Nosimpler

Note: this is a repost of a Facebook status I wrote off the cuff about a year ago, lightly edited. As such it has a different style from my other posts, but I still wanted to put it somewhere where it’d be easier to find and share than Facebook. 

Gradient descent, in its simplest where you just subtract the gradient of your loss function , is not dimensionally consistent: if the parameters you’re optimizing over have units of length, and the loss function is dimensionless, then the derivatives you’re subtracting have units of inverse length.

This observation can be used to reinvent the learning rate, which, for dimensional consistency, must have units of length squared. It also suggests that the learning rate ought to be set to something like for some kind of characteristic length scale , which loosely speaking is the length at which the curvature of starts to matter.

It might also make sense to give different parameters different units, which suggests furthermore that one might want a different learning rate for each parameter, or at least that one might want to partition the parameters into different subsets and choose different learning rates for each.

Going much further, from an abstract coordinate-free point of view the extra information you need to compute the gradient of a smooth function is a choice of (pseudo-)Riemannian metric on parameter space, which if you like is a gigantic hyperparameter you can try to optimize. Concretely this amounts to a version of preconditioned gradient descent where you allow yourself to multiply the gradient (in the coordinate-dependent sense) by a symmetric (invertible, ideally positive definite) matrix which is allowed to depend on the parameters. In the first paragraph this matrix was a constant scalar multiple of identity and in the third paragraph this matrix was constant diagonal.

This is an extremely general form of gradient descent, general enough to be equivariant under arbitrary smooth change of coordinates: that is, if you do this form of gradient descent and then apply a diffeomorphism to parameter space, you are still doing this form of gradient descent, with a different metric. For example, if you pick the preconditioning matrix to be the inverse Hessian (in the usual sense, assuming it’s invertible), you recover Newton’s method. This corresponds to choosing the metric at each point to be given by the Hessian (in the usual sense), which is the choice that makes the Hessian (in the coordinate-free sense) equal to the identity. This is a precise version of “the length at which the curvature of starts to matter” and in principle ameliorates the problem where gradient descent performs poorly in narrow valleys (regions where the Hessian (in the usual sense) is poorly conditioned), at least up to cubic and higher order effects.

In general it’s expensive to compute the inverse Hessian, so a more practical thing to do is to use a matrix which approximates it in some sense. And now we’re well on the way towards quasi-Newton methods. 

Echo-State Networks and Reservoir Computing have been studied for more than a decade. They provide a simpler yet powerful alternative to Recurrent Neural Networks, every internal weight is fixed and only the last linear layer is trained. They involve many multiplications by dense random matrices. Very large networks are difficult to obtain, as the complexity scales quadratically both in time and memory. Here, we present a novel optical implementation of Echo-State Networks using light-scattering media and a Digital Micromirror Device. As a proof of concept, binary networks have been successfully trained to predict the chaotic Mackey-Glass time series. This new method is fast, power efficient and easily scalable to very large networks.

We discuss a statistical model of evolution of wealth distribution, growing inequality and economic stratification in macro-economic scale. Evolution is driven by multiplicative stochastic fluctuations governed by the law of proportionate growth as well as by direct interactions between individuals and interactions between the private and public sectors. We concentrate on the effect of accumulated advantage and study the role of taxation and redistribution in the process of reducing wealth inequality. We show in particular that capital tax may significantly reduce wealth inequality and lead to a stabilisation of the system. The model predicts a decay of the population into economic classes of quantised wealth levels. The effect of economic stratification manifests in a characteristic multimodal form of the histogram of wealth distribution. We observe such a multimodal pattern in empirical data on wealth distribution for urban areas of China.

Author(s): Jing Chen, Song Cheng, Haidong Xie, Lei Wang, and Tao Xiang

The restricted Boltzmann machine is a fundamental building block of deep learning. The authors demonstrate its equivalence with tensor network states with explicit mappings, thus drawing a constructive connection between deep learning and quantum physics. On one side, deep learning approaches can be used to study novel states of matter. In return, investigations of tensor network states and their expressibility can be adapted to guide neural network architecture design.

[Phys. Rev. B 97, 085104] Published Fri Feb 02, 2018

We demonstrate that the Maximum Lyapunov Exponent for computable dynamical systems is isomorphic to the maximum capacity of a noiseless, memoryless channel in a Shannon communication model. The isomorphism allows the understanding of Lyapunov exponents in the simplified terms of Information Theory, rather than the traditional definitions in Chaos Theory. This work provides a bridge between fundamental physics and Information Theory to the mutual benefit of both fields. The result suggests, among other implications, that machine learning and other information theory methods can be successfully employed at the core of physics simulations.

The post Police Union Privileges appeared first on Marginal REVOLUTION.

We’re getting a lot of great posts here this week, but I also want to point out this, by one grad students:

• Christian William, Statebox: a universal language of distributed systems, Azimuth, January 22, 2018.

A brief teaser follows, in case you’re wondering what this is about.

The Azimuth Project’s research on networks and applied category theory has taken an interesting new turn. I always meant for it to do something useful, but I’m too theoretical to pull that off myself. Luckily there are plenty of other people with similar visions whose feet are a bit more firmly on the ground.

I first met the young Dutch hacktivist Jelle Herold at a meeting on network theory that I helped run in Torinio. I saw him again at a Simons Institute meeting on compositionality in computer science. He was already talking about his new startup.

