Mediaboss Marketing

Introduced by Google in 2017, Federated Learning (FL) enables mobile phones to collaboratively learn a shared prediction model while keeping all the training data on the device, decoupling the ability to do machine learning from the need to store the data in the cloud. Two years have passed, and several new research papers have proposed novel systems to boost FL performance. This March for example a team of researchers from Google suggested a scalable production system for FL to enable increasing workload and output through the addition of resources such as compute, storage, bandwidth, etc.

Earlier this month, NeurIPS 2019 in Vancouver hosted the workshop Federated Learning for Data Privacy and Confidentiality,where academic researchers and industry practitioners discussed recent and innovative work in FL, open problems and relevant approaches.

Professor Dr. Max Welling is the research chair in Machine Learning at the University of Amsterdam and VP Technologies at Qualcomm. Welling is known for his research in Bayesian Inference, Generative modeling, Deep Learning, Variational autoencoders, Graph Convolutional Networks.

Below are excerpts from the workshop talk Dr. Welling gave on Ingredients for Bayesian, Privacy Preserving, Distributed Learning, where the professor shares his views on FL, the importance of distributed learning, and the Bayesian aspects of the domain.

The question can be separated in two parts. Why do we need distributed or federated inferencing? Maybe that is easier to answer. We need it because of reliability. If you in a self-driving car, you clearly dont want to rely on a bad connection to the cloud in order to figure out whether you should brake. Latency. If you have your virtual reality glasses on and you have just a little bit of latency youre not going to have a very good user experience. And then theres, of course, privacy, you dont want your data to get off your device. Also compute maybe because its close to where you are, and personalization you want models to be suited for you.

It took a little bit more thinking why distributed learning is so important, especially within a company how are you going to sell something like that? Privacy is the biggest factor here, there are many companies and factories that simply dont want their data to go off site, they dont want to have it go to the cloud. And so you want to do your training in-house. But theres also bandwidth. You know, moving around data is actually very expensive and theres a lot of it. So its much better to keep the data where it is and move the computation to the data. And also, personalization plays a role.

There are many challenges when you want to do this. The data could be extremely heterogeneous, so you could have a completely different distribution on one device than you have on another device. Also, the data sizes could be very different. One device could contain 10 times more data than another device. And the compute could be heterogeneous, you could have small devices with a little bit of compute that now and then or you cant use because the batterys down. There are other bigger servers that you also want to have in your in your distribution of compute devices.

The bandwidth is limited, so you dont want to send huge amounts of even parameters. Lets say we dont move data, but we move parameters. Even then you dont want to move loads and loads of parameters over the channel. So you want to maybe quantize it to see this. I believe Bayesian thinking is going to be very helpful. And again, the data needs to be private so you wouldnt want to send parameters that contain a lot of information about the data.

So first of all, of course, were going to move model parameters, were not going to move data. We have data stored at places and were going to move the algorithm to that data. So basically you get your learning update, maybe privatized, and then you move it back to your central place where youre going to update it.And of course, bandwidth is another challenge that you have to solve.

We have these heterogeneous data sources and we have very variability in the speed in which we can sync these updates. Here I think the Bayesian paradigm is going to come in handy because, for instance, if you have been running an update on a very large dataset, you can shrink your posterior parameters to a very small posterior. Where on another device, you might have much less data, and you might have a very wide posterior distribution for those parameters. Now, how to combine that? You shouldnt average them, its silly. You should do a proper posterior update where the one that has a small peaked posterior has a lot more weight than the one with a very wide posterior. Also uncertainty estimates are important in that aspect.

The other thing is that with Bayesian update, if you have a very wide posterior distribution, then you know that parameter is not going be very important for making predictions. And so if youre going to send that parameter over a channel, you will have to quantize it, especially to save bandwidth. The ones that are very uncertain anyway you can quantize at a very coarse level, and the ones which have a very peak posterior need to be encoded very precisely, and so you need much higher resolution for that. So also there, the Bayesian paradigm is going to be helpful.

