Tag: MachineLearning

Commentary: Will crosswords cross up computers?
Dr. Fill is the creation of Matt Ginsberg, an artificial intelligence scientist and cruciverbalist, what you should fill in if you're ever starting at the clue: "a creator of crossword puzzles. " He set off on the project a little more than a year ago, in part because he felt Watson left the public with a false impression about the nature of artificial intelligence. That, plus improving artificial intelligence, lets us harness computers to solve increasingly complex problems. (more)

Sebastian Thrun Will Teach You How to Build Your Own Self-Driving Car, For Free
Tweet Last August, Sebastian Thrun, the brains behind Google s self-driving cars and one of the world s top AI experts, offered an online version of Stanford s Introduction to Artificial Intelligence course to absolutely anyone who wanted to take it, for free. It turned out to be just a little bit popular (over 150,000 students enrolled), and now Thrun is offering a new, totally free, seven-week online course called Programming a Robotic Car. You ll need to know a little bit of Python (the programming language, not the reptile), but it s an easy language to learn if you have any sort of programming background, and there are plenty of helpful course materials to check out. I don t know about you, but I m the suspicious type, and when something comes along that sounds too good to be true (like a totally free robotics course taught by some of the smartest guys in the world that s open for anyone to enroll in), I start wondering how they plan to pull that off. (more)

Is This Code Part of Majel, Android's Future Personal Assistant?
Someone tried to mess around with the FaceUnlock application code from the Android SDK, and stumbled upon this code, which seems to include answers, as if given by a personal assistant or such. Weve heard some rumors a few months back that Google may be working on a personal assistant as part of their Google X projects, and theyve been doing this for years. These rumors came from someone who claimed he was involved with Google X, and he wanted to give a hint about what theyve been working on: This is in total violation of the NDA, but I dont care anymore. The central focus of Google X for the past few years has been a highly advanced artificial intelligence robot that leverages the underlying technology of many popular Google programs. (more)

Controlling neuroprosthetics with your mind
Neuroscientists have found that the brain is more flexible and trainable than previously thought, opening the door to development of thought-controlled prosthetic devices to help people with spinal cord injuries, amputations and other impairments. The study advances work by researchers who have been studying the brain circuits used in natural movement, to mimic them for the development of prosthetic devices. The insights suggest that learning to control a BMI (brain-machine interface), which is inherently unnatural, may feel completely normal to a person, because this learning is using the brains existing built-in circuits for natural motor control. The rats were fitted with a brain-machine interface that converted EEG brain waves into auditory tones. (more)

The problem with investing based on pattern recognition
A famous story in artificial intelligence is how the US military developed algorithms to determine whether an image had a tank in it. They used a standard machine learning method: feed the computer a training set of photos, some of which had tanks in them and some of which didnt, and let algorithms identify which features in the photos correlated to tanks being shown. Since the features the computer identified were embedded in complicated mathematical equations, no one could figure out what it was really doing and therefore why it stopped working.Eventually someone realized that in the training set, all of the images with tanks were taken on a cloudy day, and all the images without tanks were taken on a sunny day. In the context of markets, it can cause companies and sectors with the right patterns to be overvalued, and ones with the wrong patterns to be undervalued. (more)

Skytree Uses Machine Learning To Crunch Big Data
Skytree is a machine learning program intended to serve as a replacement for SQL databases, and has the ability to take even unstructured sets of data and crunch it faster than traditional methods. The company claims that its learning algorithms can be used to process data for a wide variety of applications, from predicting the effectiveness of sales promotions to fraud detection to molecular modeling to astronomy. Which is no surprise, as algorithms to predict certain types of data are often easily portable to other types of data. The era of big data has shown that just throwing hardware at the problem is no longer working, said Martin Hack, co-founder and CEO of Skytree in the companys press release. (more)

Machine Learning for Hackers
Understanding machine learning and statistics tools through a series of hands-on case studies, instead of a traditional math-heavy presentation. Authors Drew Conway and John Myles White help you understand machine learning and statistics tools through a series of hands-on case studies, instead of a traditional math-heavy presentation. "We can see how many people are interested in learning about machine learning (ML), but don't have the mathematical background to read traditional treatments of the book," says White (@johnmyleswhite). "We wanted to get people interested in ML in a hands-on fashion in the way that chemistry sets can get children excited about chemistry before they have the scientific background to learn the subject rigorously. (more)

Have We Met? Tracing Face Blindness to Its Roots
Those with prosopagnosia, also known as face blindness, can see perfectly well, but their brains are unable to piece together the information needed to understand that a collection of features represents an individuals face. One of the keys to understanding face recognition, it seems, is understanding how the brain comes to recognize voices. But by testing for these two conditions simultaneously, researchers at the Max Planck Institute for Human Cognitive and Brain Sciences in Germany recently found evidence that face and voice recognition may be linked in a novel person-recognition system. The researchers found that regions of the brain already associated with facial recognition, like the fusiform face area in the occipital lobe, are directly linked to regions responsible for voice recognition, mostly in the temporal lobe. (more)

Tinkering with evolution: Ecological implications of modular software networks
Now, researchers in the Department of Ecology and Evolutionary Biology at Princeton University have shown the inverse namely, that network theory, when applied to software systems, provides surprising insights into biology, ecology and evolution. Specifically, they explored evolutionary behavior in complex systems by analyzing how the Debian GNU/Linux operating system utilizes modular code. While Fortuna notes that quantifying the increase of the codes modular structure time was the main insight of their study, he points out that reuse of code and softwares hierarchical structure were suggested by the pioneering work of Ricard V. The idea of using the network of dependencies and conflicts of different releases of the Debian operating system as a case study has facilitated the understanding of how code development evolves over time without the need to go deeper into the details of the code itself. (more)

Scientists Model Brain to Teach Computers to Recognize
Researchers at the beginning of the computer revolution assumed that teaching a computer to recognize something would be easy. Researchers from Los Alamos National Laboratory, Chatham University, and Emory University created a neural network with a slightly different structure than what is usually used in research. Most neural network research involves wiring artificial neurons from one layer to others in another layer. In this case, the researchers decided to wire some neurons to other neurons in the same layer, creating lateral connections. (more)