Now it’s here. It’s called Statebox. Among other things, it’s an ambitious attempt to combine categories, open games, dependent types, Petri nets, string diagrams, and blockchains into a universal language for distributed systems.

Herold is inviting academics to help. I want to. But I couldn’t go to the Croatian island of Zlarin at the drop of a hat during classes. Luckily, my grad student Christian Williams is fascinated by the idea of using category theory and blockchain technology to do something good for the world: that’s why he came to work with me! So, I sent him to the first Statebox summit as my deputy. Now he has reported back. A snippet:

Zlarin is a lovely place, but we haven’t gotten to the best part — the people. All who attended are brilliant, creative, and spirited. Everyone’s eyes had a unique spark to light. I don’t think I’ve ever met such a fascinating group in my life. The crew: Jelle, Anton, Emi Gheorghe, Fabrizio Genovese, Daniel van Dijk, Neil Ghani, Viktor Winschel, Philipp Zahn, Pawel Sobocinski, Jules Hedges, Andrew Polonsky, Robin Piedeleu, Alex Norta, Anthony di Franco, Florian Glatz, Fredrik Nordvall Forsberg. These innovators have provocative and complementary ideas in category theory, computer science, open game theory, functional programming, and the blockchain industry; and they came to share an important goal. These are people who work earnestly to better humanity, motivated by progress, not profit. Talking with them gave me hope, that there are enough intelligent, open-minded, and caring people to fix this mess of modern society. In our short time together, we connected — now, almost all continue to contribute and grow the endeavor.

Ah, the starry-eyed idealism of youth! I’m feeling a bit beaten down by the events of the last year, so it’s nice (though somewhat unnerving) to see someone who is not. Read the whole article for more details about this endeavor.

guest post by Christian Williams

A short time ago, on the Croatian island of Zlarin, there gathered a band of bold individuals—rebels of academia and industry, whose everyday thoughts and actions challenge the separations of the modern world. They journeyed from all over to learn of the grand endeavor of another open mind, an expert functional programmer and creative hacktivist with significant mathematical knowledge: Jelle |yell-uh| Herold.

The Dutch computer scientist has devoted his life to helping our species and our planet: from consulting in business process optimization to winning a Greenpeace hackathon, from updating Netherlands telecommunications to creating a website to determine ways for individuals to help heal the earth, Jelle has gained a comprehensive perspective of the interconnected era. Through a diverse and innovative career, he has garnered crucial insights into software design and network computation—most profoundly, he has realized that it is imperative that these immense forces of global change develop thoughtful, comprehensive systematization.

Jelle understood that initiating such a grand ambition requires a massive amount of work, and the cooperation of many individuals, fluent in different fields of mathematics and computer science. Enter the Zlarin meeting: after a decade of consideration, Jelle has now brought together proponents of categories, open games, dependent types, Petri nets, string diagrams, and blockchains toward a singular end: a universal language of distributed systems—Statebox.

Statebox is a programming language formed and guided by fundamental concepts and principles of theoretical mathematics and computer science. The aim is to develop the canonical process language for distributed systems, and thereby elucidate the way these should actually be designed. The idea invokes the deep connections of these subjects in a novel and essential way, to make code simple, transparent, and concrete. Category theory is both the heart and pulse of this endeavor; more than a theory, it is a way of thinking universally. We hope the project helps to demonstrate the importance of this perspective, and encourages others to join.

The language is designed to be self-optimizing, open, adaptive, terminating, error-cognizant, composable, and most distinctively—visual. Petri nets are the natural representation of decentralized computation and concurrency. By utilizing them as program models, the entire language is diagrammatic, and this allows one to inspect the flow of the process executed by the program. While most languages only compile into illegible machine code, Statebox compiles directly into diagrams, so that the user immediately sees and understands the concrete realization of the abstract design. We believe that this immanent connection between the “geometric” and “algebraic” aspects of computation is of great importance.

Compositionality is a rightfully popular contemporary term, indicating the preservation of type under composition of systems or processes. This is essential to the universality of the type, and it is intrinsic to categories, which underpin the Petri net. A pertinent example is that composition allows for a form of abstraction in which programs do not require complete specification. This is parametricity: a program becomes executable when the functions are substituted with valid terms. Every term has a type, and one cannot connect pieces of code that have incompatible inputs and outputs—the compiler would simply produce an error. The intent is to preserve a simple mathematical structure that imposes as little as possible, and still ensure rationality of code. We can then more easily and reliably write tools providing automatic proofs of termination and type-correctness. Many more aspects will be explained as we go along, and in more detail in future posts.

Statebox is more than a specific implementation. It is an evolving aspiration, expressing an ideal, a source of inspiration, signifying a movement. We fully recognize that we are at the dawn of a new era, and do not assume that the current presentation is the best way to fulfill this ideal—but it is vital that this kind of endeavor gains the hearts and minds of these communities. By learning to develop and design by pure theory, we make a crucial step toward universal systems and knowledge. Formalisms are biased, fragile, transient—thought is eternal.

Thank you for reading, and thank you to John Baez—|bi-ez|, some there were not aware—for allowing me to write this post. Azimuth and its readers represent what scientific progress can and should be; it is an honor to speak to you. My name is Christian Williams, and I have just begun my doctoral studies with Dr. Baez. He received the invitation from Jelle and could not attend, and was generous enough to let me substitute. Disclaimer: I am just a young student with big dreams, with insufficient knowledge to do justice to this huge topic. If you can forgive some innocent confidence and enthusiasm, I would like to paint a big picture, to explain why this project is important. I hope to delve deeper into the subject in future posts, and in general to celebrate and encourage the cognitive revolution of Applied Category Theory. (Thank you also to Anton and Fabrizio for providing some of this writing when I was not well; I really appreciate it.)