In terms of privacy, there is this interesting result that if you have an uncertain parameter and you draw a sample from that posterior parameter, then that single sample is more private than providing the whole distribution. Theres results that show that you can get a certain level of differential privacy by just drawing a single sample from that posterior distribution. So effectively youre adding noise to your parameter, making it more private. Again, Bayesian thinking is synergistic with this sort of Bayesian federated learning scenario.

We can do MCMC (Markov chain Monte Carlo) and variational based distributed learning. And as theres advantages to do that because it makes the updates more principled and you can combine things which, one of them might be based on a lot more data than another one.

Then we have private and Bayesian to privatize the updates of a variational Bayesian model. Many people have worked on many other of these intersections, so we have deep learning models which have been privatized, we have quantization, which is important if you want to send your parameters over a noisy channel. And its nice because the more you quantize, the more private things become. You can compute the level of quantization from your Bayesian posterior, so all these things are very nicely tied together.

People have looked at the relation between quantized models and Bayesian models how can you use Bayesian estimates to quantized better? People have looked at quantized versus deep to make your deep neural network run faster on a mobile phone you want to quantize it. People have looked at distributed versus deep, distributed deep learning. So many of these intersections have actually been researched, but it hasnt been put together. This is what I want to call for. We can try to put these things together and at the core of all of this is Bayesian thinking, we can use it to execute better on this program.

Journalist: Fangyu Cai | Editor: Michael Sarazen

Like Loading...

See the article here:

Dr. Max Welling on Federated Learning and Bayesian Thinking - Synced

Founding Father and principle architect of the U.S. Constitution James Madison stated unequivocally: Charity is no part of the legislative duty of the government.

Libertarians strongly agree with Madison and we believe that people in need are best served by grassroots private volunteer organizations, rather than by government-run, taxpayer-at-large funded operations, which tend to be heavy on bureaucracy and regulation rather than on actually delivering results and aid to those in need.

In the spirit of James Madison, the Gila County Libertarian Party has made a financial contribution to the Payson Homeless Initiative, and we encourage other political groups in the area as well as non-political civic and charitable groups to get behind the Payson Homeless Initiative and make donations or provide volunteers to help this truly grassroots private sector effort.

Americans are the most generous and charitable people on the planet, and we would do even more if we retained more of the fruits of our labor. Less taxes = more money in private hands = more contributions to productive, private voluntary charitable programs. With less of a tax burden, caring members of society will be even more charitable, and society will be strengthened.

LBJs War on Poverty has been an epic failure. 55 years later, homelessness is rampant, especially in California. Tens of thousands of people gather in homeless camps with diseases such as tuberculosis and hepatitis posing great risk to the public at large. There is a better way, and private voluntary organizations such as the Payson Homeless Initiative are a prime example.

The Gila County Libertarian Party extends a well-deserved thank you to all the individuals and organizations that are already volunteering and contributing to the Payson Homeless Initiative. Keep up the great work!

Larry Hoffenberg, secretary/treasurer, Gila County Libertarian Party

Go here to read the rest:
Kudos to the Payson Homeless Initiative | Letters To Editor - Payson Roundup

Conservative: Thanks for the Judges, Harry!

With Senate Majority Leader Mitch McConnells help, President Trump has appointed federal judges at about twice the rate of his three predecessors, notes The Washington Examiners editorial board. But Trump should be thanking McConnells predecessor, former Democratic Sen. Harry Reid of Nevada. In 2013, with President Barack Obama in the White House and Democrats controlling the Senate, Reid made the fateful and short-sighted decision to change Senate rules so that a bare majority was enough to confirm a judge, instead of 60 votes, as before. During his campaign, Trump regularly and energetically promised to appoint well-credentialed conservatives with excellent character and scholarship to judgeships a promise he has kept, much to his credit.

Culture critic: From Woodstock to Populism

Middle class in Britain was once defined by a safe, lifelong career, allegiance to the Conservative Party and defending tradition but now, Jonathan Rutherford sighs at The New Statesman, it has lost its role and the authority invested in it and has been overtaken by a new middle class fraction forged in the cultural revolution and university expansion of the 1960s. The Woodstock generation went into politics, eschewing traditionally left-wing populist economic democracy in favor of a libertarian identity politics of gender, race and sexuality. Left-wing parties became parties of the new liberal middle class, increasingly contemptuous of lives and experience of mainstream working-class voters. Yet working-class voters pushed back and voted for Brexit and their historic class enemy: the Tories. Back in 1969, no one could have believed it would turn out like this, but liberal elites have only themselves to blame.