Neural network learns to identify group sizes without knowledge of numbers
After feeding the network 51,800 images, where each was a unique layout of rectangles of various sizes, the researchers found that the new images generated by the system began to demonstrate an awareness of the relative size of different groups without having to perform any counting. Next, to demonstrate, ANS, the team fed the system another program that allowed the system to compare different groups that it had seen during the first run and found, based on new images generated, that the system was able to make educated guesses about which was bigger or smaller; Teaching computer systems to learn to use ANS is but one step towards creating machines that think rather than simply crunch numbers for us, and the hope is that one day, such systems can be put into robots to make them as useful as those weve seen in movies for decades. (more)

Artificial intelligence: Getting better at the age guessing game
The active learning algorithm is faster and more accurate in guessing the age of an individual than conventional algorithms. They have, for example, developed computer algorithms for facial age classification -- the automated assignment of individuals to predefined age groups based on their facial features as seen on video captures or still images. A person can teach a computer to make better guesses by running its algorithm through a large database of facial images of which the age is known using sets of labeled images, but acquiring such a database can be both time-consuming and expensive. The technology could find use, for example, in digital signage where the machine determines the age group of the viewer and displays targeted advertisements designed for those age groups, or in interactive games where the machine automatically presents different games based on the players' age range. (more)

How does our brain see Jesus face on a tortilla?
Whether its New Hampshires erstwhile granite Old Man of the Mountain, or a face on Mars, our brains are adept at locating images that look like faces. nonface Sinha and his students set out to investigate how that brain region decides what is and is not a face, particularly in cases where an object greatly resembles a face. To help them do that, the researchers created a continuum of images ranging from those that look nothing like faces to genuine faces. They found images that very closely resemble faces by examining photographs that machine vision systems had falsely tagged as faces. (more)

Lifebrowser: Data mining gets (really) personal at Microsoft
Lifebrowser: Data mining gets (really) personal at Microsoft (PhysOrg.com) -- Microsoft Research is doing research on software that could bring you your own personal data mining center with a touch of Proust for returns. The software uses machine learning to help a user place life events, which may span months or years, to be expanded or contracted selectively, in better context. Lifebrowser's timeline shows items that the user can associate with "landmark" events with the use of artificial intelligence algorithms. By associating an email with a relevant calendar date with a relevant document and photos, significance is gleaned from personal life events. (more)

Do-It-Yourself Cat Door Recognizes Your Feline
The ideal cat door would let the cat in most of the time, but lock the door when kitty is carrying prey or when another animal, such as a raccoon, approaches the door.Wikimedia Commons What does it take for a computer program to recognize your cat? At the time, the door connected to a desktop computer that ran the program that snapped pictures of Flo and analyzed them as she approached the door. Now, Forster is determined to build Flo's door for Timothy on weekends, in between his usual consulting work in Simi Valley, Calif. But because computer vision is still difficult, even for scientists, people who are looking for guaranteed results, rather than the fun of putting the door together, should probably stay away from a do-it-yourself door for now. (more)

Rational Retention Announces Rational Intelligence Partnership to Improve ...
Rational Retention (RR), a leading developer of retention and e-discovery software, announces today that it has partnered with the academic team of New York Universitys Center for Health Informatics and Bioinformatics and with Discovery Holdings LLC (a technology holding company specializing in Markov Boundary, Causal Graph, Kernel Ridge Regression, Support Vector Machines, and other cutting-edge high-dimensional data classification methods) to dramatically expand and enhance its predictive coding and auto-classification platform in the e-discovery space. Rational Retention and the NYU team of experts will collaborate to deploy Discovery Holdings proprietary, best-in-class machine learning algorithms and analytic protocols to intelligently and automatically code documents in the context of litigation. Last year, Rational Retention announced the release of their hosted review platform, Rational eDiscovery (ReD), which aims to provide the most powerful and cost-effective e-discovery repository in the marketplace. (more)

Daniela Witten: Using artificial intelligence to study genomes
Daniela Witten, an assistant professor of biostatistics at the University of Washington in Seattle, is developing artificial intelligence programs to sort the slurry, helping researchers develop more personalized and effective treatments for cancer and other diseases. Witten is using statistical machine learning to discover the nuggets of gold in the ore of biological research. A statistical analysis of the 3 billion base pairs of DNA making up a cancer cell may be able to identify the pairs or combination of pairs responsible for certain characteristics of the cancer. Before I got to college, I was planning to study foreign languages, Witten said in an interview with the blog Simply Statistics. (more)

Robot reveals the inner workings of brain cells: Automated way to record electrical activity inside neurons in the living brain
Gaining access to the inner workings of a neuron in the living brain offers a wealth of useful information: its patterns of electrical activity, its shape, even a profile of which genes are turned on at a given moment. Using this technique, scientists could classify the thousands of different types of cells in the brain, map how they connect to each other, and figure out how diseased cells differ from normal cells. In all these cases, a molecular description of a cell that is integrated with [its] electrical and circuit properties has remained elusive, says Boyden, who is a member of MIT's Media Lab and McGovern Institute for Brain Research. Kodandaramaiah, Boyden and Forest set out to automate a 30-year-old technique known as whole-cell patch clamping, which involves bringing a tiny hollow glass pipette in contact with the cell membrane of a neuron, then opening up a small pore in the membrane to record the electrical activity within the cell. (more)

Building machines that see: Finding edges in images
Eric Gregori, BDTI 5/9/2012 4:11 PM EDT With the emergence of increasingly capable low-cost processors and image sensors, its becoming practical to incorporate computer vision capabilities into a wide range of embedded systems, enabling them to analyze their environments via image and video inputs. Then, after explaining edge detection from an algorithmic perspective, we show how to use OpenCV, a free open-source computer vision software component library, to quickly implement application software incorporating edge detection. If you graph the values of a single row of pixels containing an edge, the strength of the edge is shown by slope of the line in the graph. Grayscale soft edge Figure 2 above and Figure 3 below represent eight pixels from two different parts of a grayscale image. (more)

