Nosimpler

The brain actively represents incoming information, but these representations are only useful to the extent that they flexibly reflect changes in the environment. How does the brain transform representations across changes, such as in size or viewing angle? We conducted a fMRI experiment and a magnetoencephalography experiment in humans (both sexes) in which participants viewed objects before and after affine viewpoint changes (rotation, translation, enlargement). We used a novel approach, representational transformation analysis, to derive transformation functions that linked the distributed patterns of brain activity evoked by an object before and after an affine change. Crucially, transformations derived from one object could predict a postchange representation for novel objects. These results provide evidence of general operations in the brain that are distinct from neural representations evoked by particular objects and scenes.

SIGNIFICANCE STATEMENT The dominant focus in cognitive neuroscience has been on how the brain represents information, but these representations are only useful to the extent that they flexibly reflect changes in the environment. How does the brain transform representations, such as linking two states of an object, for example, before and after an object undergoes a physical change? We used a novel method to derive transformations between the brain activity evoked by an object before and after an affine viewpoint change. We show that transformations derived from one object undergoing a change generalized to a novel object undergoing the same change. This result shows that there are general perceptual operations that transform object representations from one state to another.

Google unveils search engine for open data

Google unveils search engine for open data, Published online: 05 September 2018; doi:10.1038/d41586-018-06201-x

The tool, called Google Dataset Search, should help researchers to find the data they need more easily.

Probably the most famous of Grothendieck's contributions to Banach space theory is the result that he himself described as "the fundamental theorem in the metric theory of tensor products". That is now commonly referred to as "Grothendieck's theorem" (GT in short), or sometimes as "Grothendieck's inequality". This had a major impact first in Banach space theory (roughly after 1968), then, later on, in C∗-algebra theory, (roughly after 1978). More recently, in this millennium, a new version of GT has been successfully developed in the framework of "operator spaces" or non-commutative Banach spaces. In addition, GT independently surfaced in several quite unrelated fields:\ in connection with Bell's inequality in quantum mechanics, in graph theory where the Grothendieck constant of a graph has been introduced and in computer science where the Grothendieck inequality is invoked to replace certain NP hard problems by others that can be treated by "semidefinite programming" and hence solved in polynomial time. In this expository paper, we present a review of all these topics, starting from the original GT. We concentrate on the more recent developments and merely outline those of the first Banach space period since detailed accounts of that are already available, for instance the author's 1986 CBMS notes.

by Charles Mudede
According to the New York Times, the US government is surreptiously relocating migrant children to a "tent city" in Tornillo, Texas, which is near El Paso. This location is new and contains "rows of sand-colored tents." When it opened in June, it was meant to hold only 400 migrants, but it was expanded in September to hold nearly 4,000. Children staying at standard shelters around the US are being bussed there in the middle of the night. Immigration officials rudely wake the boys and girls up and force them into the dark unknown.

From the New York Times:

Several shelter workers, who spoke on condition of anonymity for fear of being fired, described what they said has become standard practice for moving the children: In order to avoid escape attempts, the moves are carried out late at night because children will be less likely to try to run away. For the same reason, children are generally given little advance warning that they will be moved.

As Americans sleep and dream, terrified children are being transported to a concentration camp.

But the mass transfers are raising the alarm among immigrant advocates, who were already concerned about the lengthy periods of time migrant children are spending in federal custody.

The roughly 100 shelters that have, until now, been the main location for housing detained migrant children are licensed and monitored by state child welfare authorities, who impose requirements on safety and education as well as staff hiring and training.

The tent city in Tornillo, on the other hand, is unregulated, except for guidelines created by the Department of Health and Human Services. For example, schooling is not required there, as it is in regular migrant children shelters.

The camp in Tornillo operates like a small, pop-up city, about 35 miles southeast of El Paso on the Mexico border, complete with portable toilets. Air-conditioned tents that vary in size are used for housing, recreation, and medical care. Originally opened in June for 30 days with a capacity of 400, it expanded in September to be able to house 3,800, and is now expected to remain open at least through the end of the year.