Foreign desk: Hurrah for the US-UK Marriage

Among elite opinion-makers, Brexit is destined to turn Britain into an isolated backwater. Not so, says Brandon J. Weichert at American Greatness. The island nation had extraordinary power on its own, and subordinating British national sovereignty to the supranational government in Brussels was always a mistake. Now that its almost out, Britain should forge a stronger relationship with the United States, an Anglo-American marriage that would ensure that Brexit is meaningful and real and not at all damaging to Britain. The good news is that President Trump has already promised a new free-trade agreement with London which will allow Britain to shake off the sclerotic superstate that is the European Union.

Libertarian: A Year of Peak Entitlement

If you listen to many politicians and pundits, you would think the United States is doing terribly while the government isnt spending a dime yet the truth is the exact opposite, argues Reasons Veronique de Rugy. Among other things, the economy is entering its 11th year of expansion, while poverty is at an all-time low, and the unemployment rate hasnt been so low since 1969. Meanwhile, the government is racking up gargantuan budget deficits, largely because both political parties are spending on a whim and condemning our free-market economy the very system that has produced the wealth that everyone takes for granted. The problem, she insists, isnt that free markets dont work, but that we may have reached peak entitlement mentality.

Urban beat: Calis Homelessness Hopelessness

Despite Californias homelessness crisis, Sacramento and city halls across the Golden State are mired in the we-need-more-money mindset a mindset, sighs Issues and Insights editorial board, that has never worked. In fact, despite all the spending, and the pleas and plans for additional money, homelessness has spiked 30% since 2017 in San Francisco, 16% in Los Angeles and a whopping 43% in San Jose. As a result, nearly half of the nations homeless who sleep on the streets today do so in California. Instead of feeding government bureaucracies with taxpayers money, government officials should follow the example of San Diego, where the city took a tough-love approach that rejected widespread street camping and watched its homeless population fall. The shift in thinking and in acting is paying off.

Compiled by Karl Salzmann

Read more from the original source:
Thanks for the judges, Harry Reid, and other commentary - New York Post

Colorados red flag law will go into effect Jan. 1, 2020 -- its the one that allows a judge to temporarily remove somebodys firearms if theyre a danger to themselves or others.

The law has raised a lot of questions for Coloradans about how itll be implemented, who will enforce it and how its different from other states with similar laws.

CPR News spoke with two experts on gun rights and gun laws to address some concerns. Dave Kopel teaches law at the University of Denver and is with the Independence Institute, a Libertarian think tank. Shannon Frattaroli works in the Center for Gun Policy and Research at the Johns Hopkins School For Public Health, where researchers are looking at ways to reduce gun violence.

These are some of the points of the law and its implementation that theyre watching.

Colorado is just one of 17 states and the District of Columbia who have some version of a red flag law. Frattaroli said Colorados law is similar to other states, for the most part.

What has been different about Colorado is the response amongst some within the state and the level of concern about due process in Colorado is certainly something that we haven't seen in other places around the country, she said.

Kopel said that from his perspective, Colorados law is better than other states in one way but worse in a number of other ways.

The way it's better is Colorado has a provision that says if somebody is accused under this law when they finally get into court, which is only later in the process, that if they want an attorney to be provided by the government, the government will provide that, he said.

Kopel said it will ultimately be up to courts to decide whether or not its unconstitutional. But he thinks it does violate due process in a common-sense way.

When a firearm is removed from somebody, only the accusers side of information is presented in court, Kopel said. There is also no opportunity for the accuser to be cross-examined and questioned.

That's a fundamental thing in any fair system is when somebody makes an accusation, the accused person, if they want, can have a lawyer cross-examine the accuser and may bring out inconsistencies of the story and give the court the opportunity to observe the accuser in person and make a determination about the accusers credibility, Kopel said.

Frattaroli said there are other protections put in place that shield those being accused.