How will we build an artificial human brain?
follow io9.comThere's an ongoing debate among neuroscientists, cognitive scientists, and even philosophers as to whether or not we could ever construct or reverse engineer the human brain. Regardless, it's fair to say that ongoing breakthroughs in brain science are steadily paving the way to the day when an artificial brain can be constructed from scratch. And if we assume that cognitive functionalism holds true as a theory the idea that our brains are a kind of computer there are two very promising approaches worth pursuing. One side wants to build a brain with code, while the other wants to recreate all the brain's important functions by emulating it on a computer. (more)

Can Automated Deep Natural-language Analysis...
13:43 GMT, May 4, 2012 Program to assist warfighters with planning and decision-making by inferring implicit information in text, filtering redundancy and connecting like documents Making sense of large amounts of data is a challenge for military operations officers and intelligence analysts whose window for processing that data is often small. In reviewing captured documents, intelligence reports, news articles and related information, conservative estimates indicate that these personnel are unable to explore as much as 90% of the information available to them due to mission, time and resource constraints. Automated, deep natural-language understanding technology may hold a solution for more efficiently processing text information. If successful, DEFT will allow analysts to move from limited, linear processing of insurmountable quantities of data to a nuanced, strategic exploration of available information. (more)

Hey Graduates: Forget Plastics
For the first time, I began to think through the practical implications of what the advent of the information explosion (now commonly referred to as Big Data) really meant to both individuals and corporations. That spark nearly five years ago led to an investment thesis here at Flybridge, where we have looked to invest behind Big Data applications and uses across a number of vertical industries. Fortunately, we have been able to find some extraordinarily talented entrepreneurs who have thought deeply about this issue of Big Datas impact on vertical industries. The company uses abig data approach to determine what the particular advertisement should be for the particular user at that particular moment across all digital channels - display, mobile, video and social. (more)

David L. Waltz, Computer Science Pioneer, Dies at 68
During his career as a teacher and a technologist at start-up companies as well as large corporate laboratories, Dr. Waltz made fundamental contributions to computer science in areas ranging from computer vision to machine learning. Dr. Waltz graduated in 1972, then taught computer science at the University of Illinois at Urbana-Champaign and, later, at Brandeis University in Massachusetts. Thinking Machines was an early maker of massive, parallel supercomputers, and by joining the company, in 1984, Dr. Waltz gained access to computers that by 80s standards held vast amounts of fast random-access memory, up to 512 megabytes. For the first time it was possible to use simple algorithms with lots and lots of data, said Brewster Kahle, a computer scientist who directs the Internet Archives and was one of the Thinking Machines researchers. (more)

The beauty of machine learning? It never stops learning
The beauty of machine learning? Machine learning makes it scalable and cost-effective to connect them in a world of unstructured data across millions of sites and products. While this machine learning sounds like a magical solution, its not bulletproof because it depends on human observation that varies from inputs.Amarnath Thombre, SVP, Strategy and Analytics at Match.com believes strongly that successful businesses will be those that observe their customers. Every organization will need to do machine learning to augment human decisions As data grows and machine learning improves, however, that will require more computing horsepower and multiple iterations of data interpretation. (more)

Brain imaging study finds evidence of basis for caregiving impulse
Distinct patterns of activitywhich may indicate a predisposition to care for infants-- appear in the brains of adults who view an image of an infant faceeven when the child is not theirs, according to a study by researchers at the National Institutes of Health and in Germany, Italy, and Japan. To collect the data, the researchers showed seven men and nine women a series of images while recording their brain activity with a functional magnetic resonance imaging scanner. When the researchers compared the areas and strength of brain activity in response to each kind of image, they found that infant images evoked more activity than any of the other images in brain areas associated with three main functions: Premotor and preverbal activity: The researchers documented increased activity in the premotor cortex and the supplemental motor area, which are regions of the brain directly under the crown of the head. (more)

On Text Analytics vs Machine Translation
About Ken Hu: Thinkudo Labs Currently, Ken is founding the text mining company Thinkudo Labs. Personally, I would claim that Text Analytics covers topics which extract and normalize text into measurable data. Instead of extracting information from the text, it transforms the text into another form. This may lead to misinterpretation that English text is a requirement for Text Analytics problems.However, that is just not true. (more)

There are two ways to make large datasets useful
Ive spent the majority of my career building technologies that try to do useful things with large datasets. * One of the most important lessons Ive learned is that there are only two ways to make useful products out of large data sets. Trading systems that hedge funds use are also often fault tolerant: if you make money 80% of the time and lose it 20% of the time, you can still usually have a profitable system. Because we knew automation would only get us 80-90% accuracy, we built 1) systems to estimate confidence levels in our ratings so we would know what to manually review, and 2) a workflow system so that our staff, an offshore team we hired, and users could flag or fixinaccuracies. (more)

Super-Turing Network to Revolutionize Computer Intelligence
A new breakthrough in neural networking might just lead to truly intelligent computers. Dubbed a super-Turing network, the new approach makes the neural networks so common to artificial intelligence research work very much like how our brains do. Statue of Turing by Stephen Kettle at Bletchley Park The model is based on a previous theoretical model that she posited in 1993. It means that her neural network learns an order of magnitude more effectively and faster than more traditional neural networks. (more)

Is Machine Learning v Domain Expertise the wrong question?
About James Taylor: CEO Decision Management Solutions James Taylor is CEO and Principal Consultant at Decision Management Solutions and one of the leading experts in decision management and decisioning technologies. Some decisions are heavily influenced by policy and regulation (deciding if a claim is complete and valid for instance) while others are more heavily influenced by the kind of machine learning insight common in analytics (deciding if the claim is fraudulent might be largely driven by a Neural Network that determines how normal the claim seems to be). We could also ask if there are really any decisions where machine learning or analytics cannot help at all (probably but only because the decision-makers dont have access to data that would help or because they are obliged to follow a precise set of regulations/policies). Or we could ask if there were any decisions that only required know-how that can be derived automatically using machine-learning (probably not, most business decisions involved some policy and regulations that are fixed even if we can replace experience with machine learning). (more)

