Machine learning and deep learning projects are gaining more and more importance in most enterprises. The complete process includes data preparation, building an analytic model and deploying it to production. This is an insights-action-loop which improves the analytic models continuously. Forrester calls the complete process and the platform behind it the Insights Platform. A key task when you want to build an appropriate analytic model using machine learning or deep learning techniques, is the integration and preparation of data sets from various sources like files, databases, big data storage, sensors or social networks. This step can take up to 80 percent of the whole analytics project. This article compares different alternative techniques to prepare data, including extract-transform-load (ETL) batch processing, streaming ingestion and data wrangling.

It's official: driverless cars have hit the race tracks. Roborace, the autonomous race car maker, had its two self-driving'DevBots' compete against each other at the Formula E Buenos Aires ePrix. The race didn't go without its own surprises: One car had to dodge a random dog that ended up on the race track, and the other ended up hitting a barrier, unable to finish the race. Roborace's self-driving car races will take place at Formula E events throughout 2017. All cars competing will be made identically.

Fujitsu today announced that it has received RIKEN's order for the "Deep learning system," which in terms of operations will be one of the largest-scale supercomputers in Japan specializing in AI research. The RIKEN Center for Advanced Intelligence Project will use the new system, scheduled to go online in April 2017, as a platform to accelerate R&D into AI technology. The system's total theoretical processing performance will reach 4 petaflops(1). The system will be comprised of two server architectures, with 24 of NVIDIA DGX-1 servers and 32 FUJITSU Server PRIMERGY RX2530 M2 servers, along with a high-reliability, high-performance storage system. Fujitsu is leveraging the extensive know-how that it and Fujitsu Laboratories Ltd. have in high-performance computing development and AI research to build and operate one of Japan's most advanced AI research systems.

With huge strides in AI--from advances in the driverless vehicle realm, to mastering games such as poker and Go, to automating customer service interactions--this advanced technology is poised to revolutionize businesses. But the terms AI, machine learning, and deep learning are often used haphazardly and interchangeably, when there are key differences between each type of technology. Here's a guide to the differences between these three tools to help you master machine intelligence. SEE: Inside Amazon's clickworker platform: How half a million people are being paid pennies to train AI (PDF download) (TechRepublic) AI is the broadest way to think about advanced, computer intelligence. In 1956 at the Dartmouth Artificial Intelligence Conference, the technology was described as such: "Every aspect of learning or any other feature of intelligence can in principle be so precisely described that a machine can be made to simulate it."

Developers from Orange Labs in France developed a deep learning system that can quickly make young faces look older, and older faces look younger. A number of techniques already exist, but they are expensive and time consuming. Using CUDA, Tesla K40 GPUs and cuDNN for the deep learning work, they trained their neural network on 5,000 faces from each age group (0-18, 19- 29, 30-39, 40-49, 50-59, and 60 years old) taken from the Internet Movie Database and from Wikipedia and then labeled with the person's age -- this helped the system learn the characteristic signature of faces in each age group. A second neural network, called the face discriminator, looks at the synthetically aged face to see whether the original identity can still be picked out. If it can't, the image is rejected, which they call the process in their paper, Age Conditional Generative Adversarial Network.

Machine Intelligence (AI, ML and Deep Learning) requires a certain calibre of computer science talent. Today, this kind of talent is at the "top of the stack" of computer science. These cutting-edge capabilities used to be found at universities, and work on publicly-funded blue sky research; today, companies have the talent, and use it for private, applied purposes. The Economist touched on this in "Million Dollar Babies": That race for talent, is catalysing acquisitions by corporates interested in adding AI to their products and services. As an example, Ex-Googler Sebastian Thurn estimated that the going rate for self-driving engineering talent is $10 million per person.

It's hard to think of a job more important that determining whether or not a patient has cancer. Yet the magnitude of the task facing pathologists is so vast that agreement between different clinicians studying the same slides can be as low as 48%. There can be many slides per patient, each of which is 10 gigapixels when digitized at 40X magnification. Imagine having to go through a thousand 10 megapixel (MP) photos, and having to be responsible for every pixel. Needless to say, this is a lot of data to cover, and often time is limited.

Deep reinforcement learning methods attain super-human performance in a wide range of environments. Such methods are grossly inefficient, often taking orders of magnitudes more data than humans to achieve reasonable performance. We propose Neural Episodic Control: a deep reinforcement learning agent that is able to rapidly assimilate new experiences and act upon them. Our agent uses a semi-tabular representation of the value function: a buffer of past experience containing slowly changing state representations and rapidly updated estimates of the value function. We show across a wide range of environments that our agent learns significantly faster than other state-of-the-art, general purpose deep reinforcement learning agents.

Deep generative models have been wildly successful at learning coherent latent representations for continuous data such as video and audio. However, generative modeling of discrete data such as arithmetic expressions and molecular structures still poses significant challenges. Crucially, state-of-the-art methods often produce outputs that are not valid. We make the key observation that frequently, discrete data can be represented as a parse tree from a context-free grammar. We propose a variational autoencoder which encodes and decodes directly to and from these parse trees, ensuring the generated outputs are always valid. Surprisingly, we show that not only does our model more often generate valid outputs, it also learns a more coherent latent space in which nearby points decode to similar discrete outputs. We demonstrate the effectiveness of our learned models by showing their improved performance in Bayesian optimization for symbolic regression and molecular synthesis.

In this paper we propose a synergistic melting of neural networks and decision trees (DT) we call neural decision trees (NDT). NDT is an architecture a la decision tree where each splitting node is an independent multilayer perceptron allowing oblique decision functions or arbritrary nonlinear decision function if more than one layer is used. This way, each MLP can be seen as a node of the tree. We then show that with the weight sharing asumption among those units, we end up with a Hashing Neural Network (HNN) which is a multilayer perceptron with sigmoid activation function for the last layer as opposed to the standard softmax. The output units then jointly represent the probability to be in a particular region. The proposed framework allows for global optimization as opposed to greedy in DT and differentiability w.r.t. all parameters and the input, allowing easy integration in any learnable pipeline, for example after CNNs for computer vision tasks. We also demonstrate the modeling power of HNN allowing to learn union of disjoint regions for final clustering or classification making it more general and powerful than standard softmax MLP requiring linear separability thus reducing the need on the inner layer to perform complex data transformations. We finally show experiments for supervised, semi-suppervised and unsupervised tasks and compare results with standard DTs and MLPs.