Insurance firm to replace human workers with AI system

http://ift.tt/2iLbwNm


Fukoku Mutual Life Insurance Co. is planning to slash nearly 30 percent of its payment assessment department's human staff after it introduces an artificial intelligence (AI) system in January 2017 to improve operating efficiency.

While concrete examples of AI systems making human workers redundant are currently rare, observers have pointed out that such cases are likely to increase.

The insurance firm will introduce an AI system based on IBM Japan Ltd.'s Watson, which according to IBM is a "cognitive technology that can think like a human," and "can analyze and interpret all of your data, including unstructured text, images, audio and video." The Watson-based system will be tasked with reading medical certificates written by doctors and other documents to collect information necessary for making payouts, such as medical histories, length of hospital stays, and surgical procedure names.

In addition to determining payment amounts, the system will also be able to check customers' cases against their insurance contracts to find any special coverage clauses -- a measure expected to prevent payment oversights. The type of payments the AI is expected to oversee at Fukoku Mutual totaled some 132,000 cases in fiscal 2015.

The company's payment assessment-related department had 131 employees as of March 2015. Final payout decisions will continue to be dealt with by a dedicated staff, but the introduction of the AI system will make reading medical certificates and other procedures more efficient.

Fukoku Mutual has already begun staff reductions in preparation for the system's installation. In total, 34 people are expected to be made redundant by the end of March 2017, primarily from a pool of 47 workers on about five-year contracts. The company is planning to let a number of the contracts run out their term and will not renew them or seek replacements.

The insurance firm will spend about 200 million yen to install the AI system, and maintenance is expected to cost about 15 million yen annually. Meanwhile, it's expected that Fukoku Mutual will save about 140 million yen per year by cutting the 34 staff.

Dai-ichi Life Insurance Co. is already using a Watson system to process payment assessments, but alongside human checks, and it appears there have been no major staff cuts or reshuffling at the firm due to the AI's introduction. Japan Post Insurance Co. is also looking to install a Watson AI for the same duties, and is set to start a trial run in March 2017.

Meanwhile, Nippon Life Insurance Co. began this month to use an AI system to analyze the best coverage plans for individual customers, based on the some 40 million insurance contracts held by its various salespeople. The system's results are then used as a reference by the sales offices.

While AI systems are expected to find applications in developing new products and insurance underwriting, there are worries the technology could put pressure on the employment market, as the machines trigger staff reshuffling or reductions. The future of AI may be bright in the business community, but it also has a shadowy side.



from Hacker News http://ift.tt/YV9WJO
via IFTTT

Russia hits back after US sanctions: Proposes kicking out 35 US diplomats

http://ift.tt/2ieR7DD


Lavrov said his ministry proposed that 31 staff members be expelled from the US Embassy in Moscow and four from the US consulate in St Petersburg.

The Russian foreign minister called US allegations of Russia's interference in the US election campaign "groundless."

The US on Thursday ordered 35 Russian diplomats to leave the country

and ordered the closure of two Russian compounds. The diplomats and their families were given 72 hours to leave the country.

Russian Prime Minister Dmitry Medvedev on Friday accused the Obama administration of ending its term in office "in anti-Russian agony."

Posting on his official Twitter and Facebook accounts, Medvedev said, "Sadly, the Obama administration, which began its life with the restoration of cooperation, ends it in anti-Russian agony. RIP."

Kremlin spokesman Dmitry Peskov insisted Thursday that the claims of election meddling were "groundless" but said there was "absolutely no alternative to the principle of reciprocity" now that sanctions had been imposed.

The steps taken by President Barack Obama Thursday mark a new low in what have become increasingly frosty relations between Russia and the United States.

The Obama administration described Russia's actions as "Significant Malicious Cyber-Enabled Activities" and sanctioned four Russian individuals and five Russian entities for what it said was election interference.

"Russia's cyberactivities were intended to influence the election, erode faith in US democratic institutions, sow doubt about the integrity of our electoral process, and undermine confidence in the institutions of the US government," a White House statement said. "These actions are unacceptable and will not be tolerated."



from CNN http://ift.tt/1UBaoaB
via IFTTT

Most Popular Features and Essays of 2016

http://ift.tt/2hxIenR

We do a lot of how-tos and explainers around here, but sometimes we just need to get a rant off our chest or explain some things from hard-earned personal experience. These are our best essays, rants, and features from 2016.

The Stuff That Costs More When You’re Poor

Spend less than you earn, save your money, and—poof!—your financial problems are solved. If only it were this easy. Being broke sucks enough on its own, and then there are obstacles that make it extra hard for poor people to fight their way to financial security. For example, here are a few expenses that actually cost more for low-income individuals.