Long Term Memory Encoding in the Brain
Researchers may finally have uncovered the brain mechanism for storage of long term memories. The researchers summarize their discoveries this way:"We demonstrate a feasible and robust mechanism for encoding synaptic information into structural and energetic changes of microtubule (MT) lattices by calcium-activated CaMKII phosphorylation. We suggest such encoded information engages in ongoing MT information processes supporting cognition and behavior, possibly by generating scale-free interference patterns via reaction-diffusion or other mechanisms. As MTs and CaMKII are widely distributed in eukaryotic cells, the hexagonal bytes and trytes suggested here may reflect a real-time biomolecular information code akin to the genetic code. (more)

Using fMRI and machine learning for brain reading
In another Minority Report-like research finding, a UCLA research team has made crucial advances in brain reading, using fMRI and machine learning methods to predict reactions of smokers experiencing nicotine cravings. For the study, smokers sometimes watched videos meant to induce cravings, sometimes watched neutral videos and at sometimes watched no video at all. By measuring the brain networks active over time during the scans, the resulting machine learning algorithms were able to anticipate changes in subjects underlying neurocognitive structure, predicting with a high degree of accuracy (90 percent for some of the models tested) what they were watching and, as far as cravings were concerned, how they were reacting to what they viewed. We detected whether people were watching and resisting cravings, indulging in them, or watching videos that were unrelated to smoking or cravings, said Anderson, who completed her Ph.D. (more)

Red Lambda eyes way to crack cloud provider complexity
Red Lambda eyes way to crack cloud provider complexity New IT security and monitoring software platform is designed to manage cloud services data Software firm Red Lambda has today announced an early release programme for its new MetaGrid security and operational intelligence software platform. Powered by the companys AppIron grid computing platform, MetaGrid is designed to unify operational silos and situational awareness, visualise and analyse network security and operational anomalies, and automate IT operations across so-called 'Big Data' environments, like those of cloud services providers. The company said MetaGrid's first four (unnamed) customers currently deployed under the early release programme work in network and cloud infrastructure, social web, and government contractor arenas. Although these early adopters operate in different industries, they all face the same challenge of handling and making sense of unprecedented data volumes and mitigating increased security risks the faster the data grows. (more)

Digital Image Processing and Analysis (2e)
In fact it is a fairly traditional account of mostly linear approaches to image processing. It makes use of CVIPtools - the Computer Vision and Image Processing Algorithm Test and Analysis Tool (CVIP-ATAT) and the CVIP Feature Extraction and Pattern Classification Tool (CVIP-FEPC), which I have to admit I hadn't encountered before reading this book. The book starts off with two chapters covering the obligatory look at computer imaging systems - mostly it is about the CVIPtools but it also covers file formats. Section II is where the real work starts. (more)

Computational biomarkers can identify at-risk heart attack victims
Subtle markers of heart damage hidden in plain sight among hours of EKG recordings could help doctors identify which heart attack patients are at high risk of dying soon, researchers from the University of Michigan, MIT, Harvard Medical School, and Brigham and Women s Hospital in Boston have discovered, The findings could help match tens of thousands of cardiac patients with life-saving treatment in time. Missing 70 percent of high-risk patients Today s methods for determining which heart attack victims need the most aggressive treatments can identify some groups of patients at a high risk of complications. Using data mining and machine learning techniques, the researchers sifted through 24-hour continuous electrocardiograms (EKGs or ECGs) from 4,557 heart attack patients. These could be prevented with medication or implantable defibrillators, which can shock the heart back into rhythm. (more)

UB's Srihari Wins Major International Computer Science Award
His speech, entitled "Probabilistic Graphical Models in Machine Learning," focused on the design of computer programs that learn and are able to modify their behavior in an environment of constantly changing information. Machine learning is crucial in fields such as document analysis and recognition due to the difficulty of expressing perceptual images, such as handwriting, in algorithms that computers can understand. Many second-generation machine learning programs were enabled by postal data collected at the Buffalo post office by UB CEDAR students. " Research by Srihari, his colleagues and students at CEDAR that allowed machines to recognize and understand handwriting was at the core of the first handwritten address-interpretation system used by the U.S. (more)

11 Unusual Ways Steve Jobs Made Apple The World's Most Admired Tech Company
The human brain is divided into two hemispheres: left and right. It's faster, mostly gray matter and is optimized to perform linear-sequential processing. That means the left hemisphere is best at processing information in a single-dimension, one sequential item after another. The left hemisphere stores words. (more)

Software to prevent abuse at the click of a mouse
Teaming up with investigators from the State Office of Criminal Investigation in Berlin, Fraunhofer researchers have come up with an automated assistance system for image and video evaluation that can detect child-pornographic images from among even large volumes of data. Many are avid data collectors: when suspects data media are confiscated, detectives must often click their way through hundreds of thousands of files to find the illegal images they seek. The algorithms use up to several thousand characteristics that describe properties such as color, texture and contours in order to analyze whether an image depicts child abuse. If the system is run on a standard PC, it classifies up to ten images per second, drastically accelerating detectives investigations. (more)

Advancements in Speech Recognition Set to Improve IVR
Advancements in Speech Recognition Set to Improve IVR As much as we like automated systems, we also like to use voice to move through steps and complete interactions. This recent Plum Voice blog focused on the advancements in IVR , thanks to improvements in speech recognition. Increases in computer processing speeds have enabled speech recognition developers to create more natural, accurate speech recognition software. Susan J. Campbell is a contributing editor for TMCnet and has also written for eastbiz.com. (more)

Language localized in the brain
Functional specificity refers to the idea that discrete parts of the brain handle distinct tasks. To determine this, the researchers analyzed each subject individually using fMRI, making sure thatpatterns of activity in one brain would only ever be compared to patterns of activity from that same brain.The researchers spent the first 10 to 15 minutes of each fMRI scan having their subject do a fairly sophisticated language task while tracking brain activity. This way, they established where the language areas lie in that individual subject, so that later, when the subject performed other cognitive tasks, they could compare those activation patterns to the ones elicited by language. The researchers said the results don t imply that every cognitive function has its own dedicated piece of cortex. (more)