Statebox Summit, Zlarin 2017, was awesome. Wish you could’ve been there. Just a short swim in the Adriatic from the old city of Šibenik |shib-enic|, there lies the small, green island of Zlarin |zlah-rin|, with just a few hundred kind inhabitants. Jelle’s friend, and part of the Statebox team, Anton Livaja and his family graciously allowed us to stay in their houses. Our headquarters was a hotel, one of the few places open in the fall. We set up in the back dining room for talks and work, and for food and sunlight we came to the patio and were brought platters of wonderful, wholesome Croatian dishes. As we burned the midnight oil, we enjoyed local beer, and already made history—the first Bitcoin transaction of the island, with a progressive bartender, Vinko.

Zlarin is a lovely place, but we haven’t gotten to the best part—the people. All who attended are brilliant, creative, and spirited. Everyone’s eyes had a unique spark to light. I don’t think I’ve ever met such a fascinating group in my life. The crew: Jelle, Anton, Emi Gheorghe, Fabrizio Genovese, Daniel van Dijk, Neil Ghani, Viktor Winschel, Philipp Zahn, Pawel Sobocinski, Jules Hedges, Andrew Polonsky, Robin Piedeleu, Alex Norta, Anthony di Franco, Florian Glatz, Fredrik Nordvall Forsberg. These innovators have provocative and complementary ideas in category theory, computer science, open game theory, functional programming, and the blockchain industry; and they came to share an important goal. These are people who work earnestly to better humanity, motivated by progress, not profit. Talking with them gave me hope, that there are enough intelligent, open-minded, and caring people to fix this mess of modern society. In our short time together, we connected—now, almost all continue to contribute and grow the endeavor.

Why is society a mess? The present human condition is absurd. We are in a cognitive renaissance, yet our world is in peril. We need to realize a deeper harmony of theory and practice—we need ideas that dare to dream big, that draw on the vast wealth of contemporary thought to guide and unite subjects in one mission. The way of the world is only a reflection of how we choose to think, and for more than a century we have delved endlessly into thought itself. If we truly learn from our thought, knowledge and application become imminently interrelated, not increasingly separate. It is imperative that we abandon preconception, pretense and prejudice, and ask with naive sincerity: “How should things be, really, and how can we make it happen?”

This pertains more generally to the irresponsibly ad hoc nature of society—we find ourselves entrenched in inadequate systems. Food, energy, medicine, finance, communications, media, governance, technology—our deepening dependence on centralization is our greatest vulnerability. Programming practice is the perfect example of the gradual failure of systems when their design is left to wander in abstraction. As business requirements evolved, technological solutions were created haphazardly, the priority being immediate return over comprehensive methodology, which resulted in ‘duct-taped’ systems, such as the Windows OS. Our entire world now depends on unsystematic software, giving rise to so much costly disorganization, miscommunication, and worse, bureaucracy. Statebox aims to close the gap between the misguided formalisms which came out of this type of degeneration, and design a language which corresponds naturally to essential mathematical concepts—to create systems which are rational, principled, universal. To explain why Statebox represents to us such an important ideal, we must first consider its closest relative, the elephant in the technological room: blockchain.

Often the best ideas are remarkably simple—in 2008, an unknown person under the alias Satoshi Nakamoto published the whitepaper Bitcoin: A Peer-to-Peer Electronic Cash System. In just a few pages, a protocol was proposed which underpins a new kind of computational network, called a blockchain, in which interactions are immediate, transparent, and permanent. This is a personal interpretation—the paper focuses on the application given in its title. In the original financial context, immediacy is one’s ability to directly transact with anyone, without intermediaries, such as banks; transparency is one’s right to complete knowledge of the economy in which one participates, meaning that each node owns a copy of the full history of the network; permanence is the irrevocability of one’s transactions. These core aspects are made possible by an elegant use of cryptography and game theory, which essentially removes the need for trusted third parties in the authorization, verification, and documentation of transactions. Per word, it’s almost peerless in modern influence; the short and sweet read is recommended.

The point of this simplistic explanation is that blockchain is about more than economics. The transaction could be any cooperation, the value could be any social good—when seen as a source of consensus, the blockchain protocol can be expanded to assimilate any data and code. After several years of competing cryptocurrencies, the importance of this deeper idea was gradually realized. There arose specialized tools to serve essential purposes in some broader system, and only recently have people dared to conceive of what this latter could be. In 2014, a wunderkind named Vitalik Buterin created Ethereum, a fully programmable blockchain. Solidity is a Turing-complete language of smart contracts, autonomous programs which enable interactions and enact rules on the network. With this framework, one can not only transact with others, but implement any kind of process; one can build currencies, websites, or organizations—decentralized applications, constructed with smart contracts, could be just about anything.

There is understandably great confidence and excitement for these ventures, and many are receiving massive public investment. Seriously, the numbers are staggering—but most of it is pure hype. There is talk of the first global computer, the internet of value, a single shared source of truth, and other speculative descriptions. But compared to the ambition, the actual theory is woefully underdeveloped. So far, implementations make almost no use of the powerful ideas of mathematics. There are still basic flaws in blockchain itself, the foundation of almost all decentralized technology. For example, the two viable candidates for transaction verification are called Proof of Work and Proof of Stake: the former requires unsustainable consumption of resources, namely hardware and electricity, and the latter is susceptible to centralization. Scalability is a major problem, thus also cost and speed of transactions. A major Ethereum dApp, Decentralized Autonomous Organization, was hacked.