Have Two Drinks at a Party

Everyone has their own relationship to and tolerance for alcohol, but next time you’re at a party, you might do well if you have exactly two drinks. If you are a person who has found that zero drinks, or one drink, is the right number for you, then that is the number to stick with. For everyone else, try two.

Why I Stopped Ghosting

In the dating world there’s a looming presence that haunts us all: ghosting. I’ve watched friends get ghosted, been ghosted on, and I’ve even been the ghost many times. But I decided to stop. Not just because I realized how impolite I was being, but because I also saw that vanishing into the ether was actually a disservice to myself.

Why the Rent vs. Buy Debate Is Completely Pointless

After saving up for a long time, I recently bought a home, which caught some of my friends off guard. “I thought you were anti-homeownership,” they said, because I think renting is underrated. Even as a homeowner, I still think renting is underrated. That doesn’t mean buying is a bad decision. The rent vs. buy debate is just silly overall. It ignores the enormous grey area that exists between the two options.

It Doesn’t Matter When You Eat

There are so many myths and strangely specific rules about when to eat to lose weight, but alone they do nothing to help. Eat a hearty breakfast and light all day.Eat small meals every few hours. Rules around when you eat are less important than you think, and even when they do help, they’re not for the reasons you think.

Yelp Is Awful for Everyone Involved

“Your review on Yelp is destroying my business,” he says to me, clearly clenching his teeth, “How long do I have to suffer because of your negative review?” A few weeks ago, I got a phone call from a contractor because of a review I’d left. What ensued was a weirdly emotional conversation that ventured between harassment and a plea for empathy.

Stop Using iCloud

Apple’s iCloud has a long and troubled past, but the company keeps pushing it for iPhone and Mac users with every new operating system update. Don’t be fooled. The service is an inconsistent mess and more trouble than it’s worth.

The Surprising Things No One Tells You About Childbirth

As much as you might prepare for the birth of your child, chances are something unexpected will happen. Some moms have had glorious orgasms giving birth, while others’ experiences were more torturous. From uncontrollable poops to the need for new shoes, here are the gritty “secrets” about childbirth you might not have heard before.

I’ve Worked Bar Security, Here’s How to Talk Your Way Out of a Bar Fight

Sun Tzu once said “the supreme art of war is to subdue the enemy without fighting.” And that goes double when the drinks are flowing. People can get a bit hot-headed when they’re on the sauce, and in some cases, physical. Based on my experiences as a security guard, these smooth talking tips will help you calm an intense situation and escape a potential beating.

How I Got Super-Vision by Shooting My Eyeballs With Lasers

Two weeks ago I paid to have lasers fired into my eyes. The next day I woke up like Peter Parker, post-spider bite. I couldn’t climb walls, but I could see so well I felt like a superhero. For the first time since I was a kid I could open my eyes and just see. My laser eye surgery story isn’t quite comic book-worthy, but here’s everything that happened, in case you want those powers too.

How to Flirt With Finesse

You might dress well, have a cool job, and be blessed with beauty, but flirting is where the real magic of attraction is, especially when it comes to first impressions. In fact, good flirting is often more effective than good looks, and it’s something anybody can learn how to do.

The Worst Ways Cable Companies Confuse You Into Paying More

Thanks to confusing bundles, hidden charges, misleading promises, and obscure terms, everyone agrees that cable TV and internet providers are the absolute worst. I tried to comparison shop for internet service in my area and ended up an awful mess of sketchy terms, and upsells. It shouldn’t be like this.

Four Rules I Followed to Stop Being a Pushover and Make Myself More Powerful

You know those people who apologize for everything, and you point it out to them, and then they apologize for apologizing? Yep, that’s me. I’ve been a pushover my whole life, but the older I get, the sicker of it I get. I finally decided to do something about it.

Free Document Scanning Apps Are Sleazy and Gross, Don’t Download Them

Mobile document scanners are possibly the most boring apps imaginable, so it’s puzzling they’re also some of the most awful, sleazy, and confusing apps you can download. It should be simple: Scan receipts, digitize notes, sync to the cloud, that’s it. Useful, but not exciting. But there are dozens, all nearly identical. Some are free. Others are a couple bucks. Most have in-app purchases. All of them are confusing as hell.

Feeling Poor Doesn’t Stop Once You Make Money

From the day I got my first job as a cart pusher at Walmart, I spent years living from one paycheck to the next. I hovered around the poverty line, hoping that I would last until next month’s rent. At the time it felt normal. It wasn’t until after I started making more money that I realized the psychological scars that living the poor life left on me.