Computers will be able to tell social traits from human faces, researchers predict
Researchers have developed new computational tools that help computers determine whether faces fall into categories like attractive or threatening, according to a recent paper published in the journal PLoS ONE. Mario Rojas and other researchers at the Computer Vision Center in the Autonomous University of Barcelona in Spain, in cooperation with researchers from the Department of Psychology of Princeton University, developed software that is able to predict those traits in some cases with accuracies beyond 90%. Facial characteristics play a central role in our everyday assessments of other people. Specifically, the task was formulated with the intention of predicting 9 facial trait judgments (attractive, competent, trustworthy, dominant, mean, frightening, extroverted, threatening, and likable) using Machine learning techniques (a branch of artificial intelligence that uses examples to teach a program how to work). (more)

A tour of the lab where IBM makes its brain chips (video)
A tour of the lab where IBM makes its brain chips(video) San Jose, Calif. At the IBM Almaden Research Center , a team of researchers is creating the artificial brains of the future. We took a tour of this lab from Dharmendra Modha (pictured), principle investigator on the brain chip project. IBM and its partners have already built a brain-like chip prototype , and Modha and his colleagues showed us how it works. IBM calls the larger project Synapse (Systems of Neuromorphic Adaptive Plastic Scalable Electronics, or SyNAPSE). (more)

Q: How does someone become a data scientist?
The author is a Forbes contributor. See also: Vardi, Science has only two legs: http://portal.acm.org/ft_gateway Here are some resources Ive collected about working with data, I hope you find them useful (note: Im an undergrad student, this is not an expert opinion in any way). 1) Learn about numerical analysis Take the Computational Linear Algebra course (it is sometimes called Applied Linear Algebra or Matrix Computations or Numerical Analysis or Matrix Analysis and it can be either CS or Applied Math course). Crays and Connection Machines of the past can now be replaced with farms of cheap cloud instances, the computing costs dropped to less than $1.80/GFlop in 2011 vs $15M in 1984: http://en.wikipedia.org/wiki/FLOPS . (more)

Ideas: Evolutionary Computing and Internet As Brain
First, the book Blondie24 , by David Fogel, describing how he and his team used evolutionary computing to develop computer programming that taught itself to play checkers. First, from Blondie24, David Fogel is making the point that the goal is not elaborating human thought into computers in sets of rules and conclusions, but generating an independent evolutionary process: For example, suppose we wanted to design a flying machine. But in emulating those specific manifestations of flight, wed be led astray. We toyed with the idea of using genetic algorithms to solve complex manufacturing scheduling problems, trying to implement a lean workflow in a situation where there were 100s of inter-dependencies in set-up, materials, etc. (more)

Speech Recognition Leaps Forward
Speech Recognition Leaps Forward During Interspeech 2011 , the 12th annual Conference of the International Speech Communication Association being held in Florence, Italy, from Aug. 28 to 31, researchers from Microsoft Research will present work that dramatically improves the potential of real-time, speaker-independent, automatic speech recognition. Dong Yu , researcher at Microsoft Research Redmond , and Frank Seide , senior researcher and research manager with Microsoft Research Asia , have been spearheading this work, and their teams have collaborated on what has developed into a research breakthrough in the use of artificial neural networks for large-vocabulary speech recognition. The notion of using ANNs to improve speech-recognition performance has been around since the 1980s, and a model known as the ANN-Hidden Markov Model (ANN-HMM) showed promise for large-vocabulary speech recognition. The new project applied CD-DNN-HMM models to speech-to-text transcription and was tested against Switchboard, a highly challenging phone-call transcription benchmark recognized by the speech-recognition research community. (more)

New App Can ID Complete Stranger's Facebook and Social Security No.
Please input the letters/numbers that appear in the image below. PittPatt jumps online and compares that picture to millions of images in Facebook and in Google Inc.'s ( GOOG ) image search, using advanced facial recognition technology. We are RAPIDLY approaching civil war in this country due to the government's refusal to operate within its constitutional limits. Gun control is unconstitutional. (more)

How Search Engines Use Machine Learning for Pattern Detection
Search engines use machine learning for pattern detection. While its impossible to explain in one short article how machine learning influences our lives, understanding the basics of machine learning can give you some insight into search algorithm updates, such as Googles Panda update. Correlation Between Variables To predict the outcome of future tests, scripts can use supervised learning on past outcomes to define a hypothetical prediction line. As the learning set grows the prediction becomes more certain. (more)

Siri's Sibling Launches Intelligent Discovery Engine
Were all familiar with the standard search engines such as Google and Yahoo, but there is a new technology on the scene that does more than just search the web it discovers it. Trapit, which is a personalized discovery engine for the web thats powered by the same artificial intelligence technology behind Apples Siri, launched its public beta last week. Trapit, which was first unveiled in June, is a system that personalizes content for its users based on keywords, URLs and reading habits. Since its first unveiling this past June, more than 10,000 participants have been testing out Trapit and using it to find and trap content. (more)

Artificial Intelligence Improves Fossil Finds
AI-Fossil Finders: The Research Team Robert Anemone, Charles Emerson, and Glenn Conroy are the researchers behind this new method of finding fossils.Robert Anemone is Professor of Anthropology and Charles Emerson is Associate Professor of Geography both at Western Michigan University.Glenn Conroy is Professor of Anatomy and Anthropology at Washington University Schoolof Medicine, in St. Louis, MO. Decoded Science: Did you train the software through supervised learning, unsupervised learning, or a combination of the two? Supervised Learning, Unsupervised Learning, or Both? Programmers can use supervised learning, unsupervised learning, or a combination of both. (more)