Those who think that the association with concentration camps is nothing but alarmist liberal nonsense, please read this sentence carefully: "The children [wear] belts etched in pen with phone numbers for their emergency contacts." Can you feel that? Trump's America is not fucking around. The camp, unlike the shelters, also offers few professional services or support. We are basically dumping children into a social black hole during the most informative years of their lives.

The New York Times:

The tent city in Tornillo... is unregulated, except for guidelines created by the Department of Health and Human Services. For example, schooling is not required there, as it is in regular migrant children shelters

None of this will end well for the children and the soul of this very rich nation. A process instituted by monsters will likely produce monsters.

[ Comment on this story ]

[ Subscribe to the comments on this story ]

Author(s): Matthew T. Eiles, Zhengjia Tong, and Chris H. Greene

A series of electric and magnetic pulses applied to an atom could cause one of its electrons to behave as if “bonded” to an empty point in space.

[Phys. Rev. Lett. 121, 113203] Published Wed Sep 12, 2018

Deep learning has been transforming our ability to execute advanced inference tasks using computers. Here we introduce a physical mechanism to perform machine learning by demonstrating an all-optical diffractive deep neural network (D²NN) architecture that can implement various functions following the deep learning–based design of passive diffractive layers that work collectively. We created 3D-printed D²NNs that implement classification of images of handwritten digits and fashion products, as well as the function of an imaging lens at a terahertz spectrum. Our all-optical deep learning framework can perform, at the speed of light, various complex functions that computer-based neural networks can execute; will find applications in all-optical image analysis, feature detection, and object classification; and will also enable new camera designs and optical components that perform distinctive tasks using D²NNs.

The self-driving cars were Uber's hail mary pass. If they get rid of that... then? Their path to profitability always required establishing monopoly - and then jacking up rates - somehow, but their path to monopoly was never clear. I don't even think self-driving cars were the path. Just another shiny object to flash to investors. Now the (some) investors are done.

Some investors have told Uber officials that it may be wise to divest the self-driving car unit, said a person familiar with the issue. Uber has invested least $2 billion in the unit over the past three years. Yet the company hasn’t yet come up with a clear path to commercializing the technology it has developed.

Will be interesting if the two major taxi app companies (Uber, Lyft) fold.

For most of us, anthrax evokes fearful memories of white powder in envelopes. The disease, however, is an ancient one. God’s fifth plague upon the Egyptians — ‘‘Behold, the hand of the Lord is upon thy cattle which is in the field, upon the horses, upon the asses, upon the camels, upon the oxen, and upon the sheep: there shall be a very grievous murrain,” Moses told Pharaoh — may well have been an anthrax outbreak. The same goes for Apollo’s bane upon the Greeks at the beginning of the Iliad. (Homer dubbed the disease “the burning wind of plague.”) Perhaps the most striking description from antiquity of what we now know as Bacillus anthracis comes from Virgil’s Georgics:

Nor was the manner of dying a simple matter: 
After the thirsty slake-seeking fever had gone 
All through the veins and withered the pitiful limbs, 
Then a fluid welled up in the suffering body, and 
Piece by piece absorbed the melting bones. 

B. anthracis is a cruel organism. In their passive form, the bacteria live as hard, oval-shaped spores with thick, nearly indestructible walls that allow them to survive for decades. When the spores colonize a victim’s bloodstream, they enter a vegetative state, dissolving their walls and gathering into neat chains that Robert Koch, the nineteenth-century German scientist whose pioneering work helped identify the disease, described as “graceful, artificially ordered strings of pearls.” In order to survive, the bacteria must kill the host and reproduce inside it before escaping back into the world and returning to a resting state.

Anthrax bacteria produce two lethal toxins in tandem, akin to those that cause tetanus and cholera. The process tends to be swift, and the chances of fatality high. The early symptoms resemble those of the common flu: your head begins to ache; your temperature rises; a general sense of weakness envelops your body; your stomach starts rumbling; you begin to cough incessantly. Then things get serious: you may go into seizures; your organs begin failing; boils break out across your skin, swelling red pustules with a trademark black center. In the fifth century bc, Hippocrates dubbed the disease anthrakes, from the ancient Greek for “charcoal.” 