Top 10 Life Hacks That Don’t Make Your Life Better At All

If you want to improve your life, there are a lot of clever little tricks you can use called “life hacks” to fix problems. Some are great! Some are very, very bad. These are the worst hacks that either don’t work, waste your time, or cause more problems than they solve.



from Lifehacker http://lifehacker.com
via IFTTT

Obama Strikes Back at Russia for Election Hacking

http://ift.tt/2htr3AD


Mr. Trump will now have to decide whether to lift the sanctions on the Russian intelligence agencies when he takes office next month, with Republicans in Congress among those calling for a public investigation into Russia’s actions. Should Mr. Trump do so, it would require him to effectively reject the findings of his intelligence agencies.

Asked on Wednesday night at his Mar-a-Lago estate in Palm Beach, Fla., about reports of the impending sanctions, Mr. Trump said: “I think we ought to get on with our lives. I think that computers have complicated lives very greatly. The whole age of computer has made it where nobody knows exactly what is going on. We have speed, we have a lot of other things, but I’m not sure we have the kind, the security we need.”

President Obama, in a statement, put in a subtle dig at Mr. Trump’s unwillingness to talk about Russia’s role. “All Americans should be alarmed by Russia’s actions,” he said. He said he acted after “repeated private and public warnings that we have issued to the Russian government” and called the moves “a necessary and appropriate response to efforts to harm U.S. interests in violation of established international norms of behavior.”

The samples of malware were in what the Obama administration called a “joint analytic report” from the F.B.I. and the Department of Homeland Security that was based in part on intelligence gathered by the National Security Agency. A more detailed report on the intelligence, ordered by President Obama, will be published in the next three weeks, though much of the detail — especially evidence collected from “implants” in Russian computer systems, tapped conversations and spies — is expected to remain classified.

The Perfect Weapon: How Russian Cyberpower Invaded the U.S.

A Times investigation reveals missed signals, slow responses and a continuing underestimation of the seriousness of a campaign to disrupt the 2016 presidential election.

In Moscow, there was a sense that the Obama administration was trying to take unseemly last-minute revenge against Russia and President Vladimir V. Putin.

“We regret that this decision was made by the U.S. administration and President Obama personally,” Dmitri S. Peskov, the spokesman for Mr. Putin, told reporters. “As we have said before, we believe such decisions and such sanctions are ungrounded and illegal from the point of view of international law.”

Russia is studying the details of what Washington did, he said, and some manner of reciprocal answer can be expected.

Konstantin Kosachyov, the head of the foreign affairs committee in the upper house of the Russian Parliament, told Interfax that “this is the agony not even of ‘lame ducks,’ but of ‘political corpses.’”

Despite the fanfare and political repercussions surrounding the announcement, it is not clear how much real effect the sanctions may have, although they go well beyond the modest sanctions imposed against North Korea for its attack on Sony Pictures Entertainment two years ago.

Graphic: Following the Links From Russian Hackers to the U.S. Election

Starting in March 2014, the United States and its Western allies levied sanctions against broad sectors of the Russian economy and blacklisted dozens of people, some of them close friends of Mr. Putin’s, after the Russian annexation of Crimea and its activities to destabilize Ukraine. Mr. Trump suggested in an interview with The New York Times earlier this year that he believed those sanctions were useless, and left open the possibility he might lift them.

Mr. Obama and his staff have debated for months when and how to impose what they call “proportionate” sanctions for the remarkable set of events that took place during the election, as well as how much of them to announce publicly. Several officials, including Vice President Joseph R. Biden Jr., have suggested that there may also be a covert response, one that would be obvious to Mr. Putin but not to the public.

While that may prove satisfying, many outside experts have said that unless the public response is strong enough to impose a real cost on Mr. Putin, his government and his vast intelligence apparatus, it might not deter further activity.

“They are concerned about controlling retaliation,” said James A. Lewis, a cyberexpert at the Center for Strategic and International Studies in Washington.

The Obama administration was riven by an internal debate about how much of its evidence to make public. Although the announcement risks revealing sources and methods, it was the best way, some officials inside the administration argued, to make clear to a raft of other nations — including China, Iran and North Korea — that their activities can be tracked and exposed.

In the end, Mr. Obama decided to expand an executive order that he issued in April 2015, after the Sony hacking. He signed it in Hawaii on Thursday morning, specifically giving himself and his successor the authority to issue travel bans and asset freezes on those who “tamper with, alter, or cause a misappropriation of information, with a purpose or effect of interfering with or undermining election processes or institutions.”