Riverain gets FDA approval of lung cancer-detection software
Riverain Technologies has received regulatory approval for the next-generation version of its imaging software that suppresses bones to help radiologists detect cancerous lung nodules. Advertisement The new version of the software offers greater sensitivity, meaning it can better detect nodules, and better specificity, meaning it yields fewer false positives, said Steve Worrell, Riverains chief technology officer. A recent study of the new version of OnGuard by researchers at University of Chicago Medical Center found that the software identified 25 percent more lung cancers than radiologists found when reviewing the same X-rays without OnGuard. Brandon Glenn MedCity News Brandon Glenn is the Ohio bureau chief for MedCity News. (more)

Developing artificial intelligence systems that can interpret images
Today, Torralba is a tenured associate professor of electrical engineering and computer science at MIT, and an affiliate of the Computer Science and Artificial Intelligence Laboratory (CSAIL), where he develops AI systems that can interpret images to understand what scenes and objects they contain. I wanted to build systems that could put objects into context, to try to understand how different objects relate to each other, he says. So he began developing systems that used information gathered from the entire image to help identify individual objects. If an image contains an object perched on top of a table, for example, that object is unlikely to be anything very large, narrowing down considerably the number of things it could possibly be. (more)

Artificial intelligence for quantum chemistry
Drawing on a database of quantum chemical results for over 7000 molecules, their program could give the atomisation energies of unfamiliar molecules to within 1% - and in a billionth of the time required for a full approximation. Von Lilienfeld's team trained the algorithm on a subset of molecules in the database, comparing their matrices to find 'distances' between molecules - a measure of the difference between the molecules in terms of their matrices. In the case of finding an unknown molecule's atomisation energy, the distances between the unknown molecule and all the known molecules gave weights for how much each known atomisation energy could contribute to an estimate for the unknown molecule. The researchers found that with a landscape of more than 5000 molecules, the error for predicting atomisation energies of new molecules drops below 10kcal/mol, approaching the 5kcal/mol accuracy of hybrid DFT. (more)

Secret Formula For Writing A Hit Song
Mathematicians DiscoverThough math is usually used to tell you if a song was a hit based on sales, scientists from the University of Bristol believe they have " The Hit Equation. " A group focused on artificial intelligence developed machine learning algorithms that mined the characteristics of top hits to develop a formula for key musical features that determine whether a song will be a hit. I don't know how Hit Song Science does its magic, but they claim to analyze " fundamental characteristics of all music " including brightness, tempo and changes over time, to identify hit potential in relationship to changing trends. (more)

Blogging the Stanford Machine Learning Class
If there is a way to split baseball teams into five groups, the machine will find it Photograph by Nick Laham/Getty Images. This is a classic example of a supervised learning problem: Were telling the computer ahead of time what its looking forone of 10 symbolsand helping it train by feeding it data in which the correct answer has already been provided. Lets say you were asked to take your favorite sport and divide the teams into two categories based on their style of play, ignoring structural groupings like leagues and divisions. As a baseball fan, my instinct would be to divide the major leagues into teams that emphasize pitching and those that focus on hitting. (more)

Facial recognition software spots family resemblance
FACIAL recognition software that's as good as people at spotting family resemblances could help to reunite lost family members - or help the likes of Facebook work out which of your friends are blood relatives. To do this, the team used a database of public figures and their parents or children - such as French president Nicolas Sarkozy and his son Jean - and fed the program 320 pairs each of parent-child matches and mismatches. The software then compared the difference between a test pair of photos with pairs of photos in its database. In tests using 160 pairs - 80 parent-child matches and 80 mismatches - the system had a success rate of 68 per cent. (more)

Silicon Valley's Kinect Contributions
I got very excited about those demos, Budiu says, and I kept telling everyone how cool they were. Budiu, a researcher at Microsoft Research Silicon Valley, soon got his chance. When Williams approached Budiu in July 2009 about solving some of the challenges with the project, Budiu jumped at the opportunity. I said, Yes, absolutely, Budiu says, and I put everything else I was doing on hold. (more)

Why Allow Customers to Suffer the Consequences from Bad Chat Translations?
SMT, Lionbridge officials explain, has contributed to the growth in machine translation applications in recent years since it can analyze vast amounts of previously translated material, and target-language texts to create what Lionbridge officials say is a real-time translation in a fraction of the time that it takes to produce traditional rules- based translation systems. The way it works, is customer use the Customizer along with customer-supplied translation memories and glossary assets, GeoFluent then translates for each customer by application domain, applying machine-learning methods to analyze your translation memories, glossaries and target- language documents, Lionbridge officials explain, consistent with your terminology and branding. Through this just-inked agreement, the two companies will create Lionbridge GeoFluent for LivePerson Chat, an integrated cloud-based multilingual chat application that enables contact centers and enterprises to gain access to on-demand, quality translation that can take place directly within their eLivePerson chat application. (more)

University adopts predictive technology
Cynthia Karena Siri was not the first speech recognition application to grace our phones, but its success will help increase expectations for more artificial intelligence in business. Rather than answering simple questions such as 'how many students are enrolled', IBM's predictive model can answer more complex questions such as 'which students are showing signs of needing extra support? ' and 'what are the causes? ' "The university can intervene when a flag goes up," says Kittle, "for example, low performance over one semester. (more)

Sue Me, I've Been Sick
He has worked in IT in Southern California since 1993, and is currently working on a Ph.D. He's been a board game player since he could walk, a poker player since he was thirteen, an amateur card-counter, and has had Strong Opinions about video games since 1980. If you've got a song for Wednesday, a commercial for Saturday, a recommendation for Tuesday, an essay for Monday, or, heck, just a handful a questions, fire off an email to AskJaybird@gmail.com Patrick is an about-40 year old geek with an undergraduate degree in mathematics and a master's degree in Information Systems. He has worked in IT in Southern California since 1993, and is currently working on a Ph.D. (more)

Blogging the Stanford Machine Learning Class
It occurred to me halfway through this weeks machine learning lectures that I could actually use the stuff were learning for something other than the homework assignments. This is known as your training setthe information youll feed to your algorithm so that it can slowly adjust its many partsthat is, learn the quirks of the housing market. Advertisement In the case of Robottke, we already had a very thorough training set of real blog posts by Kottke going back five years, with information like how often he would link to a given site, what keywords he likes to assign to his posts, and which bloggers and Twitter users he frequently turned to for ideas. (Essentially, we were doing the same thing a machine-learning algorithm does, just very badly. (more)

