Ask a member of Facebook's growth team what feature played the biggest role in getting the company to a billion daily users, and they'll likely tell you it was photos.

Model ensembling is a very powerful technique to increase accuracy on a variety of ML tasks. In this article I will share my ensembling approaches for Kaggle Competitions. For the first part we look at creating ensembles from submission files. The second part will look at creating ensembles through stacked generalization/blending. I answer why ensembling reduces the generalization error. Finally I show different methods of ensembling, together with their results and code to try it out for yourself. This is how you win ML competitions: you take other peoples' work and ensemble them together." The most basic and convenient way to ensemble is to ensemble Kaggle submission CSV files. You only need the predictions on the test set for these methods -- no need to retrain a model. This makes it a quick way to ensemble already existing model predictions, ideal when teaming up. Let's see why model ensembling reduces error rate and why it works better to ensemble low-correlated model predictions. During space missions it is very important that all signals are correctly relayed. A coding solution was found in error correcting codes. The simplest error correcting code is a repetition-code: Relay the signal multiple times in equally sized chunks and have a majority vote. Signal corruption is a very rare occurrence and often occur in small bursts. So then it figures that it is even rarer to have a corrupted majority vote. As long as the corruption is not completely unpredictable (has a 50% chance of occurring) then signals can be repaired. Suppose we have a test set of 10 samples. The ground truth is all positive ("1?):

The top 10 deep learning projects on Github include a number of libraries, frameworks, and education resources. Have a look at the tools others are using, and the resources they are learning from.

The post links to 6 articles covering a number of related topics.

The story of how the FBI finally tracked down notorious fugitive Lynn Cozart, using its brand-new, 1 billion facial recognition system, seems tailor-made to disarm even the staunchest of skeptics. Cozart, a former security guard in Beaver County, Pennsylvania, was convicted of deviant sexual intercourse in 1996. According to court filings, he had molested his three juvenile children, two girls and one boy, from 1984 through 1994. It wasn't until May 11, 1995, that the children's mother came forward and told the Pennsylvania State Police what Cozart had been doing. He was convicted, but he failed to show up for his sentencing hearing in April 1996. Federal agents raided his home, interviewed family members and released photos of the man to the general public. In August 2006, the Cozart case was featured in "America's Most Wanted," the national television program, under a segment titled "Ten Years of Hell for Three Children."

Google's Cloud Speech API will allow developers to convert audio to text within their own apps. The offering from Google will bring its neural network smarts to apps large and small, and opens up a wide range of interesting new possibilities. It also brings the fight to Nuance Communications' front door. Google is providing access to the limited preview of the Cloud Speech API through its developer website. Developers can take advantage of the API for free, for now, though presumably Google will start charging for access at some point.

Lawrence Livermore National Laboratory (LLNL) has purchased a new brain-inspired supercomputing platform developed by International Business Machines Corp (NYSE:IBM). Based on a breakthrough neurosynaptic computer chip called IBM TrueNorth, the scalable platform will process the equivalent of 16 million neurons and 4 billion synapses while consuming only the energy equivalent of a tablet computer. The brain-like, neural network design of the IBM neuromorphic system is able to run complex cognitive tasks such as pattern recognition and integrated sensory processing far more efficiently than conventional chips. "The delivery of this advanced computing platform represents a major milestone as we enter the next era of cognitive computing," said Dharmendra Modha, IBM fellow and chief scientist of Brain-inspired Computing, IBM Research. "We value our partnerships with the national labs. In fact, prior to design and fabrication, we simulated the IBM TrueNorth processor using LLNL's Sequoia supercomputer. This collaboration will push the boundaries of brain-inspired computing to enable future systems that deliver unprecedented capability and throughput, while minimizing the capital, operating and programming costs – keeping our nation at the leading edge of science and technology."

Machine learning, the application of artificial intelligence in computerized processes, has gained a foothold in data operations in the financial services industry. The way machine learning is set up to analyze and work with data to yield insights, analysis and decision support is still developing, as executives from several service providers and machine-learning users tell Inside Reference Data....

The rise of the robot writers is here. Computers are replacing human writers without you realising it. Hal just took my job. The Associated Press (AP) already uses them. These robot journalists don't need insurance or a pension, pay or annual leave – or even sleep. This lets publications react to unforeseen events such as earthquakes, in real time, without stirring a human hack from their slumbers.

A sci-fi UFO landscape, styled after Van Gogh? Ever felt like adding some Van Gogh-style swirls to your photos? A Russian project called Ostagram allows Internet users to combine photos and works of art to create fantastical images that wouldn't be out of place in the world of Alice in Wonderland or in a sci-fi space opera. The Ostagram project, created by user Sergey Morugin, is a web service that uses a computer algorithm to combine the content of one image with the style of another image using convolutional neural networks. This means you can get a photo of your dog to look like a Monet painting, if you pick the dog pic as the source for content, and the Monet artwork as a source for style.