The judge listens to the case, listens to the perspective of the petitioner and considers that case and the risks to the public, the risks to the person, she said. Then there's this secondary line of due process where the person has an opportunity that's built into the system to be heard in court.

She compared those protections to the same ones used in restraining orders for domestic violence cases.

The law says only law enforcement and family or household members can ask a judge to remove somebodys firearms. Lesley Hollywood, founder of the gun-rights group Rally For Our Rights, strongly opposes the law and said the definition of who counts as a family member is broad.

There's a lot of concern that people will be falsely reported by ex-husbands former roommates, dating partners, people they've had an affair with, she said. One of the things I worry about as a woman is that stalkers and abusers can actually use it to have their victims disarmed.

Frattaroli said a family or household member would not include a dating partner unless they're living with them. Only a domestic partner who has lived with the respondent within the last six months can file a petition, according to the law. If somebody has a child with a respondent, they can file a petition at any time.

A domestic partner can also file a petition but its not clear if theres a statute of limitations. Kopel argued somebody who dated a person 15 years ago, for example, could file a petition.

In order for an extreme risk protection order to be considered by the court, a petitioner would have to fill out paperwork, go to court to submit it and then appear in front of a judge to have the case heard, Frattaroli said.

That isn't an easy thing to do, she said. It's not something that I think people enjoy doing.

She said from talking with law enforcement around the country who deal with extreme risk protection orders and see how theyre actually used, false reports havent come through.

Kopel said the law is structured in a way that promotes false accusations. An earlier version of the bill had a protection against false accusations. He said the likelihood of somebody being prosecuted for perjury is not high.

Kopel said it would depend on the case and evidence presented.

Frattaroli said the law isnt designed to target specific groups of people and whether or not someone has PTSD is irrelevant.

It doesn't have anything to do with if someone is depressed or if someone is sort of unpopular, she said. It's all about what are people doing? What are people saying and are those actions, are those words, do they constitute a real actionable threat to their own safety or to the safety of others?

Kopel compared law enforcement refusing to comply with other sanctuary issues.

You can argue about it all you want, but it's pretty hard to make government officials do something they don't want to do, especially when there's a lot of support within their jurisdiction for them not wanting to enforce a law that they and a lot of the public see is abusive and unfair.

Frattaroli said she hasnt seen as much pushback elsewhere from law enforcement about a red flag law as she has in Colorado. She said law enforcement and the community will ultimately decide if and how theyll use the extreme risk protection order as a tool for when somebody is in crisis or threatening harm.

Frattaroli said its hard to look back and say definitively if it would have made a difference but cases like Parishs are examples of the driving force behind extreme risk laws.

Kopel cited Douglas County Sheriff Tony Spurlock, who thinks it would have.

He knows more about the facts of that particular case then than I do so I wouldn't disagree with him, he said. And that's why I'm in favor of the (red flag) laws if they're well structured, which this one is not.

Read more from the original source:
Colorado's Red Flag Law Goes Into Effect Jan. 1. This Is What People Are Watching For - Colorado Public Radio

The Stark County Board of Elections denied to certify two candidates in the race the 7th Congressional District due to petition signature issues.

CANTON U.S. Rep. Bob Gibbs, R-Lakeville, is now the only candidate for Congress to represent the 7th Congressional District on the March 17 primary ballot.

On Monday morning, the Stark County Board of Elections certified a ballot that didnt include the names of the Democratic and Libertarian candidates for Congress representing the 7th. The board staff said they had failed to submit enough valid signatures.

However, a Knox County elections official said her office mistakenly designated 10 signatures submitted by a Libertarian candidate as invalid.

Brandon Lape, 38, of Danville, filed a formal request Monday afternoon asking the Stark County Board of Elections to reverse its decision to kick him off the ballot.

Three candidates met Wednesdays filing deadline to seek their partys nomination to be congressman for the 7th District, which includes most of Stark County.

They were Gibbs, Lape and Democrat Patrick Pikus, of Plain Township, who also ran in 2018.

Gibbs and Pikus, as members of major parties, had to submit 50 valid petition signatures by registered voters in their party who lived in the congressional district.

Lape, 38, as a member of a minor party, had to submit 25 valid petition signatures.