Google and Microsoft Talk Artificial Intelligence
Technology Review : You both spoke on stage of how AI has been advanced in recent years through the use of machine-learning techniques that take in large volumes of data and figure out things like how to translate text or transcribe speech. What about the areas we want AI to help where there isn't lots of data to learn from? Eric Horvitz: I've often thought that if you had a cloud service in the sky that recorded every speech request and what happened nextevery conversation in every taxi in Beijing, for exampleit could be possible to have AI learn how to do everything. Isn't it difficult to use machine learning if the training data isn't already labeled and explained, to give the AI a "truth" to get started from? (more)

Artificial intelligence joins the fossil hunt
Success in finding bones boils down to a lot of luck, says Robert Anemone of Western Michigan University in Kalamazoo, who once blundered into "the best locality we ever found" - a cache of early primate bones from between 40 and 50 million years ago - after making a wrong turn during a trip in the Great Divide basin of south-western Wyoming. So the team began by feeding the software a list of known locations in the 10,000-square-kilometre Great Divide basin, labelling them either as being fossil-rich or belonging to one of four other categories - barren, forest, scrub or wetland. Using only the satellite data, the computer had learned that the area's fossil sites were in sandstone - but not all sandstone has fossils at the surface. It correctly identified 79 per cent of the known fossil sites as likely to contain fossils, and correctly classified 84 per cent of all the other locations, Emerson says. (more)

Computers found more accurate than doctors in breast-cancer diagnosis
Computer analyses of breast cancer microscopic images were found more accurate than those conducted by humans, computer scientists at the Stanford School of Engineering and pathologists at the Stanford School of Medicine report. Medical science has long used three specific features for evaluating breast cancer cells: what percentage of the tumor is comprised of tube-like cells, the diversity of the nuclei in the outermost (epithelial) cells of the tumor and the frequency with which those cells divide (a process known as mitosis). These three factors are judged by sight with a microscope and scored qualitatively to stratify breast cancer patients into three groups that predict survival rates. In fact, they discovered that the characteristics of the cancer cells and the surrounding cells, known as the stroma, were both important in predicting patient survival. (more)

AGIRI 2006 Workshop Keynote Speaker: Dr. Stan Franklin (Dir. Institute for Intelligent Systems, University of Memphis) - A Cognitive Theory of Everything: The LIDA Technology as an Artificial General Intelligence.
"Implementing and fleshing out a number of psychological and neuroscience theories of cognition, the LIDA conceptual model aims at being a cognitive 'theory of everything.' With modules or processes for perception, working memory, episodic memories, 'consciousness,' procedural memory, action selection, perceptual learning, episodic learning, deliberation, volition, and non-routine problem solving, the LIDA model is ideally suited to provide a working ontology that would allow for the discussion, design, and comparison of AGI systems. The LIDA technology is based on the LIDA cognitive cycle, a sort of 'cognitive atom.' The more elementary cognitive modules play a role in each cognitive cycle. Higher-level processes are performed over multiple cycles. This talk will give a quick overview of the LIDA conceptual model, and its underlying computational technology." May 20, 2006. (more)

Alex (Sandy) Pentland, director of the Human Dynamics Group at MIT, describes Reality Mining.
"Alex (Sandy) Pentland, director of the Human Dynamics Group at MIT, describes a future in which cell phones log data about their owners' behavior. He reasons that this data can be used to strengthen social networks, generate recommendations, help track diseases, and monitor personal health." 2008?. (more)

ArsDigita University Curriculum - The Structure and Interpretation of Computer Programs course: Holly Yanco's lecture about Data Structures (Trees, Trees, Trees). This is lecture video #8 (of 19) for the course.
The Structure and Interpretation of Computer Programs course: "An introduction to programming and the power of abstraction, using Abelson and Sussman's classic textbook of the same name. Key concepts include: building abstractions, computational processes, higher-order procedures, compound data, data abstractions, controlling interactions, generic operations, self-describing data, message passing, streams and infinite data structures, meta-linguistic abstraction, interpretation of programming languages, machine model, compilation, and embedded languages." October 12, 2000. (more)

CSE Colloquia - 2005: Learning, Logic, and Probability - A Unified View.
"Artificial intelligence systems must be able to learn, reason logically, and handle uncertainty. Research has focused on each of these goals individually, and only recently have attempts been made to achieve all three at once. In this colloquium, Pedro Domingos, UW Computer Science & Engineering, describes Markov logic: a representation that combines the full power of first-order logic and probabilistic graphical models, and algorithms for learning and inference in it. Experiments in a real-world university domain." November 2, 2004. (more)

CSE Colloquia - 2005: Natural Language Processing.
Natural language processing offers a rich problem domain for machine learning approaches. Many NLP problems require the induction of a mapping that involves complex, discrete structures such as strings, labeled sequences, or trees. In this distinguished lecture, Michael Collins [Massachusetts Institute of Technology] describes how 'large margin' methods in machine learning can be generalized to 'structured' problems found in NLP. December 7, 2005. (more)

Computer Chronicles: Neural Networks.
"Neural networks are artificial intelligence systems modeled after the human brain. This program looks at several examples and applications. Included are Braincel 1.1 from Promised Land Technologies [demonstrated by Murray Ruggiero], BrainMaker Professional 2.0 from California Scientific Software [demonstrated by Mark Lawrence], MacBrain 3.0 from Neurix [demonstrated by Matt Jensen], NeuroSMARTS from Cognition Technology [demonstrated by Richard Mansfield], and ExploreNet from HNC. Also includes visits to NASA [Max Reid describes HONN: Higher Order Neural Network] and Intel [Mark Holler describes ETANN: Electronically Trainable Analog Neural Network] to see the work they're doing on neural networks." Also appearing on the show is Tom J. Schwartz (The Schwartz Assoc.). Hosted by Stewart Cheifet and Jan Lewis. May 15, 1991. (more)