The disease has triumphed once the blood begins spilling from your orifices. When the medical examiners or the veterinarians cut you open, they will find that your blood has gone black, and that certain organs, particularly the spleen, have turned into masses of melting flesh.

Now the melting of the Siberian permafrost is unleashing anthrax bacilli that have been frozen there for centuries.  Vast herds of reindeer are being decimated and a way of life destroyed for native subarctic peoples.  Details in a longread at Harper's Magazine.

Josh Mitchell, a consultant at the security firm Nuix, analyzed five body camera models from five different companies: Vievu, Patrol Eyes, Fire Cam, Digital Ally, and CeeSc. The companies all market their devices to law enforcement groups around the US...

In all but the Digital Ally device, the vulnerabilities would allow an attacker to download footage off a camera, edit things out or potentially make more intricate modifications, and then re-upload it, leaving no indication of the change. Or an attacker could simply delete footage they don't want law enforcement to have...

Additionally, Mitchell says that some of the more sophisticated models, which contain radios for Bluetooth or cellular data connectivity, also have vulnerabilities that can be exploited to remotely stream live footage off the cameras, or to modify, add, and delete the footage stored on the devices...

Then, when the camera connects to a PC for syncing, it could deliver all sorts of malicious code: a Windows exploit that could ultimately allow an attacker to gain remote access to the police network, ransomware to spread across the network and lock everything down, a worm that infiltrates the department's evidence servers and deletes everything, or even cryptojacking software to mine cryptocurrency using police computing resources...

"These are full-feature computers walking around on your chest, and they have all of the issues that go along with that."

Via BoingBoing.

Just need to keep getting gullible local officials to throw money at you for a "study." No good way to get to Dodger stadium so...how about low capacity cars on sleds! They're good for everything!

The Boring Company is proposing to build Dugout Loop, a zero-emissions, high-speed, underground public transportation system from the Los Feliz, East Hollywood, or Rampart Village neighborhoods ("western terminus") to Dodger Stadium in the City of Los Angeles.

Let's take a look. The route is 3.6 miles.

Loop is a zero-emissions, high-speed underground public transportation system in which passengers are transported on autonomous electric skates traveling at 125-150 miles per hour. Electric skates will carry between 8 and 16 passengers.

I highly doubt they'll travel this fast, but the real point is that it doesn't matter. Boarding is the real bottleneck for things like this. Picture the taxi line at the airport, or the line for the rollercoaster. That's what you get when you can only board a dozen people at a time. The line's gonna be long, Brant.

Oh, sorry, line? No there will be an app for that which will totally solve this problem (hahahaahahaha).

Initially, riders will be able to reserve times and purchase Dugout Loop tickets in advance similar to booking seats at a movie theater via a mobile app, over the phone, or in person (e.g. 5:45pm PT Dugout Loop ticket).

Remember this is primarily a baseball game transportation device. What time would you like to go to the baseball game? And sure, arrivals can be staggered a bit, but everybody wants to leave at the same time...enjoy the line!

Initially, Dugout Loop will be limited to approximately 1,400 people (approximately 2.5% of Stadium capacity) per event.

I love how it doesn't even say "per hour" but "per event" which probably includes at least a 2 hour window (guessing!). One real subway train can easily carry 1000, board them all quickly, and you can run one ever 2 minutes. One attraction to these "sleds" is the weird idea that if you have lower capacity you can run them more often, but headways aren't really a technical constraint of subway systems. Any modern subway system can run 24 trains per hour easy, and plenty do 32. At that point it's the boarding time that makes running them more often be impractical. Even our pretty antiquated trolley system in Philly runs through the tunnel with <3 minute headways at peak, and they carry about 70 people per train.