Pikus submitted the minimum of 50 but board staff found nine signatures to be invalid.

Pikus, 54, a business manager for the Timken Co., said he accepted responsibility for "rookie mistake. ... I thought I had 50 good ones."

He said he had been reluctant to run again, but when he saw that no other Democrat was apparently interested in running, he started collecting signatures and "took them at their word that they were registered (to vote)."

"Nobody is at fault besides me for not double checking," he said, adding that he would review his options.

Because all of the people who signed Lapes petition lived in Ashland and Knox counties, the Ashland and Knox county boards of election verified the signatures. Stark County, as the most populous county in the 7th District, decides whose candidacies to certify.

Elections staff in Ashland and Knox counties initially invalidated 23 of 42 petition signatures submitted by Lape, leaving him with only 19, six short of the minimum. They found two signatures were by people not registered to vote, two were not registered to vote from the addresses provided and one signature was invalid because it was by the petition circulator. They found 18 were signed by voters who had voted in a recent Republican or Democratic primary, which made them ineligible to sign Lapes petition.

But Kim Horn, the director of the Knox County Board of Elections, said hours later her office mistakenly concluded that 10 of candidate Lapes signatures were invalid. Horn said her offices voter registration database wrongly designated several voters as Republican or Democratic voters when they hadnt voted in a Republican or Democratic primary since 2016.

By law, a person ceases to be affiliated with a political party if they dont vote in that partys primary for at least two calendar years.

Regine Johnson, the Stark County Board of Elections deputy director, said the board would choose whether to hold the hearing of reconsideration. If it takes place, it would likely happen by Jan. 13.

Stark County Commissioner Janet Weir Creighton, a Republican whos not up for re-election in 2020, said in an interview about the petition process this month that she always advises candidates to get more signatures than the minimum. And to have an official at party headquarters check the signatures against the rolls of registered voters.

"If youre required to get 50 good signatures, you dont get 50. You get more than what you need," she said. "If you cant follow these rules then I question why youre running for office. ... I would be sick if it cost my candidacy because of an error."

Creighton added the earlier petition filing deadline made it harder to get signatures as people were more focused on the holidays.

Write-in candidates have until Jan. 6 to file for the March 17 primary. They have to fill out a form but do not have to submit signatures. They are not listed on the ballot. In the Democratic primary for 7th Congressional District congressman, which now has no candidates, the write-in candidate who has at least 50 voters write in their name on the ballot and wins a plurality of the write-in vote would then become the Democratic candidate. That persons name would be listed on the ballot in November along with Gibbs name, said Travis Secrest, an administrative assistant for the Stark County Board of Elections

Non-partisan candidates who wish to run for Congress and be listed on the November ballot have until March 16 to file, said Secrest. They must submit at least 2,616 valid signatures by any registered voters in the district.

Candidates denied a place on the ballot due to signature issues are not eligible to run as write-in candidates or non-partisan candidates.

The Board also chose to leave off the ballot liquor options for Aldi Ohio in the Canton 6-C precinct and the Palace Theatre in Canton 2-B due to both entities submitting an insufficient number of valid petition signatures.

In addition, the Board also declined to certify three candidates seeking party committee positions due to an insufficient number of valid signatures. That included Patrick J. Glasgow, who was seeking to be the male member on the Libertarian Party State Central Committee for the 7th Congressional District. Glasgow would have been unopposed. And also denied a spot on the primary ballot were Gloria Ann Jeter of precinct Canton 6-C and Patrick Hoch of Canton 7-F for Stark County Democratic Party Central Committee.

Patrick Dorosky is now the only candidate for the Canton 6-C Stark County Democratic Party Central Committeeman. No one else sought the Canton 7-F seat. Eligible write-in candidates may seek the spot or if no one does Stark County Democratic Party

Chairman Samuel J. Ferruccio, whos also the chairman of the Stark County Board of Elections, can appoint any Democratic registered voter in Canton 7-F to fill the seat.

Reach Repository writer Robert Wang at (330) 580-8327 or robert.wang@cantonrep.com. Twitter: @rwangREP.

Read this article:
Bob Gibbs now unopposed in 2020; other candidates removed from ballot - Canton Repository