Mr. Obama used that order to immediately impose sanctions on four Russian intelligence officials: Igor Valentinovich Korobov, the current chief of a military intelligence agency, the G.R.U., and three deputies: Sergey Aleksandrovich Gizunov, the deputy chief of the G.R.U.; Igor Olegovich Kostyukov, a first deputy chief, and Vladimir Stepanovich Alekseyev, also a first deputy chief of the G.R.U.

But G.R.U. officials rarely travel to the United States, or keep their assets here, so the effects may be largely symbolic. It is also unclear if any American allies will impose parallel sanctions on Russia.

The administration also put sanctions on three companies and organizations that it said supported the hacking operations: the Special Technologies Center, a signals intelligence operation in St. Petersburg; a firm called Zor Security that is also known as Esage Lab; and the Autonomous Noncommercial Organization Professional Association of Designers of Data Processing Systems, whose lengthy name, American officials said, was cover for a group that provided special training for the hacking.

“It is hard to do business around the world when you are named like this,” a senior administration official with long experience in Russia sanctions said on Thursday morning. The official spoke on the condition of anonymity because of the sensitive nature of the intelligence.

Got a confidential news tip?

The New York Times would like to hear from readers who want to share messages and materials with our journalists.

But the question will remain whether the United States acted too slowly — and then, perhaps, with not enough force. Members of Hillary Clinton’s election campaign argue that the distractions caused by the leaks of emails, showing infighting in the D.N.C., and later the private communications of John D. Podesta, the campaign chairman, absorbed an American press corps more interested in the leaks than in the phenomena of a foreign power marrying new cybertechniques with old-style information warfare.

Certainly the United States had early notice. The F.B.I. first informed the D.N.C. that it saw evidence that the committee’s email systems had been hacked in the fall of 2015. Months of fumbling and slow responses followed. Mr. Obama said at a new conference he was first notified early this summer. But one of his top cyberaides met Russian officials in Geneva to complain about cyberactivity in April.

By the time the leadership of the D.N.C. woke up to what was happening, the G.R.U. had not only obtained those emails through a hacking group that has been closely associated with it for years, but, investigators say, also allowed them to be published on a number of websites, from a newly created one called DC Leaks to the far more established WikiLeaks. Meanwhile, several states reported the “scanning” of their voter databases — which American intelligence agencies also attributed to Russian hackers. But there is no evidence, American officials said, that Russia sought to manipulate votes or voter rolls on Nov. 8.

Mr. Obama decided not to issue sanctions ahead of the elections, for fear of Russian retaliation ahead of Election Day. Some of his aides now believe that was a mistake. But the president made clear before leaving for Hawaii that he planned to respond.

The question now is whether the response he has assembled will be more than just symbolic, deterring not only Russia but others who might attempt to influence future elections.

Continue reading the main story


from Hacker News http://ift.tt/YV9WJO
via IFTTT

The Cars People Regret Buying Most, According to Consumer Reports

http://ift.tt/2hpSICi

Buying a new car is always exciting at first, but it can take a little while before you realize your ride is more lemon than luxury. Consumer Reports surveyed owners of recently-purchased cars to see which models caused the most buyer’s remorse.

After reaching out to over 300,000 new car owners, Consumer Reports ended up with a list of seven cars you might want to reconsider before you drive them off the lot. In the video above, autos editor Mike Monticello goes over each one, in seven different categories:

  1. Small Cars: Dodge Dart. Owners cited sluggish acceleration and weak A/C.
  2. Midsized Sedans: Chrysler 200. Owners said that it’s difficult to get in and out of, and has a rear seat that’s too small.
  3. Small SUVs: Jeep Compass. Owners cited feeble acceleration and disappointing fuel mileage.
  4. Midsized SUVs: Nissan Pathfinder. Owners mentioned too many unscheduled dealer trips, uncomfortable seats, and low fuel mileage as common issues.
  5. Minivans: Dodge Grand Caravan. Owners cited rough shifting, uncomfortable seats, and a cheap-looking interior.
  6. Pickups: Nissan Frontier. Owners complained about too much road noise and clumsy steering.
  7. Worst overall: Acura ILX. Less than half surveyed said they’d buy the car again, mentioning a general lack of quality, pokey acceleration, abundant road noise, and a rough ride as issues.

Remember, a good price doesn’t mean a good deal. Always do plenty of research before you decide to put money down on a car.

Buyer’s Remorse: 7 Cars Owners Regret | YouTube



from Lifehacker http://lifehacker.com
via IFTTT

Berkeley: Machine Learning Crash Course

http://ift.tt/2i8BOfu

Machine Learning Crash Course: Part 2

By Daniel Geng and Shannon Shih 24 Dec 2016

Perceptrons, Logistic Regression, and SVMs

In this post we’ll talk about one of the most fundamental machine learning algorithms: the perceptron algorithm. This algorithm forms the basis for many modern day ML algorithms, most notably neural networks. In addition, we’ll discuss the perceptron algorithm’s cousin, logistic regression. And then we’ll conclude with an introduction to SVMs, or support vector machines, which are perhaps one of the most flexible algorithms used today.