DARPA LAGR Program: Learning Applied to Long-Range Vision using a Collision-Free Navigation Platform.
New York University uses machine learning to extend the vision of the DARPA LAGR robots. July 14, 2008. (more)

Discussion of and Demonstrations of Learning Programs for Robots.
The first half of the film is a lecture by Marvin Minsky describing the basic ideas of Patrick Winston's learning program, using examples and "near misses" to refine the program's model of what an "arch" is. The second half of the film is a narration by Dave Waltz describing other robotics research at MIT. He discusses Tim Finin's program that uses Winston-like models to recognize objects that match the model even when parts of the object are obscured. It uses hypotheses about dimensions of the objects that it can not directly observe. 1975??. (more)

Eric Horvitz with Microsoft Research on “Surprise Modeling”.
Eric Horvitz, head of the Adaptive Systems and Interaction group at Microsoft Research, talks about surprise modeling. 2008?. (more)

ICML 2007 - The 24th Annual International Conference on Machine Learning.
Collection of over 40 lectures on machine learning given at the Intl. Conf. on Machine Learning, Corvallis, OR, 2007. Invited talks by Bernhard Schölkopf, Josh Tenenbaum, and David Heckerman. June 20-24, 2007. (more)

Interview with Tom Mitchell.
8 min. interview with Tom Mitchell about machine learning, from CMU's 2006 ML Autumn School. 2006. (more)

Lighthill Controversy Debate at the Royal Institution with Professor Sir James Lighthill, Professor Donald Michie, Professor Richard Gregory and Professor John McCarthy.
Professors Donald Michie [Edinburgh], Richard Gregory [Bristol] and John McCarthy [Stanford] challenge the pessimistic findings & views of Professor Sir James Lighthill [Cambridge], author of "The Lighthill Report" [Artificial Intelligence: A General Survey, in Artificial Intelligence: a paper symposium, Science Research Council (1973)]. June 1973. (more)

Mind Reading.
“As pollsters have so well demonstrated this presidential primary season, reading minds, whether of voters or the person next to you, is close to impossible. However as this ScienCentral News video explains, scientists are actually one step closer to reading our thoughts. … [T]he new research is aimed at the biology underlying thoughts-- or, as scientists call them, ‘cognitive processes.’ Carnegie Mellon cognitive psychologist Marcel Just teamed up with machine learning expert Tom Mitchell to conduct the research.” February 2, 2008. (more)

NATO Advanced Study Institute Workshop on Mining Massive Data Sets for Security (MMDSS 2007) presentation by Ekrem Duman (Dogus University, Turkey) - Detecting Money Laundering Actions Using Data Mining and Expert Systems.
"Nowadays terrorism is one of the biggest troubles that almost every country faces. It mainly influences the economy and the well being of the citizens and this effect is relatively larger in the developed countries. Since the financial sources of terrorist groups can be regarded as black money, the solutions against the money laundering actions can be expected to identify the transactions of the terrorists. Then, blocking their accounts could slow down their actions if cannot stop. In many countries, the financial institutions are expected to inform compliance regulation bodies about any persons or transactions that they think suspicious. To cope with this necessity, various software packages for anti money laundering (AML) have been developed and are commercially available." In this talk, Ekrem Duman explores the factors that must be addressed in building these programs. Q&A follows the talk.. September 17, 2007. (more)

Overview Talk on Informatics by Edward "Ted" H. Shortliffe, MD, PhD., presented at the Biomedical Informatics @ Arizona State University Symposium 2006.
An overview of the field, from inception to current trends, and suggestions for how to establish a new Biomedical Informatics academic program. January 19, 2006. (more)

Scientific American Frontiers with Alan Alda: "Almost Human" segment from the "Robots Alive!" broadcast.
Rodney Brooks is beginning to build the first robot with human-like senses, allowing it to learn about the world for itself, like a human baby. April 9, 1997. (more)

Scientific American Frontiers with Alan Alda: "Robot Independence" segment from the "Life's Really Big Questions" broadcast.
Natural selection is at work in the artificial world, as robots learn to reproduce without us. December 19, 2000. (more)

Self-Improving Artificial Intelligence
Lecture by Steve Omohundro for the Stanford University Computer Systems Colloquium (EE 380). Steve presents fundamental principles that underlie the operation of "self-improving systems," i.e., computer software and hardware that improve themselves by learning from their own operations. October 24, 2007. (more)

Self-Improving Artificial Intelligence.
Lecture at Stanford by Stephen Omohundro, Self-Aware Systems. "We are on the verge of a radical new paradigm for both computer software and hardware. "Self-improving systems" will have detailed models ... all » ... all » of their own designs and will improve themselves by learning from their own operation. They will continuously adapt themselves to the tasks they need to perform. Eventually they will be able to improve every aspect of themselves: their programs, programming languages, specification logics, instruction sets, and hardware architectures. In this talk we present fundamental principles that underlie the operation of this kind of system. ... We conclude with a discussion of some of the broader social implications of this kind of system.". October 31, 2007. (more)

Technology Review Documentary: Evolutionary Design.
Computers can provide design variations that no human would have imagined. September 2006. (more)

The Age of Intelligent Machines: The Film. By Raymond Kurzweil.
From the original video notes: A survey of Artificial Intelligence showing AI at work and under development. The paradoxes, promise and challenges of advanced computer science, with authorities Marvin Minsky, Roger Schank, Raj Reddy and other leaders in the field. 1987. (more)

The Discipline and Future of Machine Learning.
Seminar talk by Tom Mitchell at the Carnegie Mellon University School of Computer Science Machine Learning Department. March 1, 2007. (more)

The Future of Robotics and Artificial Intelligence
Andrew Ng (Stanford University) is building robots to improve the lives of millions. From autonomous helicopters to robotic perception, Ng's research in machine learning and artificial intelligence could result one day in a robot that can clean your house. Control and Perception are the two main problems; machine learning is essential. May 21, 2011. (more)