The past 30 years have seen incredible growth and innovation in the tech industry. We've gone from pocket calculators and PCs to pocket computers more powerful than the mammoth mainframe computers of the 1980s. The Atari 800XL computer I used in high school to develop games was powered by a microprocessor with 3,500 transistors – the one running my iPhone today has 2 billion transistors. The cost of a gigabyte of storage was in the range of 100,000 and the size of a refrigerator. Today, it's basically free and size is measured in millimetres.

This article studies the achievable guarantees on the error rates of certain learning algorithms, with particular focus on refining logarithmic factors. Many of the results are based on a general technique for obtaining bounds on the error rates of sample-consistent classifiers with monotonic error regions, in the realizable case. We prove bounds of this type expressed in terms of either the VC dimension or the sample compression size. This general technique also enables us to derive several new bounds on the error rates of general sample-consistent learning algorithms, as well as refined bounds on the label complexity of the CAL active learning algorithm. Additionally, we establish a simple necessary and sufficient condition for the existence of a distribution-free bound on the error rates of all sample-consistent learning rules, converging at a rate inversely proportional to the sample size. We also study learning in the presence of classification noise, deriving a new excess error rate guarantee for general VC classes under Tsybakov's noise condition, and establishing a simple and general necessary and sufficient condition for the minimax excess risk under bounded noise to converge at a rate inversely proportional to the sample size.

Generalization error bounds are probabilistically valid, non-asymptotic tools for characterizing the predictive ability of forecasting models. This methodology is fundamentally about choosing particular prediction functions out of some class of plausible alternatives so that, with high reliability, the resulting predictions will be nearly as accurate as possible ("probably approximately correct"). While many of these results are aimed at classification problems with independent and identically distributed (i.i.d.) data, this paper adapts and extends these methods to time-series models, so that economic and financial forecasting techniques can be evaluated rigorously. In particular, these methods control the expected accuracy of future predictions from mis-specified models based on finite samples. This allows for immediate model comparisons which neither appeal to asymptotics nor make strong assumptions about the data-generating process, in stark contrast to such popular model-selection tools as AIC.

If you've watched enough reruns of shows like CSI, Bones, and Law and Order, you probably know by now that when forensic specialists find DNA evidence, the suspect is often identified within the next couple of minutes--as soon as the team sticks the results of DNA analysis into a computer program. Although the real life process isn't quite as speedy, DNA certainly has been the highest bar for identification in forensics. But when it comes to hair samples of missing persons or those found at crime scenes, sequencing the proteins in those locks may work better than DNA. In a study published in the journal PLOS One on September 7, researchers at Lawrence Livermore National Laboratory in California demonstrated a method of extracting genetic information from proteins found in hair that is remarkably reliable. "Currently forensic science is very dependent on DNA," says primary author Glendon Parker, a biochemist at Livermore.

Ed. note: This is the latest installment in a series of posts from Lateral Link's team of expert contributors. Michael Allen is Managing Principal at Lateral Link, focusing exclusively on partner placements with Am Law 200 clients and placements for in-house attorneys. There's a new attorney named ROSS in BakerHostetler's bankruptcy practice and it doesn't eat, sleep, or complain about bonuses. Back in May, the Texas firm announced it would be the first to integrate artificial intelligence into its practice. Since then, Latham has entered into the fray, along with the Milwaukee-based Von Briesen & Roper.

There's no single book that can help you fully master machine learning. It's a complicated subject that spans many topics, purposes, and of course benefits in real-world applications. But this post should help novices and experts alike find the right book to continue their education. With so many resources available it can be tough knowing where to start. I'll consider 10 books and look at each one's teaching style, subject matter, and recency of publication.

In this Q&A Ari explains why the use of machine learning for automation is unlikely to overtake most organizational roles in the near future, why data is so incredibly important, the continucontinuous need for data scientists and his priorities for 2016.

Ford has agreed to acquire Chariot, an on-demand shuttle service based in San Francisco. This morning, the automaker said it would expand Chariot's shuttle service beyond San Francisco to at least five more markets in the next 18 months. At the moment, Chariot operates 100 Ford Transit shuttles along 28 routes in the Bay Area. It chooses routes by collecting votes from potential riders, but Ford also said it will move to a more complex algorithm in the future. The goal of a service like this is to fill the gap between taxi and bus services, which are often woefully inadequate for point-to-point transportation in some cities.

Concerned about potential regulatory fines and skyrocketing costs, the world's largest banks are turning to artificial intelligence to improve their compliance with know-your-customer and anti-money laundering regulations. "The value proposition for AI solutions is highest for large banks with significant volumes, complexity, multiple lines of business and geographical reach as these banks are affected most by the current challenges and stand to benefit the most by adopting new and innovative solutions," write Arin Ray and Neil Katkov, analysts with Celent in a new research report entitled "Artificial Intelligence in KYC-AML: Enabling the Next Level of Operational Efficiency." The analysts predict that global tier-one and large regional banks will be early adopters of AI over the next three years. Ray and Katkov's conclusions match the findings of a survey of 424 executives from financial services and fintech companies released in March by Chicago-based law firm Baker McKenzie. The firm found that 29 percent are thinking about using AI in know-your customer and anti-money laundering monitoring.

They say you should never judge a book by its cover, but this cliché may not apply when it comes to making decision about friends. Researchers have found that individuals feel it is justified to ostracize others from a group if they appear cold and incompetent - a look deemed'resting b*tch face'. These findings suggest that people with this appearance are perceived as troublemakers or selfish and need to be excluded in order to restore harmony and cohesion in the group. A study has found individuals feel it is justified to ostracize others from a group if they appear cold and incompetent - a look deemed'resting b*tch face'. The University of Basel reveals that humans really do judge a book by its cover.