Instructor: Eric Grimson In this lecture, Prof. Grimson introduces machine learning and shows examples of supervised learning using feature vectors.

According to experts in the field, the brain is considered to be one of the most complex systems in existence. While significant headway has been made to understand how the brain works, researchers tend to generate more questions than answers about this entity. However, the creators of the machine learning model - a team from Kyoto University - believe it has the potential to explain the difference in neuronal computation in different brain regions more clearly. To comprehend the brain, neurologists must look at the neurons that construct it. Our entire world of perception runs across these billions of cells in our head and that is compounded by the exponentially larger number of connections – known as synapses – between them.

Tesla's Autopilot software received an upgrade, making it possible for its electric vehicles to maneuver around construction cones. According to media reports, Tesla has been issuing upgrades to its software regularly during the past few years. The software is powered by sensors placed on the car and its computer vision system that shows the vehicle surroundings on the screen. In the summer it added the ability to zoom in and out and get a 360-degree view. In the latest addition, Tesla included the ability to detect, visualize and maneuver around traffic cones as seen in this video uploaded to Twitter. When releasing the update, Tesla wrote: "in cases where a traffic cone is detected and Navigate on Autopilot is engaged (requires Full Self Driving Capability), the vehicle is designed to suggest a lane change (or attempt a lane if REQUIRE LANE CHANGE CONFIRMATION is set to NO) to avoid cones.

AI and machine learning algorithms are becoming increasingly good at predicting next actions in videos. The very best can anticipate fairly accurately where a baseball might travel after it's been pitched, or the appearance of a road miles from a starting position. To this end, a novel approach proposed by researchers at Google, the University of Michigan, and Adobe advances the state of the art with large-scale models that generate high-quality videos from only a few frames. All the more impressive, it does so without relying on techniques like optical flows (the pattern of apparent motion of objects, surfaces or edges in a scene) or landmarks, unlike previous methods. "In this work, we investigate whether we can achieve high quality video predictions … by just maximizing the capacity of a standard neural network," wrote the researchers in a preprint paper describing their work.

Washington, DC (November 7, 2019) -- Machine learning, a form of artificial intelligence, may help improve care for patients with kidney failure. The findings come from a study that will be presented at ASN Kidney Week 2019 November 5–November 10 at the Walter E. Washington Convention Center in Washington, DC. For the study, Ollie Fielding (pulseData, in New York) and his colleagues deployed a machine learning model to identify patients at risk of requiring kidney replacement therapy, such as dialysis or kidney transplantation. An electronic health record database of 110,998 patients was used to create a machine learning model to predict progression to kidney failure. The system calculates weekly risk scores for patients, and for those with high risk scores, an alert is sent so that treatment discussions can be made by a multidisciplinary team of clinicians.

A bill that U.S. Sen. Cory Gardner has co-sponsored to develop and guide the use of artificial intelligence in the federal government passed out of committee on Wednesday. "Our bill will bring agencies, industry, and others to the table to discuss government adoption of artificial intelligence and emerging technologies," Gardner said in a statement. The AI in Government Act defines artificial intelligence as any type of computer programming that would enable the computer to carry out tasks of the sort that "would require intelligence if performed by a human." The bill would create an AI Center of Excellence within the General Services Administration, which would coordinate AI use in the public interest and house the government's technical expertise. The center's responsibility would also include analyzing the ethical and civil liberties implications of artificial intelligence, helping state and local governments as needed.

ISLAMABAD (Web Desk) Aqsa Kausar, an electrical engineering graduate from NUST (National University of Science and Technology), has become the first female Google Developer Expert in Machine language from Pakistan. According to local news agency, she has risen to acclaim at such a young age through her contributions to the field of Machine Learning. Besides, there are other various awards to her credit for holding workshops in events like Google DevFest 2018 and Google Cloud Next Extended 2019. Moreover, not too long ago, she also participated in Google's Machine Learning Train-The-Trainer session which was held in Singapore. Aqsa currently working as an AI developer with a software organization named Red Buffer.

These are two examples of how NASA hopes to use artificial intelligence. As far-fetched as the concept sounds, the agency is already using AI in missions on both Earth and Mars. And there are other missions in the works that could see AI exploring icy moons in search of life. This bot-friendly future stands counter to some of the fuss in the press this past week, after Facebook shut down an experiment because two artificially intelligent bots began communicating in a shorthand language instead of English. Many in the media portrayed the bots as coming up with their own language.

Combining data from varied sources has considerable potential for knowledge discovery: collaborating data parties can mine data in an expanded feature space, allowing them to explore a larger range of scientific questions. However, data sharing among different parties is highly restricted by legal conditions, ethical concerns, and / or data volume. Fueled by these concerns, the fields of cryptography and distributed learning have made great progress towards privacy-preserving and distributed data mining. However, practical implementations have been hampered by the limited scope or computational complexity of these methods. In this paper, we greatly extend the range of analyses available for vertically partitioned data, i.e., data collected by separate parties with different features on the same subjects. To this end, we present a novel approach for privacy-preserving generalized linear models, a fundamental and powerful framework underlying many prediction and classification procedures. We base our method on a distributed block coordinate descent algorithm to obtain parameter estimates, and we develop an extension to compute accurate standard errors without additional communication cost. We critically evaluate the information transfer for semi-honest collaborators and show that our protocol is secure against data reconstruction. Through both simulated and real-world examples we illustrate the functionality of our proposed algorithm. Without leaking information, our method performs as well on vertically partitioned data as existing methods on combined data -- all within mere minutes of computation time. We conclude that our method is a viable approach for vertically partitioned data analysis with a wide range of real-world applications.

We introduce a framework for studying how distributional assumptions on the process by which data is partitioned into a training and test set can be leveraged to provide accurate estimation or learning algorithms, even for worst-case datasets. We consider a setting of $n$ datapoints, $x_1,\ldots,x_n$, together with a specified distribution, $P$, over partitions of these datapoints into a training set, test set, and irrelevant set. An algorithm takes as input a description of $P$ (or sample access), the indices of the test and training sets, and the datapoints in the training set, and returns a model or estimate that will be evaluated on the datapoints in the test set. We evaluate an algorithm in terms of its worst-case expected performance: the expected performance over potential test/training sets, for worst-case datapoints, $x_1,\ldots,x_n.$ This framework is a departure from more typical distributional assumptions on the datapoints (e.g. that data is drawn independently, or according to an exchangeable process), and can model a number of natural data collection processes, including processes with dependencies such as "snowball sampling" and "chain sampling", and settings where test and training sets satisfy chronological constraints (e.g. the test instances were observed after the training instances). Within this framework, we consider the setting where datapoints are bounded real numbers, and the goal is to estimate the mean of the test set. We give an efficient algorithm that returns a weighted combination of the training set---whose weights depend on the distribution, $P$, and on the training and test set indices---and show that the worst-case expected error achieved by this algorithm is at most a multiplicative $\pi/2$ factor worse than the optimal of such algorithms. The algorithm, and its proof, leverage a surprising connection to the Grothendieck problem.