These statements are absolutely not to disregard all of the great work of this community; it is primarily rhetoric to distinguish the high ideals of Statebox, and I lack the eloquence to make the point diplomatically, nor near the knowledge to give a real account of this huge endeavor. We now return to the rhetoric.

What seems to be lost in the commotion is the simple recognition that we do not yet really know what we should make, nor how to do so. The whole idea is simply too big—the space of possibility is almost completely unknown, because this innovation can open every aspect of society to reform. But as usual, people try to ignore their ignorance, imagining it will disappear, and millions clamor about things we do not yet understand. Most involved are seeing decentralization as an exciting business venture, rather than our best hope to change the way of this broken world; they want to cash in on another technological wave. Of the relatively few idealists, most still retain the assumptions and limitations of the blockchain.

For all this talk, there is little discussion of how to even work toward the ideal abstract design. Most mathematics associated to blockchain is statistical analysis of consensus, while we’re sitting on a mountain of powerful categorical knowledge of systems. At the summit, Prof. Neil Ghani said “it’s like we’re on the Moon, talking about going to Mars, while everyone back on Earth still doesn’t even have fire.” We have more than enough conceptual technology to begin developing an ideal and comprehensive system, if the right minds come together. Theory guides practice, practice motivates theory—the potential is immense.

Fortunately, there are those who have this big picture in mind. Long before the blockchain craze, Jelle saw the fundamental importance of both distributed systems and the need for academic-industrial symbiosis. In the mid-2000’s, he used Petri nets to create process tools for businesses. Employees could design and implement any kind of abstract workflow to more effectively communicate and produce. Jelle would provide consultation to optimize these processes, and integrate them into their existing infrastructure—as it executed, it would generate tasks, emails, forms and send them to designated individuals to be completed for the next iteration. Many institutions would have to shell out millions of dollars to IBM or Fujitsu for this kind of software, and his was more flexible and intuitive. This left a strong impression on Jelle, regarding the power of Petri nets and the impact of deliberate design.

Many experiences like this gradually instilled in Jelle a conviction to expand his knowledge and begin planning bold changes to the world of programming. He attended mathematics conferences, and would discuss with theorists from many relevant subjects. On the island, he told me that it was actually one of Baez’s talks about networks which finally inspired him to go for this huge idea. By sincerely and openly reaching out to the whole community, Jelle made many valuable connections. He invited these thinkers to share his vision—theorists from all over Europe, and some from overseas, gathered in Croatia to learn and begin to develop this project—and it was a great success.

By now you may be thinking, alright kid spill the beans already. Here they are, right into your brain—well, most will be in the next post, but we should at least have a quick overview of some of the main ideas not already discussed.

The notion of open system complements compositionality. The great difference between closure and openness, in society as well as theory, was a central theme in many of our conversations during the summit. Although we try to isolate and suspend life and cognition in abstraction, the real, concrete truth is what flows through these ethereal forms. Every system in Statebox is implicitly open, and this impels design to idealize the inner and outer connections of processes. Open systems are central to the Baez Network Theory research team. There are several ways to categorically formalize open systems; the best are still being developed, but the first main example can be found in The Algebra of Open and Interconnected Systems by Brendan Fong, an early member of the team.

Monoidal categories, as this blog knows well, represent systems with both series and parallel processes. One of the great challenge of this new era of interconnection is distributed computation—getting computers to work together as a supercomputer, and monoidal categories are essential to this. Here, objects are data types, and morphisms are computations, while composition is serial and tensor is parallel. As Dr. Baez has demonstrated with years of great original research, monoidal categories are essential to understanding the complexity of the world. If we can connect our knowledge of natural systems to social systems, we can learn to integrate valuable principles—a key example being complete resource cognizance.

Petri nets are presentations of free strict symmetric monoidal categories, and as such they are ideal models of “normal” computation, i.e. associative, unital, and commutative. Open Petri nets are the workhorses of Statebox. They are the morphisms of a category which is itself monoidal—and via openness it is even richer and more versatile. Most importantly it is compact closed, which introduces a simple but crucial duality into computation—input-output interchange—which is impossible in conventional cartesian closed computation, and actually brings the paradigm closer to quantum computation

Petri nets represent processes in an intuitive, consistent, and decentralized way. These will be multi-layered via the notion of operad and a resourceful use of Petri net tokens, representing the interacting levels of a system. Compositionality makes exploring their state space much easier: the state space of a big process can be constructed from those of smaller ones, a technique that more often than not avoids state space explosion, a long-standing problem in Petri net analysis. The correspondence between open Petri nets and a logical calculus, called place/transition calculus, allows the user to perform queries on the Petri net, and a revolutionary technique called information-gain computing greatly reduces response time.

Dependently typed functional programming is the exoskeleton of this beautiful beast; in particular, the underlying language is Idris. Dependent types arose out of both theoretical mathematics and computer science, and they are beginning to be recognized as very general, powerful, and natural in practice. Functional programming is a similarly pure and elegant paradigm for “open” computation. They are fascinating and inherently categorical, and deserve whole blog posts in the future.