Electric skates are zero-emission vehicles, and thus do not output hazardous gases like internal combustion cars do.

Wow electric powered underground vehicles. What will Elon think of next?

The fares are not finalized but will cost around $1.

Their own projections put it at 250,000 riders per year. Let's say each does roundtrip, so 500,000 total. Time for some math. All that grad school must have been good for something. Let's see if I remember how to do this.

Oh yes. 500,000×$1= $500,000. Sure most transit systems are subsidized, but, uh...

The post Spiders Can Fly! appeared first on Marginal REVOLUTION.

The post Every law is violent appeared first on Marginal REVOLUTION.

JUST IN: Rep. Chris Collins (R-NY), Trump’s earliest congressional backer, arrested by the FBI on securities fraud-related charges, via @jonathan4ny

— Peter Alexander (@PeterAlexander) August 8, 2018

This paper explores the physics of the what-if question "what if the entire Earth was instantaneously replaced with an equal volume of closely packed, but uncompressed blueberries?" While the assumption may be absurd, the consequences can be explored rigorously using elementary physics. The result is not entirely dissimilar to a small ocean-world exoplanet.

Related Articles

Kick Control: Using the Attracting States Arising Within the Sensorimotor Loop of Self-Organized Robots as Motor Primitives.

Front Neurorobot. 2018;12:40

Authors: Sándor B, Nowak M, Koglin T, Martin L, Gros C

Abstract
Self-organized robots may develop attracting states within the sensorimotor loop, that is within the phase space of neural activity, body and environmental variables. Fixpoints, limit cycles and chaotic attractors correspond in this setting to a non-moving robot, to directed, and to irregular locomotion respectively. Short higher-order control commands may hence be used to kick the system from one self-organized attractor robustly into the basin of attraction of a different attractor, a concept termed here as kick control. The individual sensorimotor states serve in this context as highly compliant motor primitives. We study different implementations of kick control for the case of simulated and real-world wheeled robots, for which the dynamics of the distinct wheels is generated independently by local feedback loops. The feedback loops are mediated by rate-encoding neurons disposing exclusively of propriosensoric inputs in terms of projections of the actual rotational angle of the wheel. The changes of the neural activity are then transmitted into a rotational motion by a simulated transmission rod akin to the transmission rods used for steam locomotives. We find that the self-organized attractor landscape may be morphed both by higher-level control signals, in the spirit of kick control, and by interacting with the environment. Bumping against a wall destroys the limit cycle corresponding to forward motion, with the consequence that the dynamical variables are then attracted in phase space by the limit cycle corresponding to backward moving. The robot, which does not dispose of any distance or contact sensors, hence reverses direction autonomously.

PMID: 30050427 [PubMed]

Heroes.

In America’s less affluent neighborhoods, low-income minorities are often left to grapple with the gentrification juggernaut all on their own. But these marginalized people will be happy to know they have an ally in Ryan and Caitlin Hubbard and their son, Corey—a white, upper-middle-class family that is helping to fight the good fight by refusing to live anywhere that isn’t already 100 percent white.

We introduce an all-optical Diffractive Deep Neural Network (D2NN) architecture that can learn to implement various functions after deep learning-based design of passive diffractive layers that work collectively. We experimentally demonstrated the success of this framework by creating 3D-printed D2NNs that learned to implement handwritten digit classification and the function of an imaging lens at terahertz spectrum. With the existing plethora of 3D-printing and other lithographic fabrication methods as well as spatial-light-modulators, this all-optical deep learning framework can perform, at the speed of light, various complex functions that computer-based neural networks can implement, and will find applications in all-optical image analysis, feature detection and object classification, also enabling new camera designs and optical components that can learn to perform unique tasks using D2NNs.

Excerpts from a stunning article at Vice's Motherboard:

The nation's top voting machine maker has admitted in a letter to a federal lawmaker that the company installed remote-access software on election-management systems it sold over a period of six years, raising questions about the security of those systems and the integrity of elections that were conducted with them.

