Deep learning has become the method of choice in many application domains of machine learning in recent years, especially for multi-class classification tasks. The most common loss function used in this context is the cross-entropy loss, which reduces to the log loss in the typical case when there is a single correct response label. While this loss is insensitive to the identity of the assigned class in the case of misclassification, in practice it is often the case that some errors may be more detrimental than others. Here we present the bilinear-loss (and related log-bilinear-loss) which differentially penalizes the different wrong assignments of the model. We thoroughly test this method using standard models and benchmark image datasets. As one application, we show the ability of this method to better contain error within the correct super-class, in the hierarchically labeled CIFAR100 dataset, without affecting the overall performance of the classifier.

In applications of graphical models, we typically have more information than just the samples themselves. A prime example is the estimation of brain connectivity networks based on fMRI data, where in addition to the samples themselves, the spatial positions of the measurements are readily available. With particular regard for this application, we are thus interested in ways to incorporate additional knowledge most effectively into graph estimation. Our approach to this is to make neighborhood selection receptive to additional knowledge by strengthening the role of the tuning parameters. We demonstrate that this concept (i) can improve reproducibility, (ii) is computationally convenient and efficient, and (iii) carries a lucid Bayesian interpretation. We specifically show that the approach provides effective estimations of brain connectivity graphs from fMRI data. However, providing a general scheme for the inclusion of additional knowledge, our concept is expected to have applications in a wide range of domains.

The analysis in Part I revealed interesting properties for subgradient learning algorithms in the context of stochastic optimization when gradient noise is present. These algorithms are used when the risk functions are non-smooth and involve non-differentiable components. They have been long recognized as being slow converging methods. However, it was revealed in Part I that the rate of convergence becomes linear for stochastic optimization problems, with the error iterate converging at an exponential rate $\alpha^i$ to within an $O(\mu)-$neighborhood of the optimizer, for some $\alpha \in (0,1)$ and small step-size $\mu$. The conclusion was established under weaker assumptions than the prior literature and, moreover, several important problems (such as LASSO, SVM, and Total Variation) were shown to satisfy these weaker assumptions automatically (but not the previously used conditions from the literature). These results revealed that sub-gradient learning methods have more favorable behavior than originally thought when used to enable continuous adaptation and learning. The results of Part I were exclusive to single-agent adaptation. The purpose of the current Part II is to examine the implications of these discoveries when a collection of networked agents employs subgradient learning as their cooperative mechanism. The analysis will show that, despite the coupled dynamics that arises in a networked scenario, the agents are still able to attain linear convergence in the stochastic case; they are also able to reach agreement within $O(\mu)$ of the optimizer.

Researchers from The United Kingdom stated that a self-taught artificial intelligence machine could pave the way in predicting heart attacks better than doctors. The mentioned machine was said to possibly save thousand to millions of people if implemented. In which aside from heart attacks, blocked arteries and strokes were mentioned as well. Yet, thanks to the team, the future of predicting heart attacks better are on the way. The study was reported to be done by the University of Nottingham who created a bunch of programs that could predict heart attack better and train themselves to learn more. The AI machine included four machine learning algorithms namely: random forest, logistic regression, gradient boosting, and neural networks.

"Will artificial intelligence replace marketers in the near future?" This is the compelling question posted by Loren McDonald of IBM Watson Marketing during his presentation at the recent Digital Summit conference in Los Angeles. While many marketers might consider this a provocative presentation opener, there are some blunt realities marketers need to consider if they want to remain in the field and be competitive. Artificial Intelligence is about the development of computers systems that are able to perform tasks that would normally require human intelligences such as visual identification speech recognition, decision-making and translating between languages. AI performs a role in many of the stems that you use everyday from using Siri on your phone, a chatbot on an ecommerce site like Staples or 1-800-Flowers or every time you use Google.

Link this page Would you like to put a link to this lecture on your homepage?

Link this page Would you like to put a link to this lecture on your homepage?

Chinese PC maker Lenovo plans to pour over US$ 1.2 billion into artificial intelligence, Internet of Things and big data in the next four years, as part of its efforts to diversify their operations amid the stalled growth of its PC and smartphone business, local media is reporting (in Chinese). Lenovo CEO Yang Yuanqing said the annual investment in the above three areas will represent over one-fifth of the company's total annual R&D expenditure by March 2021. Lenovo remained the top PC vendor in the first quarter of 2017, garnering a 19.9% share in the global market by shipping 12.377 million units, IT research firm Gartner noted. Yet its rival HP has narrowed Lenovo's lead with shipments of 12.118 million. Among the company's three main lines of business, namely data centers, mobile devices, and PCs and smart devices (PCSD), revenue from PCSD business accounted for around 70% of its total revenue for the three months ended Dec. 31, 2016, according to the firm's Q3 FY 2016/17 results released this February.

Nearly three-fourths (74 percent) of game sales in the U.S. market now come from digital storefronts, the Entertainment Software Association's annual Essential Facts About the Computer and Video Game Industry showed Wednesday. The report noted the number of gamers who regularly bought games digitally has increased steadily. The percentage of gamers who bought digital content (defined as downloadable content, full versions of games, subscriptions, mobile apps and Facebook games) rose from 31 percent in 2010 to 74 percent in 2016. The news fits in with larger trends in the gaming industry as more gamers have taken advantage of increasing console storage capacity and digital storefronts to buy their games. This trend has also come in the face of declining game sales from traditional brick and mortar retailers.

The first was the steam engine-driven Industrial Revolution; the second involved the innovations from Henry Ford's assembly line. Third, microelectronics and computer power appeared on factory floors. Now, manufacturing businesses are beginning to integrate robotics, automation and other data-driven technologies into their workflows. Now, manufacturing businesses are beginning to integrate robotics, automation and other data-driven technologies into their workflows. Robots have taken over difficult, dangerous and repetitive physical tasks, improving factory safety, worker comfort and product quality.