Supervised and Unsupervised Algorithms

In machine learning, there are two general classes of algorithms. You’ll remember that in our last post we discussed regression and classification. These two methods fall under the larger umbrella of supervised learning algorithms, one of two classes of machine learning algorithms. The other class of algorithms is called unsupervised algorithms.

Supervised algorithms learn from labeled training data. The algorithms are “supervised” because we know what the correct answer is. If the algorithm receives a bunch of images labeled as apples or oranges it can first guess the object in the image, then use the label to check if its guess is correct.

Unsupervised learning is a bit different in that it finds patterns in data. It works similarly to the way we humans observe patterns (or objects) in random phenomena. Unsupervised learning algorithms do the same thing by looking at unlabeled data. Just like we don’t have a particular goal when looking at an object (other than identifying it), the algorithm doesn’t have a particular goal other than inferring patterns from the data itself.

We’ll talk about unsupervised algorithms in a later blog post. For now, let’s look at a very simple supervised algorithm, called the perceptron algorithm.

Perceptrons

Model of a perceptron

One of the goals of machine learning and AI is to do what humans do and even surpass them. Thus, it makes sense to try and copy what makes humans so great at what they do–their brains.

The brain is composed of billions of interconnected neurons that are constantly firing, passing signals from one neuron to another. Together, they allow us to do incredible things such as recognize faces, patterns, and most importantly, think.

The job of an individual neuron is simple: if its inputs match certain criteria, it fires, sending a signal to other neurons connected to it. It’s all very black and white. Of course, the actual explanation is much more complicated than this, but since we’re using computers to simulate a brain, we only need to copy the idea of how a brain works.

The perceptron mimics the function of neurons in machine learning algorithms. The perceptron is one of the oldest machine learning algorithms in existence. When it was first used in 1957 to perform rudimentary image recognition, the New York Times called it:

the embryo of an electronic computer that [the Navy] expects will be able to walk, talk, see, write, reproduce itself and be conscious of its existence.

We’re still a little far off from that, but the Times did recognize the potential of a perceptron. Perceptrons now form the basis for more complicated neural networks, which we’ll be talking about in the next post.

Like neurons, perceptrons take in several inputs and spit out an output. For example, a perceptron might take as input temperature and try to answer the question, “Should I wear a sweater today?” If the input temperature is below a certain threshold (say, 70˚F), the perceptron will output 1 (yes). If the temperature is above the threshold, the perceptron will output a 0 (no).

Of course, we could consider more variables than just temperature when deciding whether or not to wear a sweater. Like how a biological neuron typically can have more than one input electrical impulse, we can make our perceptrons have multiple inputs. In that case, we’ll have to also weight each input by a certain amount. If we were to use temperature, wind speed, and something completely random like the number of people showering in Hong Kong as our inputs, we would want to use different weights for each input. Temperature would probably have a negative weight, because the lower the temperature the more you should probably wear a sweater. Wind speed should have a positive weight, because the higher the wind speeds the more likely you will need to put on a sweater. And as for the number of people showering in Hong Kong, the weight should probably be zero (unless you normally factor in the number of people taking showers in Hong Kong into your decision making).

But someone from Canada might be very used to the cold, so their threshold for wearing a sweater might be a lower temperature than someone from, say, Australia. To express this, we use a bias to specify the threshold for the Canadian and the Australian. You can think of the bias as a measure of how difficult it is for the perceptron to say ‘yes’ or ‘no’.

The Canadian is fairly insensitive to cold temperatures, and thus has a low bias

The Australian is biased towards hotter temperatures and thus has a higher threshold temperature

Logistic Regression

However, life is not as black and white as perceptrons indicate. There is uncertainty in just about anything, even choosing whether or not to put on a sweater. In reality, we don’t immediately put on a sweater the moment it drops below some predefined temperature. It’s more as if at any temperature we have a certain “chance” of putting on a sweater. Maybe at 45 F somebody will have a 95% chance of putting on a sweater, and at 60 F the same person will have a 30% chance of putting on a sweater.

To better model life’s complexities, we use logistic regression to find these probabilities. This involves fitting a logistic curve (like the one below) to our data. To do this, we again use gradient descent to choose the best parameters for the model. That is, we find parameters that minimize some cost function.

The general form of the logistic model is

$$ h(\theta) = \frac{1}{1+e^{ \theta^Tx}} $$

where is a vector of parameters, is the input variables, and is the model probabilities. For more information, we suggest you check out Andrew Ng’s notes on logistic regression.