In a letter sent to Sen. Ron Wyden (D-OR) in April and obtained recently by Motherboard, Election Systems and Software acknowledged that it had "provided pcAnywhere remote connection software … to a small number of customers between 2000 and 2006," which was installed on the election-management system ES&S sold them.

The statement contradicts what the company told me and fact checkers for a story I wrote for the New York Times in February. At that time, a spokesperson said ES&S had never installed pcAnywhere on any election system it sold. "None of the employees, … including long-tenured employees, has any knowledge that our voting systems have ever been sold with remote-access software," the spokesperson said.

ES&S did not respond on Monday to questions from Motherboard, and it’s not clear why the company changed its response between February and April. Lawmakers, however, have subpoena powers that can compel a company to hand over documents or provide sworn testimony on a matter lawmakers are investigating, and a statement made to lawmakers that is later proven false can have greater consequence for a company than one made to reporters...

ES&S is the top voting machine maker in the country, a position it held in the years 2000-2006 when it was installing pcAnywhere on its systems. The company's machines were used statewide in a number of states, and at least 60 percent of ballots cast in the US in 2006 were tabulated on ES&S election-management systems...

Election-management systems are not the voting terminals that voters use to cast their ballots, but are just as critical: they sit in county election offices and contain software that in some counties is used to program all the voting machines used in the county; the systems also tabulate final results aggregated from voting machines...

But election-management systems and voting machines are supposed to be air-gapped for security reasons—that is, disconnected from the internet and from any other systems that are connected to the internet. ES&S customers who had pcAnywhere installed also had modems on their election-management systems so ES&S technicians could dial into the systems and use the software to troubleshoot, thereby creating a potential port of entry for hackers as well...

Wyden told Motherboard that installing remote-access software and modems on election equipment “is the worst decision for security short of leaving ballot boxes on a Moscow street corner.”
In 2006, the same period when ES&S says it was still installing pcAnywhere on election systems, hackers stole the source code for the pcAnyhere software, though the public didn’t learn of this until years later in 2012 when a hacker posted some of the source code online, forcing Symantec, the distributor of pcAnywhere, to admit that it had been stolen years earlier...

He notes that election officials who purchased the systems likely were not aware of the potential risks they were taking in allowing this and didn’t understand the threat landscape to make intelligent decisions about installing such software.

All of this raises questions about how many counties across the US had remote-access software installed—in addition to ES&S customers—and whether intruders had ever leveraged it to subvert elections...

Wyden says he’s still waiting for ES&S to respond to the outstanding questions he sent the company in March. “ES&S needs to stop stonewalling and provide a full, honest accounting of equipment that could be vulnerable to remote attacks,” he told Motherboard. “When a corporation that makes half of America’s voting machines refuses to answer the most basic cyber security questions, you have to ask what it is hiding.”

Plan your experiment, gather your data, do your inference for all effects and interactions of interest. When all is said and done, accept some level of uncertainty in your conclusions: you might not be 97.5% sure that the treatment effect is positive, but that’s fine. For one thing, decisions need to be made. You were already going to make some decision with much less information—that is, with much more uncertainty. Now that you have more information, you can make a more informed decision. The other thing is, even if you did have a super-clean experiment with excellent measurements and a large and stable effect, so that you had that 95% interval excluding zero for your quantity of interest . . . so what? Whatever you care about is in the future, so even if your treatment was so great compared to the alternative in your sample being studied, there’s no saying what it will be in future populations under different conditions.

That’s not to say that you can’t learn from data; I’m not saying that at all. You can learn a lot from data. But forget about 95%. Just do your best, live your life, and be open about your uncertainties. You might get run over by a bus tomorrow anyway.

The post On this 4th of July, let’s declare independence from “95%” appeared first on Statistical Modeling, Causal Inference, and Social Science.

joint post with Jonathan Lorand, Blake Pollard, and Maru Sarazola

In the cell, chemical reactions are often ‘coupled’ so that reactions that release energy drive reactions that are biologically useful but involve an increase in energy. But how, exactly, does coupling work?