Even economics has opened its mind to categories. Statebox is very fortunate to have several of these pioneers—open game theory is a categorical, compositional version of game theory, which allows the user to dynamically analyze and optimize code. Jules’ choice of the term “teleological category” is prescient; it is about more than just efficiency—it introduces the possibility of building principles into systems, by creating game-theoretical incentives which can guide people to cooperate for the greater good, and gradually lessen the influence of irrational, selfish priorities.

Categories are the language by which Petri nets, functional programming, and open games can communicate—and amazingly, all of these theories are unified in an elegant representation called string diagrams. These allow the user to forget the formalism, and reason purely in graphical terms. All the complex mathematics goes under the hood, and the user only needs to work with nodes and strings, which are guaranteed to be formally correct.

Category theory also models the data structures that are used by Statebox: Typedefs is a very lightweight—but also very expressive—data structure, that is at the very core of Statebox. It is based on initial F-algebras, and can be easily interpreted in a plethora of pre-existing solutions, enabling seamless integration with existing systems. One of the core features of Typedefs is that serialization is categorically internalized in the data structure, meaning that every operation involving types can receive a unique hash and be recorded on the blockchain public ledger. This is one of the many components that make Statebox fail-resistant: every process and event is accounted for on the public ledger, and the whole history of a process can be rolled back and analyzed thanks to the blockchain technology.

The Statebox team is currently working on a monograph that will neatly present how all the pertinent categorical theories work together in Statebox. This is a formidable task that will take months to complete, but will also be the cleanest way to understand how Statebox works, and which mathematical questions have still to be answered to obtain a working product. It will be a thorough document that also considers important aspects such as our guiding ethics.

The team members are devoted to creating something positive and different, explicitly and solely to better the world. The business paradigm is based on the principle that innovation should be open and collaborative, rather than competitive and exclusive. We want to share ideas and work with you. There are many blooming endeavors which share the ideals that have been described in this article, and we want them all to learn from each other and build off one another.

For example, Statebox contributor and visionary economist Viktor Winschel has a fantastic project called Oicos. The great proponent of applied category theory, David Spivak, has an exciting and impressive organization called Categorical Informatics. Mike Stay, a past student of Dr. Baez, has started a company called Pyrofex, which is developing categorical distributed computation. There are also somewhat related languages for blockchain, such as Simplicity, and innovative distributed systems such as Iota and RChain. Even Ethereum is beginning to utilize categories, with Casper. And of course there are research groups, such as Network Theory and Mathematically Structured Programming, as well as so many important papers, such as Algebraic Databases. This is just a slice of everything going on; as far as I know there is not yet a comprehensive account of all the great applied category theory and distributed innovations being developed. Inevitably these endeavors will follow the principle they share, and come together in a big way. Statebox is ready, willing, and able to help make this reality.

If you are interested in Statebox, you are welcomed with open arms. You can contact Jelle at jelle@statebox.io, Fabrizio at fabrizio@statebox.org, Emi at emi@statebox.io, Anton at anton@statebox.io; they can provide more information, connect you to the discussion, or anything else. There will be a second summit in 2018 in about six months, details to be determined. We hope to see you there. Future posts will keep you updated, and explain more of the theory and design of Statebox. Thank you very much for reading.

P.S. Found unexpected support in Šibenik! Great bar—once a reservoir.

Investigating wave propagation in fluid enables a variety of important applications in underwater communications, object detections and unmanned robot control. Conventionally, momentum and spin reveal fundamental physical properties about propagating waves. Yet, vast previous research focused on the orbital angular momentum of acoustics without thinking about the existence possibility of spin due to the longitudinal wave nature. Here, we show that underwater acoustic wave processes the non-trivial spin angular momentum intrinsically, which is associated with its special spin-orbital coupling relation for longitudinal waves. Furthermore, we demonstrate that this intrinsic spin, although unobservable in plane wave form, can be detected by four approaches: wave interference, Gaussian exponential decay form, boundary evanescent wave, and waveguide mode. We further show that the strong spin-orbital coupling can be exploited to achieve unidirectional excitation and backscattering immune transport. We hope the present results can improve the geometric and topological understanding about underwater acoustic wave and pave the way on the spin-related underwater applications.

... your priorities change.  The graph shows viewership at Pornhub at the time of the (false) nuclear
bomb warning.  Via Boing Boing.

Abstract

This article is very interesting:

• Ed Yong, Brain cells share information with virus-like capsules, Atlantic, January 12, 2018.

Your brain needs a protein called Arc. If you have trouble making this protein, you’ll have trouble forming new memories. The neuroscientist Jason Shepherd noticed something weird:

He saw that these Arc proteins assemble into hollow, spherical shells that look uncannily like viruses. “When we looked at them, we thought: What are these things?” says Shepherd. They reminded him of textbook pictures of HIV, and when he showed the images to HIV experts, they confirmed his suspicions. That, to put it bluntly, was a huge surprise. “Here was a brain gene that makes something that looks like a virus,” Shepherd says.

That’s not a coincidence. The team showed that Arc descends from an ancient group of genes called gypsy retrotransposons, which exist in the genomes of various animals, but can behave like their own independent entities. They can make new copies of themselves, and paste those duplicates elsewhere in their host genomes. At some point, some of these genes gained the ability to enclose themselves in a shell of proteins and leave their host cells entirely. That was the origin of retroviruses—the virus family that includes HIV.

It’s worth pointing out that gypsy is the name of a specific kind of retrotransposon. A retrotransposon is a gene that can make copies of itself by first transcribing itself from DNA into RNA and then converting itself back into DNA and inserting itself at other places in your chromosomes.