Logistic regression and the perceptron algorithm are very similar to each other. It’s common to think of logistic regression as a kind of perceptron algorithm on steroids, in that a logistic model can predict probabilities while a perceptron can only predict yes or no. In fact, taking a logistic model and setting all values less than .5 to zero, and all values above .5 to one gives a very similar result to just the perceptron algorithm.

Support Vector Machines

Support vector machines, or SVMs for short, are a class of machine learning algorithms that have become incredibly popular in the past few years. They are based on a very intuitive idea. Here, we’ll introduce SVMs and go through the key ideas in the algorithm.

Margins

If you remember the section on classification from our last post, we classify data by drawing a line, called a decision boundary, to separate them.

Once we’ve drawn a decision boundary, we can find the margin for each datapoint. The margin for a single data point is the distance from that data point to the decision boundary.

In a way, a margin is a way to determine how confident we are in our classification. If a data point has a large margin, and hence is very far away from the decision boundary we can be pretty confident about our prediction. However, if a data point has a very small margin, and is hence very close to the decision boundary then we probably aren’t as sure of our classification.

Now that we’ve defined margins, we can talk about support vectors. A support vector is a vector from the data point with the smallest margin to the decision boundary. In other words, it’s a vector between that data point that is closest to the decision boundary and the boundary itself. This vector, in fact, any margin, will be perpendicular to the decision boundary because the smallest distance between a point and a line is a perpendicular line.

The idea behind a support vector machine is to classify data by drawing a decision boundary such that it maximizes the support vectors. By maximizing the support vectors, we’re also maximizing the margins in a data set, and thus the decision boundary is as far away as possible from the data points.

Linear Separability

If you’ve played around with the simulation enough, you’ll notice that sometimes the algorithm fails completely. This happens only when the data points aren’t linearly separable. Think of it this way: when you have a set of data points which you can’t draw a straight line to separate them with, then you have a linearly inseparable dataset, and since it’s impossible to draw a line to split them, then the SVM algorithm fails.

So how do we deal with linear inseparability? Turns out we can reformulate our optimization problem. Before, we wanted every single data point to be as far away (to the correct side) from the decision boundary as possible. Now, we’ll allow a data point to stray toward the wrong side of the boundary, but we’ll add a “cost” to that happening (remember cost functions from the last post?). This is something that happens very often in machine learning and is called regularization. It basically allows our model to be a bit more flexible when trying to classify the data. The cost of violating the decision boundaries can be as high or as low as we want it to be, and is controlled by something called the regularization parameter.

Mathematically, we implement regularization by adding a term to our cost function. For instance

could be the regularized cost function, where C (the cost) is a function of all the parameters and all the training data, R is the regularization (or penalty) for each data point, and c is the regularization parameter. A large c would mean the penalty for violating the decision boundary would be very high, and vice versa.

Kernels

This part is a bit mathy, which is why it’s hidden, but it’s also very fascinating. Now the SVMs shown so far can only make straight decision boundaries. But what if we wanted to classify data using a different kind of boundary? Well, that’s where kernels come in. Kernels are (loosely speaking) a fancy term for an inner product.

So why do we need kernels? It turns out that the SVM algorithm never depends on just a data point itself, but rather it depends on the inner product between data points. For that reason, we never need to know the coordinates of a data point, we only need to know the inner products between all the data points.

It turns out that for certain representations of data (or certain coordinate systems), it’s actually much easier to calculate the kernel than it is to calculate the actual data point. For example, we can use data points in an infinite dimensional space. As long as we have a way of computing the inner product, or the kernel, quickly. And luckily, there are ways to do this! One example of this is the Radial Basis Function kernel, which calculates the inner product for a certain infinite dimensional space. The form of the RBF kernel is

$$ K(x, x') = e^{-\frac{||x-x’||}{2\sigma^2}} $$

where x and x’ are the two data points in the original coordinate system, ||x-x’|| is the distance between the two points, and sigma is a parameter that controls the model’s behavior. The feature space for the RBF kernel is actually infinite, and would be completely unrepresentable completely in a computer, but using the kernel we can actually use SVMs in that feature space.

Another example of a kernel is the polynomial kernel, which takes the form

$$ K(x, x') = (x^Tx'+c)^d $$

where x and x' are the two data points, and c and d are parameters for the kernel. In this case, the feature space is not actually infinite, but the kernel is still helpful because it allows us to change our representation of the data.

Intuitively, you can think of a kernel as mapping a set data from one coordinate system to another coordinate system. In the original coordinate system the data may not be linearly seperable at all, whereas in the new coordinate system if we choose the correct kernel, we should get a set a data set is very easily linearly seperable.