Much is known about this question, but the literature is also full of vague explanations and oversimplifications. Coupling cannot occur in equilibrium; it arises in open systems, where the concentrations of certain chemicals are held out of equilibrium due to flows in and out. One might thus suspect that the simplest mathematical treatment of this phenomenon would involve non-equilibrium steady states of open systems. However, Bazhin has shown that some crucial aspects of coupling arise in an even simpler framework:

• Nicolai Bazhin, The essence of ATP coupling, ISRN Biochemistry 2012 (2012), article 827604.

He considers ‘quasi-equilibrium’ states, where fast reactions have come into equilibrium and slow ones are neglected. He shows that coupling occurs already in this simple approximation.

In this series of blog articles we’ll do two things. First, we’ll review Bazhin’s work in a way that readers with no training in biology or chemistry should be able to follow. (But if you get stuck, ask questions!) Second, we’ll explain a fact that seems to have received insufficient attention: in many cases, coupling relies on emergent conservation laws.

Conservation laws are important throughout science. Besides those that are built into the fabric of physics, such as conservation of energy and momentum, there are also many ’emergent’ conservation laws that hold approximately in certain circumstances. Often these arise when processes that change a given quantity happen very slowly. For example, the most common isotope of uranium decays into lead with a half-life of about 4 billion years—but for the purposes of chemical experiments in the laboratory, it is useful to treat the amount of uranium as a conserved quantity.

The emergent conservation laws involved in biochemical coupling are of a different nature. Instead of making the processes that violate these laws happen more slowly, the cell uses enzymes to make other processes happen more quickly. At the time scales relevant to cellular metabolism, the fast processes dominate, while slowly changing quantities are effectively conserved. By a suitable choice of these emergent conserved quantities, the cell ensures that certain reactions that release energy can only occur when other ‘desired’ reactions occur. To be sure, this is only approximately true, on sufficiently short time scales. But this approximation is enlightening!

Following Bazhin, our main example involves ATP hydrolysis. We consider this following schema for a whole family of reactions:

Some concrete examples of this schema include:

• The synthesis of glutamine (XY) from glutamate (X) and ammonium (Y). This is part of the important glutamate-glutamine cycle in the central nervous system.

• The synthesis of sucrose (XY) from glucose (X) and fructose (Y). This is one of many processes whereby plants synthesize more complex sugars and starches from simpler building-blocks.

In these and other examples, the two reactions, taken together, have the effect of synthesizing a larger molecule XY out of two parts X and Y while ATP is broken down to ADP and the phosphate ion P_i Thus, they have the same net effect as this other pair of reactions:

The first reaction here is just the synthesis of XY from X and Y. The second is a deliberately simplified version of ATP hydrolysis. The first involves an increase of energy, while the second releases energy. But in the schema used in biology, these processes are ‘coupled’ so that ATP can only break down to ADP + P_i if X and Y combine to form XY.

As we shall see, this coupling crucially relies on a conserved quantity: the total number of Y molecules plus the total number of P_i ions is left unchanged by reactions (1) and (2). This fact is not a fundamental law of physics, nor even a general law of chemistry (such as conservation of phosphorus atoms). It is an emergent conservation law that holds approximately in special situations. Its approximate validity relies on the fact that the cell has enzymes that make reactions (1) and (2) occur more rapidly than reactions that violate this law, such as (3) and (4).

In the series to come, we’ll start by providing the tiny amount of chemistry and thermodynamics needed to understand what’s going on. Then we’ll raise the question “what is coupling?” Then we’ll study the reactions required for coupling ATP hydrolysis to the synthesis of XY from components X and Y, and explain why these reactions are not yet enough for coupling. Then we’ll show that coupling occurs in a ‘quasiequilibrium’ state where reactions (1) and (2), assumed much faster than the rest, have reached equilibrium, while the rest are neglected. And then we’ll explain the role of emergent conservation laws!