About 40% of your genes are retrotransposons! They seem to mainly be ‘selfish genes’, focused on their own self-reproduction. But some are also useful to you.

So, Arc genes are the evolutionary cousins of these viruses, which explains why they produce shells that look so similar. Specifically, Arc is closely related to a viral gene called gag, which retroviruses like HIV use to build the protein shells that enclose their genetic material. Other scientists had noticed this similarity before. In 2006, one team searched for human genes that look like gag, and they included Arc in their list of candidates. They never followed up on that hint, and “as neuroscientists, we never looked at the genomic papers so we didn’t find it until much later,” says Shepherd.

I love this because it confirms my feeling that viruses are deeply entangled with our evolutionary past. Computer viruses are just the latest phase of this story.

As if that wasn’t weird enough, other animals seem to have independently evolved their own versions of Arc. Fruit flies have Arc genes, and Shepherd’s colleague Cedric Feschotte showed that these descend from the same group of gypsy retrotransposons that gave rise to ours. But flies and back-boned animals co-opted these genes independently, in two separate events that took place millions of years apart. And yet, both events gave rise to similar genes that do similar things: Another team showed that the fly versions of Arc also sends RNA between neurons in virus-like capsules. “It’s exciting to think that such a process can occur twice,” says Atma Ivancevic from the University of Adelaide.

This is part of a broader trend: Scientists have in recent years discovered several ways that animals have used the properties of virus-related genes to their evolutionary advantage. Gag moves genetic information between cells, so it’s perfect as the basis of a communication system. Viruses use another gene called env to merge with host cells and avoid the immune system. Those same properties are vital for the placenta—a mammalian organ that unites the tissues of mothers and babies. And sure enough, a gene called syncytin, which is essential for the creation of placentas, actually descends from env. Much of our biology turns out to be viral in nature.

Here’s something I wrote in 1998 when I was first getting interested in this business:

RNA reverse transcribing viruses

RNA reverse transcribing viruses are usually called retroviruses. They have a single-stranded RNA genome. They infect animals, and when they get inside the cell’s nucleus, they copy themselves into the DNA of the host cell using reverse transcriptase. In the process they often cause tumors, presumably by damaging the host’s DNA.

Retroviruses are important in genetic engineering because they raised for the first time the possibility that RNA could be transcribed into DNA, rather than the reverse. In fact, some of them are currently being deliberately used by scientists to add new genes to mammalian cells.

Retroviruses are also important because AIDS is caused by a retrovirus: the human immunodeficiency virus (HIV). This is part of why AIDS is so difficult to treat. Most usual ways of killing viruses have no effect on retroviruses when they are latent in the DNA of the host cell.

From an evolutionary viewpoint, retroviruses are fascinating because they blur the very distinction between host and parasite. Their genome often contains genetic information derived from the host DNA. And once they are integrated into the DNA of the host cell, they may take a long time to reemerge. In fact, so-called endogenous retroviruses can be passed down from generation to generation, indistinguishable from any other cellular gene, and evolving along with their hosts, perhaps even from species to species! It has been estimated that up to 1% of the human genome consists of endogenous retroviruses! Furthermore, not every endogenous retrovirus causes a noticeable disease. Some may even help their hosts.

It gets even spookier when we notice that once an endogenous retrovirus lost the genes that code for its protein coat, it would become indistinguishable from a long terminal repeat (LTR) retrotransposon—one of the many kinds of “junk DNA” cluttering up our chromosomes. Just how much of us is made of retroviruses? It’s hard to be sure.

For my whole article, go here:

• Subcellular life forms.

It’s about the mysterious subcellular entities that stand near the blurry border between the living and the non-living—like viruses, viroids, plasmids, satellites, transposons and prions. I need to update it, since a lot of new stuff is being discovered!

Jason Shepherd’s new paper has a few other authors:

• Elissa D. Pastuzyn, Cameron E. Day, Rachel B. Kearns, Madeleine Kyrke-Smith, Andrew V. Taibi, John McCormick, Nathan Yoder, David M. Belnap, Simon Erlendsson, Dustin R. Morado, John A.G. Briggs, Cédric Feschotte and Jason D. Shepherd, The neuronal gene Arc encodes a repurposed retrotransposon gag protein that mediates intercellular RNA transfer, Cell 172 (2018), 275–288.

We develop a framework for the general interpretation of the stochastic dynamical system near a limit cycle. Such quasi-periodic dynamics are commonly found in a variety of nonequilibrium systems, including the spontaneous oscillations of hair cells in the inner ear. We demonstrate quite generally that in the presence of noise, the phase of the limit cycle oscillator will diffuse while deviations in the directions locally orthogonal to that limit cycle will display the Lorentzian power spectrum of a damped oscillator. We identify two mechanisms by which these stochastic dynamics can acquire a complex frequency dependence, and discuss the deformation of the mean limit cycle as a function of temperature. The theoretical ideas are applied to data obtained from spontaneously oscillating hair cells of the amphibian sacculus.

It has been extensively shown in past literature that Bayesian Game Theory and Quantum Non-locality have strong ties between them. Pure Entangled States have been used, in both common and conflict interest games, to gain advantageous payoffs, both at the individual and social level. In this paper we construct a game for a Mixed Entangled State such that this state gives higher payoffs than classically possible, both at the individual level and the social level. Also, we use the I-3322 inequality so that states that aren't helpful as advice for Bell-CHSH inequality can also be used. Finally, the measurement setting we use is a Restricted Social Welfare Strategy (given this particular state).