Computationally intensive simulation! Press start to begin the simulation. (Unfortunately doesn't work on mobile). Left click to add red data point, shift-left click to add green data point. Click and drag to move the data points around. Use the sliders to adjust the regularization parameter (C), the sigma of the RBF kernel (must be in RBF kernel mode), and the C and Degree for the polynomial kernel (must be in polynomial kernel mode). Press 'k' to switch kernels. Press 'r' to reset the data points.

This simulation is an adaptation of Andrej Karpathy's SVM simulator which can be found here

SVM Simulation!

Start
Browser not supported for Canvas. Get a real browser.

Polynomial Kernel C = 0.0
Polynomial Kernel Degree = 2.0

Things to try with the simulator

  • Use the linear kernel and set the regularization parameter (C) to 1.0e+6. Put two red data points and two green data points. Can the SVM classify linearly inseperable data? How about for other values of C?
  • Try using the RBF kernel with different sigmas. Does the RBF kernel overfit or underfit the data?
  • Try using different degrees polynomial kernels. Does the shape for a second degree polynomial kernel look familiar?

Sources

Andrew Ng’s notes on SVMs



from Hacker News http://ift.tt/YV9WJO
via IFTTT

Is 100% Of "US Warming" Due To NOAA Data Tampering?

http://ift.tt/2iG25Py

Submitted by Tony Heller via RealClimateScience.com,

Climate Central just ran this piece, which the Washington Post picked up on. They claimed the US was “overwhelmingly hot” in 2016, and temperatures have risen 1,5°F since the 19th century.

The U.S. Has Been Overwhelmingly Hot This Year | Climate Central

The first problem with their analysis is that the US had very little hot weather in 2016. The percentage of hot days was below average, and ranked 80th since 1895. Only 4.4% of days were over 95°F, compared with the long term average of 4.9%. Climate Central is conflating mild temperatures with hot ones.

They also claim US temperatures rose 1.5°F since the 19th century, which is what NOAA shows.

Climate at a Glance | National Centers for Environmental Information (NCEI)

The problem with the NOAA graph is that it is fake data. NOAA creates the warming trend by altering the data. The NOAA raw data shows no warming over the past century

The adjustments being made are almost exactly 1.5°F, which is the claimed warming in the article.

The adjustments correlate almost perfectly with atmospheric CO2. NOAA is adjusting the data to match global warming theory. This is known as PBEM (Policy Based Evidence Making.)

The hockey stick of adjustments since 1970 is due almost entirely to NOAA fabricating missing station data. In 2016, more than 42% of their monthly station data was missing, so they simply made it up. This is easy to identify because they mark fabricated temperatures with an “E” in their database.

When presented with my claims of fraud, NOAA typically tries to arm wave it away with these two complaints.

  1. They use gridded data and I am using un-gridded data.
  2. They “have to” adjust the data because of Time Of Observation Bias and station moves.

Both claims are easily debunked. The only effect that gridding has is to lower temperatures slightly. The trend of gridded data is almost identical to the trend of un-gridded data.

Time of Observation Bias (TOBS) is a real problem, but is very small. TOBS is based on the idea that if you reset a min/max thermometer too close to the afternoon maximum, you will double count warm temperatures (and vice-versa if thermometer is reset in the morning.) Their claim is that during the hot 1930’s most stations reset their thermometers in the afternoon.

This is easy to test by using only the stations which did not reset their thermometers in the afternoon during the 1930’s. The pattern is almost identical to that of all stations. No warming over the past century. Note that the graph below tends to show too much warming due to morning TOBS.

NOAA’s own documents show that the TOBS adjustment is small (0.3°F) and goes flat after 1990.

http://ift.tt/2hNyyDt

Gavin Schmidt at NASA explains very clearly why the US temperature record does not need to be adjusted.

You could throw out 50 percent of the station data or more, and you’d get basically the same answers.

One recent innovation is the set up of a climate reference network alongside the current stations so that they can look for potentially serious issues at the large scale – and they haven’t found any yet.

NASA – NASA Climatologist Gavin Schmidt Discusses the Surface Temperature Record

NOAA has always known that the US is not warming.

U.S. Data Since 1895 Fail To Show Warming Trend – NYTimes.com

All of the claims in the Climate Central article are bogus. The US is not warming and 2016 was not a hot year in the US. It was a very mild year.



from Zero Hedge http://ift.tt/qouXdu
via IFTTT

Why Nothing Sticks to Teflon Pans and Cookware

http://ift.tt/2iG0zg3

If you own non-stick cookware, it’s probably coated with Teflon, which, as you probably know already, is the material that makes it hard for food to stick. What you might not know is why it works so well.

