Despite their attractiveness, popular perception is that techniques for nonparametric function approximation do not scale to streaming data due to an intractable growth in the amount of storage they require. To solve this problem in a memory-affordable way, we propose an online technique based on functional stochastic gradient descent in tandem with supervised sparsification based on greedy function subspace projections. The method, called parsimonious online learning with kernels (POLK), provides a controllable tradeoff? between its solution accuracy and the amount of memory it requires. We derive conditions under which the generated function sequence converges almost surely to the optimal function, and we establish that the memory requirement remains finite. We evaluate POLK for kernel multi-class logistic regression and kernel hinge-loss classification on three canonical data sets: a synthetic Gaussian mixture model, the MNIST hand-written digits, and the Brodatz texture database. On all three tasks, we observe a favorable tradeoff of objective function evaluation, classification performance, and complexity of the nonparametric regressor extracted the proposed method.

Brain tests at the age of three appear to predict a child's future chance of success in life, say researchers. Low cognitive test scores for skills like language indicate less developed brains, possibly caused by too little stimulation in early life, they say. These youngsters are more likely to become criminals, dependent on welfare or chronically ill unless they are given support later on, they add. Their study in New Zealand appears in the journal, Nature Human Behaviour. The US researchers from Duke University say the findings highlight the importance of early life experiences and interventions to support vulnerable youngsters.

2048 is an engaging single-player, nondeterministic video puzzle game, which, thanks to the simple rules and hard-to-master gameplay, has gained massive popularity in recent years. As 2048 can be conveniently embedded into the discrete-state Markov decision processes framework, we treat it as a testbed for evaluating existing and new methods in reinforcement learning. With the aim to develop a strong 2048 playing program, we employ temporal difference learning with systematic n-tuple networks. We show that this basic method can be significantly improved with temporal coherence learning, multi-stage function approximator with weight promotion, carousel shaping, and redundant encoding. In addition, we demonstrate how to take advantage of the characteristics of the n-tuple network, to improve the algorithmic effectiveness of the learning process by i) delaying the (decayed) update and applying lock-free optimistic parallelism to effortlessly make advantage of multiple CPU cores. This way, we were able to develop the best known 2048 playing program to date, which confirms the effectiveness of the introduced methods for discrete-state Markov decision problems.

Every day, K–12 educators juggle a bevy of tasks: teaching, developing lesson plans, grading papers, writing tests. And often, they battle with technology to do all of these things. With its more streamlined G Suite for Education -- formerly known as Apps for Education -- Google has updated its commonly used applications, hoping to save teachers a bit of time and energy with their everyday tasks. As announced in a blog post by Jonathan Rochelle, the director of product management at Google, G Suite's goal was to harness the intelligence of computers to create a smarter, easier and more efficient technology experience for educators and students. "G Suite for Education is the same set of apps that you know and love -- Gmail, Docs, Drive, Calendar, Hangouts, and more -- but designed with new intelligent features that make work easier and bring teachers and students together," writes Rochelle.

UPDATE: Mar 20, 2016 - Added my new follow-up course on Deep Learning, which covers ways to speed up and improve vanilla backpropagation: momentum and Nesterov momentum, adaptive learning rate algorithms like AdaGrad and RMSProp, utilizing the GPU on AWS EC2, and stochastic batch gradient descent. We look at TensorFlow and Theano starting from the basics - variables, functions, expressions, and simple optimizations - from there, building a neural network seems simple! Deep learning is all the rage these days. What exactly is deep learning? Well, it all boils down to neural networks.

In this post you discovered artificial neural networks for machine learning. How neural networks are not models of the brain but are instead computational models for solving complex machine learning problems. That neural networks are comprised of neurons that have weights and activation functions. The networks are organized into layers of neurons and are trained using stochastic gradient descent. That it is a good idea to prepare your data before training a neural network model.

Since the 1970s, when automated teller machines arrived, the number of bank tellers in America has more than doubled. James Bessen, an economist who teaches at Boston University School of Law, points to that seeming paradox amid new concerns that automation is "stealing" human jobs. To the contrary, he says, jobs and automation often grow hand in hand. Sometimes, of course, machines really do replace humans, as in agriculture and manufacturing, says Massachusetts Institute of Technology labor economist David Autor in a succinct and illuminating TED talk, which could have served as the headline for this column. Across an entire economy, however, Dr. Autor says that's never happened.

No discussion of ML would be complete without at least mentioning neural networks. Not only do neural nets offer an extremely powerful tool to solve very tough problems, but they also offer fascinating hints at the workings of our own brains, and intriguing possibilities for one day creating truly intelligent machines. Neural networks are well suited to machine learning problems where the number of inputs is gigantic. The computational cost of handling such a problem is just too overwhelming for the types of systems we've discussed above. As it turns out, however, neural networks can be effectively tuned using techniques that are strikingly similar to gradient descent in principle. A thorough discussion of neural networks is beyond the scope of this tutorial, but I recommend checking out our previous post on the subject.

Ignoring everything they've seen throughout the Terminator franchise, a group of Japanese researchers have come up with a computer that writes short stories… and it's actually produced a piece of work that got through the first round of a literary competition. Creative writing as a manufactured commodity – that's a scary thought, but perhaps it's inevitable. Companies love automation: feed a few instructions in one end and get a finished product out of the other. It's always been a popular notion that there are only a handful of stories and that everything written is a variation on those; if that's true then why can't a machine just write a half decent story? And does it even have to be half-decent to sell by the thousands?

Machine Learning is one of the most exciting fields in the world. Every week we discover something new, something amazing, something revolutionary. It's incredible, but it can also be overwhelming. That's why we created This Week in Machine Learning! Each week we publish a curated list of Machine Learning stories as a resource to help you keep pace with all these exciting developments.