If you are not familiar with these ideas, we suggest you go to this Machine Learning course and complete sections II, III, IV (up to Logistic Regression) first.

Thanks to Michal Malohlava, Amy Wang, and Avni Wadhwa of H20.ai for providing the following guest post about building ML apps using Sparkling Water and Apache Spark on CDH. The Sparkling Water project is nearing its one-year anniversary, which means Michal Malohlava, our main contributor, has been very busy for the better part of this past year. The Sparkling Water project combines H2O machine-learning algorithms with the execution power of Apache Spark. This means that the project is heavily dependent on two of the fastest growing machine-learning open source projects out there. With every major release of Spark or H2O there are API changes and, less frequently, major data structure changes that affect Sparkling Water.

As a topic, AI's popularity tends to rise and fall. Some have even argued that it rises at times when there's a lull in tangible technological innovations, perhaps to give the geeks something to talk about. In the past, AI has often been a blank sheet onto which we project our fears and fantasies of the future. We get excited about artificial intelligence or we get fearful about it, yet those hopes or threats have always felt abstract and faraway. This time, though, it feels different.

BENGALURU: Mumbai-based Arya.ai offers its deep learning algorithms for developers to build intelligent AI systems that can adapt and do multiple things with minimal inputs from humans. From creating a diagnostic assistant for radiologists to a mathematical assistant for science academicians and on to drone image processing abilities, the uses appear to be really diverse. "We have already launched the advanced AI software tools in a closed group beta phase with developers internationally and researchers in select software companies, these developers are using these softwares for building robots that can assist professionals from different fields in their task," said Vinay Sankarapu, cofounder of Arya.ai (in picture). API for developers can be used for four specific categories. From creating custom APIs to use cases within computer vision, this could range from classifying or searching for products on e-commerce platforms by using visual inputs to security based face matching techniques, as well as language and reasoning, where event prediction can take place.

Research at Google is at the forefront of innovation in Machine Intelligence, with active research exploring virtually all aspects of machine learning, including deep learning and more classical algorithms. Exploring theory as well as application, much of our work on language, speech, translation, visual processing, ranking and prediction relies on Machine Intelligence. In all of those tasks and many others, we gather large volumes of direct or indirect evidence of relationships of interest, applying learning algorithms to understand and generalize. Machine Intelligence at Google raises deep scientific and engineering challenges, allowing us to contribute to the broader academic research community through technical talks and publications in major conferences and journals. Contrary to much of current theory and practice, the statistics of the data we observe shifts rapidly, the features of interest change as well, and the volume of data often requires enormous computation capacity.

An AI research lab owned by Google is going to start meeting NHS patients as it looks to be more open and transparent about how it helps doctors and clinicians to do their jobs. DeepMind, as the startup is known, intends to hold four patient meet ups a year at Google's new London office in King's Cross, with the first "patient engagement forum" taking place in September in an event that will be streamed on YouTube, alongside a live Twitter Q&A. The patient meet ups are taking place because DeepMind wants to get the public on side as it looks to expand its relationship with the NHS. The events will give members of the public the opportunity to ask DeepMind staff about its NHS partnerships and to learn how the Google-owned company intends to improve their healthcare. Founded in 2011 by Demis Hassabis, Mustafa Suleyman and Shane Legg, DeepMind faced criticism from privacy campaigners and some patients after it emerged in May that it had access to millions of NHS patient records for a kidney monitoring project.

The amount of energy consumed by big data centers has always been a headache for tech companies. Keeping the servers cool as they crunch numbers is such a challenge that Facebook even built one of its facilities on the edge of the Arctic Circle. Well, Google has a different solution to this problem: putting its DeepMind artificial intelligence unit in charge and using AI to manage power usage in parts of its data centers. A 40 percent reduction in the amount of electricity needed for cooling, which Google describes as a "phenomenal step forward." After accounting for "electrical losses and other non-cooling inefficiencies," this 40 percent reduction translated into a 15 percent reduction in overall power saving, says Google.

For Google, the massive network of data centers that powers the web giant's operations run up a similarly massive energy tab. The company has been improving server farm efficiency for years, but it recently adopted a novel technique for trimming usage: letting the robots take control. In particular, Google is giving the reins to an artificial intelligence system developed by its subsidiary DeepMind. The AI overlord succeeded in shaving several percentage points off of data center energy consumption, Bloomberg reported. This led to a 15 percent improvement in power-usage efficiency, the metric of how much power goes to the actual computing as opposed to auxiliary services at the data centers.

Thanks to an abundance of digital data, and powerful GPUs, we are now capable of teaching computers to read, see, and hear. Just this year, a handful of high-profile experiments came into the spotlight, including Microsoft Tay, Google DeepMind AlphaGo, and Facebook M. These experiments all relied on a technique known as deep learning, which attempts to mimic the layers of neurons in the brain's neocortex. This idea – to create an artificial neural network by simulating how the neocortex works – has been around since the 1980s. During the training process, the algorithm learns to discover useful patterns in the digital representation of data, like sounds and images.