Related Articles

Representation Learning of Logic Words by an RNN: From Word Sequences to Robot Actions.

Front Neurorobot. 2017;11:70

Authors: Yamada T, Murata S, Arie H, Ogata T

Abstract
An important characteristic of human language is compositionality. We can efficiently express a wide variety of real-world situations, events, and behaviors by compositionally constructing the meaning of a complex expression from a finite number of elements. Previous studies have analyzed how machine-learning models, particularly neural networks, can learn from experience to represent compositional relationships between language and robot actions with the aim of understanding the symbol grounding structure and achieving intelligent communicative agents. Such studies have mainly dealt with the words (nouns, adjectives, and verbs) that directly refer to real-world matters. In addition to these words, the current study deals with logic words, such as "not," "and," and "or" simultaneously. These words are not directly referring to the real world, but are logical operators that contribute to the construction of meaning in sentences. In human-robot communication, these words may be used often. The current study builds a recurrent neural network model with long short-term memory units and trains it to learn to translate sentences including logic words into robot actions. We investigate what kind of compositional representations, which mediate sentences and robot actions, emerge as the network's internal states via the learning process. Analysis after learning shows that referential words are merged with visual information and the robot's own current state, and the logical words are represented by the model in accordance with their functions as logical operators. Words such as "true," "false," and "not" work as non-linear transformations to encode orthogonal phrases into the same area in a memory cell state space. The word "and," which required a robot to lift up both its hands, worked as if it was a universal quantifier. The word "or," which required action generation that looked apparently random, was represented as an unstable space of the network's dynamical system.

PMID: 29311891 [PubMed]

Cocaine abuse disrupts dopamine system function, and reduces cocaine inhibition of the dopamine transporter (DAT), which results in tolerance. Although tolerance is a hallmark of cocaine addiction and a DSM-V criterion for substance abuse disorders, the molecular adaptations producing tolerance are unknown, and testing the impact of DAT changes on drug taking behaviors has proven difficult. In regard to treatment, amphetamine has shown efficacy in reducing cocaine intake; however, the mechanisms underlying these effects have not been explored. The goals of this study were twofold; we sought to (1) identify the molecular mechanisms by which cocaine exposure produces tolerance and (2) determine whether amphetamine-induced reductions in cocaine intake are connected to these mechanisms. Using cocaine self-administration and fast-scan cyclic voltammetry in male rats, we show that low-dose, continuous amphetamine treatment, during self-administration or abstinence, completely reversed cocaine tolerance. Amphetamine treatment also reversed escalated cocaine intake and decreased motivation to obtain cocaine as measured in a behavioral economics task, thereby linking tolerance to multiple facets of cocaine use. Finally, using fluorescence resonance energy transfer imaging, we found that cocaine tolerance is associated with the formation of DAT-DAT complexes, and that amphetamine disperses these complexes. In addition to extending our basic understanding of DATs and their role in cocaine reinforcement, we serendipitously identified a novel therapeutic target: DAT oligomer complexes. We show that dispersion of oligomers is concomitant with reduced cocaine intake, and propose that pharmacotherapeutics aimed at these complexes may have potential for cocaine addiction treatment.

SIGNIFICANCE STATEMENT Tolerance to cocaine's subjective effects is a cardinal symptom of cocaine addiction and a DSM-V criterion for substance abuse disorders. However, elucidating the molecular adaptions that produce tolerance and determining its behavioral impact have proven difficult. Using cocaine self-administration in rats, we link tolerance to cocaine effects at the dopamine transporter (DAT) with aberrant cocaine-taking behaviors. Further, tolerance was associated with multi-DAT complexes, which formed after cocaine exposure. Treatment with amphetamine deconstructed DAT complexes, reversed tolerance, and decreased cocaine seeking. These data describe the behavioral consequence of cocaine tolerance, provide a putative mechanism for its development, and suggest that compounds that disperse DAT complexes may be efficacious treatments for cocaine addiction.

Johannes Kepler loved geometry, so of course he was fascinated by Platonic solids. His early work Mysterium Cosmographicum, written in 1596, includes pictures showing how the 5 Platonic solids correspond to the 5 elements:

Five elements? Yes, besides earth, air, water and fire, he includes a fifth element that doesn’t feel the Earth’s gravitational pull: the ‘quintessence’, or ‘aether’, from which heavenly bodies are made.

In the same book he also tried to use the Platonic solids to explain the orbits of the planets:

The six planets are Mercury, Venus, Earth, Mars, Jupiter and Saturn. And the tetrahedron and cube, in case you’re wondering, sit outside the largest sphere shown above. You can see them another picture from Kepler’s book:

These ideas may seem goofy now, but studying the exact radii of the planets’ orbits led him to discover that these orbits aren’t circular: they’re ellipses! By 1619 this led him to what we call Kepler’s laws of planetary motion. And those, in turn, helped Newton verify Hooke’s hunch that the force of gravity goes as the inverse square of the distance between bodies!

In honor of this, the problem of a particle orbiting in an inverse square force law is called the Kepler problem.

So, I’m happy that Greg Egan, Layra Idarani and I have come across a solid mathematical connection between the Platonic solids and the Kepler problem.

But this involves a detour into the 4th dimension!

It’s a remarkable fact that the Kepler problem has not just the expected conserved quantities—energy and the 3 components of angular momentum—but also 3 more: the components of the Runge–Lenz vector. To understand those extra conserved quantities, go here:

• Greg Egan, The ellipse and the atom.

