Sure, there are lots of tutorials and overviews on gaining the insight you need into picking up machine learning, but many (most?) of them take the long view: get a foundation first, learn the basics next, then learn a bit of complementary theory before getting too far ahead of yourself in practical terms, take a step back, try your hand at a few examples, undertake a project on your own... This is all great advice, and a great approach to learning... well, almost anything. But let's say you're not starting from scratch. Or you don't have the patience to go through all of the motions. Let's say you want to hit the ground running and scramble under pressure to learn everything right now.

We propose an algorithm to separate simultaneously speaking persons from each other, the "cocktail party problem", using a single microphone. Our approach involves a deep recurrent neural networks regression to a vector space that is descriptive of independent speakers. Such a vector space can embed empirically determined speaker characteristics and is optimized by distinguishing between speaker masks. We call this technique source-contrastive estimation. The methodology is inspired by negative sampling, which has seen success in natural language processing, where an embedding is learned by correlating and de-correlating a given input vector with output weights. Although the matrix determined by the output weights is dependent on a set of known speakers, we only use the input vectors during inference. Doing so will ensure that source separation is explicitly speaker-independent. Our approach is similar to recent deep neural network clustering and permutation-invariant training research; we use weighted spectral features and masks to augment individual speaker frequencies while filtering out other speakers. We avoid, however, the severe computational burden of other approaches with our technique. Furthermore, by training a vector space rather than combinations of different speakers or differences thereof, we avoid the so-called permutation problem during training. Our algorithm offers an intuitive, computationally efficient response to the cocktail party problem, and most importantly boasts better empirical performance than other current techniques.

We present a data mining approach for reducing the search space of local search algorithms in a class of binary integer programs including the set covering and partitioning problems. The quality of locally optimal solutions typically improves if a larger neighborhood is used, while the computation time of searching the neighborhood increases exponentially. To overcome this, we extract variable associations from the instance to be solved in order to identify promising pairs of flipping variables in the neighborhood search. Based on this, we develop a 4-flip neighborhood local search algorithm that incorporates an efficient incremental evaluation of solutions and an adaptive control of penalty weights. Computational results show that the proposed method improves the performance of the local search algorithm for large-scale set covering and partitioning problems.

We introduce a method to stabilize Generative Adversarial Networks (GANs) by defining the generator objective with respect to an unrolled optimization of the discriminator. This allows training to be adjusted between using the optimal discriminator in the generator's objective, which is ideal but infeasible in practice, and using the current value of the discriminator, which is often unstable and leads to poor solutions. We show how this technique solves the common problem of mode collapse, stabilizes training of GANs with complex recurrent generators, and increases diversity and coverage of the data distribution by the generator.

Support vector machines (SVMs) are an important tool in modern data analysis. Traditionally, support vector machines have been fitted via quadratic programming, either using purpose-built or off-the-shelf algorithms. We present an alternative approach to SVM fitting via the majorization--minimization (MM) paradigm. Algorithms that are derived via MM algorithm constructions can be shown to monotonically decrease their objectives at each iteration, as well as be globally convergent to stationary points. We demonstrate the construction of iteratively-reweighted least-squares (IRLS) algorithms, via the MM paradigm, for SVM risk minimization problems involving the hinge, least-square, squared-hinge, and logistic losses, and 1-norm, 2-norm, and elastic net penalizations. Successful implementations of our algorithms are presented via some numerical examples.

It is a truth universally acknowledged that a single chatbot, powered by artificial intelligence, will inevitably learn from human users to be a sexist pig. Microsoft learned this lesson the hard way in 2016 when it released the Twitter chatbot known as Tay. Within hours, Tay was tweeting sentences like "Zoe Quinn is a Stupid Whore" and "I fucking hate feminists and they should all die and burn in hell." A recent study published in the journal Science, conducted by researchers at Princeton University and the University of Bath, found machine learning inherently "absorbs stereotyped biases" from human internet users. That includes racial bias, as proved by Google search algorithms' propensity to prioritize images of white women and babies over people of color.

No sooner had I returned from Los Angeles and the Vision AR/VR Summit than I was back on a plane to fly to Malta. Along with Oculus, I was there to present a session discussing AiSolve's ground-breaking medical VR simulator at the Children's Hospital Los Angeles at the eHealth Week event for health professionals. The project has really captured people's attention. Facebook featured it at the company's recent F8 annual get together and the collaboration with Bioflight, the Hollywood-based VFX specialist, and Children's Hospital Los Angeles has received extensive media coverage around the world. The project is a breakthrough in intelligent and responsive VR training for emergency paediatric trauma situations and has been supported by social media giant Facebook and uses an Oculus Rift headset and Touch controllers.

My dad got me my first computer for Christmas when I was 11 years old. It was a Texas Instruments 99-4a. He must have gotten it second hand because it had no box or any form of storage like a tape drive. I had been obsessed with the movie War Games, so my first program on the TI was designed to replicate the computer in the movie which uttered the famous line "shall we play a game." My program gave static responses to a huge variety of programmed questions, and mostly followed the script from the movie.

On May 11, 1997, a computer showed that it could outclass a human in that most human of pursuits: playing a game. The human was World Chess Champion Garry Kasparov, and the computer was IBM's Deep Blue, which had begun life at Carnegie Mellon University as a system called ChipTest. One of Deep Blue's creators, Murray Campbell, talked to the IDG News Service about the other things computers have learned to do as well as, or better than, humans, and what that means for our future. What follows is an edited version of that conversation. IDGNS: Is it true that you and Deep Blue joined IBM at the same time?

In 2014, Google fired a shot heard all the way to Detroit. Google's newest driverless car prototype had no steering wheel and no brakes. The message was clear: Cars of the future will be born fully autonomous, with no human driver needed or desired. Even more jarring, rather than retrofit a Prius or a Lexus as Google did to build its previous two generations of driverless cars, the company custom-built the body of its youngest driverless car with a team of subcontracted automotive suppliers. Best of all, the car emerged from the womb already an expert driver, with roughly 700,000 miles of experience culled from the brains of previous prototypes. Now that Google's self-driving cars have had another few more years of practice, the fleet's collective drive-time equals more than 1.3 million miles, the equivalent of a human logging 15,000 miles a year behind the wheel for 90 years. In response, car companies are pouring billions of dollars into software development and the epicenter of automotive innovation has moved from Detroit to Silicon Valley.