Teflon, also known as polytetrafluoroethylene (PTFE), is a polymer with a carbon chain in the center and two fluorines that wrap around the outside. The carbon and fluorine have a strong bond that leads them to primarily stick to each other rather than other substances PTFE comes in contact with. It doesn’t react with most other chemicals and has a low friction coefficient, which makes it ideal as a non-stick coating on cookware and other equipment.

PTFE is typically applied by roughing the equipment’s surface with a sand or chemical blast, applying a primer, spraying on the PTFE and then heating it so everything solidifies. Once it’s solid, it makes a smooth cooking surface that conducts heat, and is slick enough to resist sticking. For more on why it works so well as a cooking surface, check out the video above.

Why Doesn’t Anything Stick to Teflon? | TED-Ed (YouTube)



from Lifehacker http://lifehacker.com
via IFTTT

Obama Set To Announce Economic Sanctions And "Covert Cyber Ops" Against Russia For "Election Hacking"

http://ift.tt/2huWKhF

Just a week after Obama held a press conference announcing that he sent a stern warning to Vladamir Putin regarding his alleged "election hacking" efforts (see "Obama Told Putin To "Cut It Out" On Hacking"), the Washington Post is reporting that the Obama administration is close to announcing a series of economic sanctions and other measures to punish Russia for its "interference" in the 2016 presidential election.  Quoting "U.S. officials," WaPo said that an announcement from the Obama administration could come as early as this week and would likely include "covert cyber operations."

According to WaPo's "sources", the delay in sanctions against Russia have come from Obama's inability to take unilateral actions under current laws.  While Obama previously signed an executive order that would allow him to freeze the assets in the United States of people overseas who have engaged in cyber acts, it only applies to actions that have threatened U.S. national security or financial stability.  Further, per a "senior administration official," use of the existing law would require (1) actual election infrastructure to be designated as 'critical infrastructure' and (2) the administration to prove that such infrastructure was actually "harmed," conditions which the National Security Council say have not been met. 

The White House is still finalizing the details of the sanctions package. Holding up the announcement is an internal debate over how best to adapt a 2015 executive order that gave the president the authority to levy sanctions against foreign actors who carry out cyberattacks against the United States.

 

The order was used as the “stick” in negotiations over a highly-publicized 2015 agreement with China that neither nation would hack the other for economic gain.

 

But officials concluded this fall that the order does not cover the kind of covert influence operation that the Intelligence Community believes Russia carried out during the election — hacking political organizations and leaking stolen emails with the goal of influencing the outcome.

 

The April 2015 order allows the Treasury Department to freeze the assets of individuals or entities who used digital means to damage U.S. critical infrastructure or engage in economic espionage.

 

The National Security Council concluded that it would not be able to use the authority against Russian hackers because their malicious activity did not clearly fit under its terms, which require harm to critical infrastructure or the theft of commercial secrets.

 

“You would (a) have to be able to say that the actual electoral infrastructure, such as state databases, was critical infrastructure, and (b) that what the Russians did actually harmed it,” a senior administration official told The Post. “Those are two high bars.”

Obama Putin

 

Of course, laws are merely suggestions for an Obama administration that has grown quite comfortable legislating through executive action from the White House.  As Zachary Goldman, a sanctions and national security expert at New York University School of Law, points out the current laws simply require the Obama administration to "engage in some legal acrobatics to fit the DNC hack into an existing authority, or they need to write a new authority."

“Fundamentally, it was a low-tech, high-impact event,” said Zachary Goldman, a sanctions and national security expert at New York University School of Law. And the 2015 executive order was not crafted to target hackers who steal emails and dump them on WikiLeaks or seek to disrupt an election. “It was an authority published at a particular time to address a particular set of problems,” he said.

 

So officials “need to engage in some legal acrobatics to fit the DNC hack into an existing authority, or they need to write a new authority,” Goldman said.

 

Administration officials would like Obama to use the power before leaving office to demonstrate its utility.

And, not surprisingly, another administration official points out that “part of the goal here is to make sure that we have as much of the record public or communicated to Congress in a form that would be difficult to simply walk back."  Yes, that is the problem with legislating through executive action rather than acknowledging the will of the American people and trying to work with Congress.

And while Obama and Democrats continue their crusade to deligitamize the Trump administration, we would point out once again that, despite all the rhetoric, not a single person has gone on the record and/or presented a single shred of tangible evidence to confirm Russian involvement in the DNC and/or John Podesta email hacks



from Zero Hedge http://ift.tt/qouXdu
via IFTTT