 

---

 
The paper:

• John Baez, Jonathan Lorand, Blake S. Pollard and Maru Sarazola,
Biochemical coupling through emergent conservation laws.

The blog series:

• Part 1 – Introduction.

• Part 2 – Review of reaction networks and equilibrium thermodynamics.

• Part 3 – What is coupling?

• Part 4 – Interactions.

• Part 5 – Coupling in quasiequilibrium states.

• Part 6 – Emergent conservation laws.

• Part 7 – The urea cycle.

• Part 8 – The citric acid cycle.

Pryor unhooks the deer’s skull from the wall above his still-curled-up companion. Examines it. Not a good specimen –the back half of the lower jaw’s missing, a gap that, with the open cranial cavity, makes room enough for Pryor’s head.

He puts it on. – Will Eaves, Murmur

So as we roll into the last dying embers of the flaming glory that is (North American) Pride Month (Europe you’ve still got some fun ahead), I’ve doing my best to make both good and questionable decisions.

The good decision was to read an absolutely fabulous book by a British author named Will Eaves (on a run of absolutely stunning books) who fairly recently released a book called Murmur.  Murmur is an odd beast, I guess you could say it’s a novel about Alan Turing but that would be moderately inaccurate and probably unfair.

Firstly, because even from one of my favourite authors, I am basically allergic to people writing about maths and mathematicians. It’s always so stodgy and wooden, as if they are writing about a world they don’t understand but also can’t convincingly fake. (Pride month analogy: I mostly avoid straight people writing queer stories for the same reason.) And Turing is a particular disaster for cultural portrayals: he intersects with too many complicated things (world war 2, cryptography, computers, pre-1967 British homosexuality) for him to ever be anything but an avatar.

So this is not a book about Alan Turing being a tortured genius. It’s a book about a guy called Alec Pryor who just happens to share a bunch of biographical details with Turing (Bletchley, Cambridge, ex-Fiance, arrest and chemical castration, Jungian therapist). And it’s not a book about him being sad, wronged, gay genius who kills himself. It’s a story about him living and him processes the changes in his internal life due to his punishment and his interactions with the outside world and his past and his musings on consciousness and computation.

All of which is to say Murmur is a complex, wonderfully written book that over its hundred and seventy something pages sketches out a fully realized world that doesn’t make me want to hide under the sofa in despair. And it does that rare thing for people telling this story: it doesn’t flatten out the story by focusing on the punchline but rather the person and the life behind it.

(As an aside, I’d strongly recommend Hannah Gadsby’s Netflix special Nanette, which talks about the damage we do by truncating our stories to make other people happy. It’s the only cultural document so far of 2018 worth going out of your way to see.)

I would recommend you find yourself a copy. It’s published by a small press, so order it from them or, if you’re in London, pop into Gays The Word and get yourself a copy.

And now that you’ve made your way through the unexpected book review, let’s get to the point

But  the point of writing this post wasn’t to do a short review of a wonderful book (it was to annoy Aki who is waiting for me to finish something). But Murmur is a book that spends some time (as you inevitably do when considering an ersatz Turing) considering the philosophical implications of artificial intelligence.  (Really selling it there Daniel.) And this parallels some discussion that I’ve been seeing around the traps about what we talk about when we talk about neural networks.

Also because the quote that I ripped from an absolutely wonderful run in the novel to unceremoniously shove at the top of this post made me think of how we use methods that we don’t fully understand.

The first paper I fell into (and incidentally reading papers on neural nets is my aforementioned questionable decision) has the direct title Polynomial Regression As an Alternative to Neural Nets, where Cheng, Khomtchouk, and Matloff argue that we might as well use polynomial regression as it’s easier to interpret than a NN and basically gives the same answer.

The main mathematical argument in the paper is that if you build a NN with a polynomial activation function, each layer gives a higher-order polynomial. They argue that the Stone-Weierstrass approximation theorem suggests that any activation function will lead to a NN that can be well approximated by a high-order polynomial.

Now as a general rule, anytime someone whips out Stone-Weierstrass I feel a little skeptical. Because the bit of me that remembers my approximation theory remembers that the construction in this theorem is very slow to converge. I’m also alarmed by the use of high-degree polynomial regression using the natural basis and no regularization. Both of these things are a very bad idea.

But the whole story–that neural networks can also be understood as being quite like polynomial regression and that analogy can allow us to improve our NN techniques–is the sort of story that we need to tell to understand how to make these methods better in a principled way.

(Honestly, I’m not a giant fan of a lot of the paper–it does polynomial regression in exactly the way people have been telling applied people they should never do it. But hey, there’s some interesting stuff in there.)

The other paper I read was a whole lot more interesting. Rather than trying to characterize a neural network as something else, it instead tries to argue that NNs should be used to process electronic health records. And it gets good results compared to standard methods, which is always a good sign. But that’s not what’s interesting.

The interesting thing comes via Twitter. Uri Shalit, who’s a prof at Technion, noticed something very interesting in the appendix. Table 1 in the appendix showed that regularized logistic regression performs almost exactly as well as the complicated deep net.

Once again, the Goliath of  AI is slain by the David of “really boring statistical methods”.

But again, there’s more to this than that. Firstly, the logistic regression required some light “feature engineering” (ie people who knew what they were doing had to do something to the covariates). In particular, they had to separate them into time bins to allow the model to do a good job at modelling time. The Deep Net didn’t need that.

This particular case of feature engineering is trivial, but in a lot of cases it’s careful understanding of the structure of the problem that lets us use simple statistical techniques instead of something weird and complex.  My favourite example of this is Bin Yu’s work where she (read: her and a lot of collaborators) basically reconstructed movies a person was watching from an fMRI scan! The way they did it was to understand how certain patterns excite certain pathway (basically using existing science) and putting those activation patterns in as covariates in a LASSO regression. So the simplest modern technique + feature engineering (aka science) gave fabulous results.

The argument for all of these complex AI and deep learning methods is that they allow us to be a little more sloppy with the science. And it seems to work quite well for images and movies, which are extremely structured. But electronic health records are not particularly structured and can have some really weird missingness problems, so it’s not clear that the same methods will have as much room to move. In this case a pretty boring regularized logistic regression does almost exactly as well, which suggests that the deep net is not able to really fly.

The path forward now is to start understanding these cases, working out how things work, when things work, and what techniques we can beg borrow and steal from other areas of stats and machine learning. Because deep learning is not a panacea, it’s just a boy standing in front of a girl asking her to love him.

The post In my role as professional singer and ham appeared first on Statistical Modeling, Causal Inference, and Social Science.

We provide a vivid demonstration of the mechanical effect of transverse spin momentum in an optical beam in free space. This component of the Poynting momentum was previously thought to be virtual, and unmeasurable. Here, its effect is revealed in the inertial motion of a probe particle in a circularly polarized Gaussian trap, in vacuum. Transverse spin forces combine with thermal fluctuations to induce a striking range of non-equilibrium phenomena. With increasing beam power we observe (i) growing departures from energy equipartition, (ii) the formation of coherent, thermally excited orbits and, ultimately, (iii) the ejection of the particle from the trap. Our results complement and corroborate recent measurements of spin momentum in evanescent waves, and extend them to a new geometry, in free space. In doing so, we exhibit fundamental, generic features of the mechanical interaction of circularly polarized light with matter. The work also shows how observations of the under-damped motion of probe particles can provide detailed information about the nature and morphology of momentum flows in arbitrarily structured light fields as well as providing a test bed for elementary non-equilibrium statistical mechanics.

Bazhin has analyzed ATP coupling in terms of quasiequilibrium states where fast reactions have reached an approximate steady state while slow reactions have not yet reached equilibrium. After an expository introduction to the relevant aspects of reaction network theory, we review his work and explain the role of emergent conserved quantities in coupling. These are quantities, left unchanged by fast reactions, whose conservation forces exergonic processes such as ATP hydrolysis to drive desired endergonic processes.