Noether proved that conserved quantities come from symmetries. Energy comes from time translation symmetry. Angular momentum comes from rotation symmetry. Since the group of rotations in 3 dimensions, called SO(3), is itself 3-dimensional, it gives 3 conserved quantities, which are the 3 components of angular momentum.

None of this is really surprising. But if we take the angular momentum together with the Runge–Lenz vector, we get 6 conserved quantities—and these turn out to come from the group of rotations in 4 dimensions, SO(4), which is itself 6-dimensional. The obvious symmetries in this group just rotate a planet’s elliptical orbit, while the unobvious ones can also squash or stretch it, changing the eccentricity of the orbit.

(To be precise, all this is true only for the ‘bound states’ of the Kepler problem: the circular and elliptical orbits, not the parabolic or hyperbolic ones, which work in a somewhat different way. I’ll only be talking about bound states in this post!)

Why should the Kepler problem have symmetries coming from rotations in 4 dimensions? This is a fascinating puzzle—we know a lot about it, but I doubt the last word has been spoken. For an overview, go here:

• John Baez, Mysteries of the gravitational 2-body problem.

This SO(4) symmetry applies not only to the classical mechanics of the inverse square force law, but also the quantum mechanics! Nobody cares much about the quantum mechanics of two particles attracting gravitationally via an inverse square force law—but people care a lot about the quantum mechanics of hydrogen atoms, where the electron and proton attract each other via their electric field, which also obeys an inverse square force law.

So, let’s talk about hydrogen. And to keep things simple, let’s pretend the proton stays fixed while the electron orbits it. This is a pretty good approximation, and experts will know how to do things exactly right. It requires only a slight correction.

It turns out that wavefunctions for bound states of hydrogen can be reinterpreted as functions on the 3-sphere, S³ The sneaky SO(4) symmetry then becomes obvious: it just rotates this sphere! And the Hamiltonian of the hydrogen atom is closely connected to the Laplacian on the 3-sphere. The Laplacian has eigenspaces of dimensions n² where n = 1,2,3,…, and these correspond to the eigenspaces of the hydrogen atom Hamiltonian. The number n is called the principal quantum number, and the hydrogen atom’s energy is proportional to -1/n².

If you don’t know all this jargon, don’t worry! All you need to know is this: if we find an eigenfunction of the Laplacian on the 3-sphere, it will give a state where the hydrogen atom has a definite energy. And if this eigenfunction is invariant under some subgroup of SO(4), so will this state of the hydrogen atom!

The biggest finite subgroup of SO(4) is the rotational symmetry group of the 600-cell, a wonderful 4-dimensional shape with 120 vertices and 600 dodecahedral faces. The rotational symmetry group of this shape has a whopping 7,200 elements! And here is a marvelous moving image, made by Greg Egan, of an eigenfunction of the Laplacian on S³ that’s invariant under this 7,200-element group:

We’re seeing the wavefunction on a moving slice of the 3-sphere, which is a 2-sphere. This wavefunction is actually real-valued. Blue regions are where this function is positive, yellow regions where it’s negative—or maybe the other way around—and black is where it’s almost zero. When the image fades to black, our moving slice is passing through a 2-sphere where the wavefunction is almost zero.

For a full explanation, go here:

• Greg Egan, In the chambers with seven thousand symmetries, 2 January 2018.

Layra Idarani has come up with a complete classification of all eigenfunctions of the Laplacian on S³ that are invariant under this group… or more generally, eigenfunctions of the Laplacian on a sphere of any dimension that are invariant under the even part of any Coxeter group. For the details, go here:

• Layra Idarani, SG-invariant polynomials, 4 January 2018.

All that is a continuation of a story whose beginning is summarized here:

• John Baez, Quantum mechanics and the dodecahedron.

So, there’s a lot of serious math under the hood. But right now I just want to marvel at the fact that we’ve found a wavefunction for the hydrogen atom that not only has a well-defined energy, but is also invariant under this 7,200-element group. This group includes the usual 60 rotational symmetries of a dodecahedron, but also other much less obvious symmetries.

I don’t have a good picture of what these less obvious symmetries do to the wavefunction of a hydrogen atom. I understand them a bit better classically—where, as I said, they squash or stretch an elliptical orbit, changing its eccentricity while not changing its energy.

We can have fun with this using the old quantum theory—the approach to quantum mechanics that Bohr developed with his colleague Sommerfeld from 1920 to 1925, before Schrödinger introduced wavefunctions.

In the old Bohr–Sommerfeld approach to the hydrogen atom, the quantum states with specified energy, total angular momentum and angular momentum about a fixed axis were drawn as elliptical orbits. In this approach, the symmetries that squash or stretch elliptical orbits are a bit easier to visualize:

This picture by Pieter Kuiper shows some orbits at the 5th energy level, n = 5: namely, those with different eigenvalues of the total angular momentum, ℓ.

While the old quantum theory was superseded by the approach using wavefunctions, it’s possible to make it mathematically rigorous for the hydrogen atom. So, we can draw elliptical orbits that rigorously correspond to a basis of wavefunctions for the hydrogen atom. So, I believe we can draw the orbits corresponding to the basis elements whose linear combination gives the wavefunction shown as a function on the 3-sphere in Greg’s picture above!

We should get a bunch of ellipses forming a complicated picture with dodecahedral symmetry. This would make Kepler happy.

As a first step in this direction, Greg drew the collection of orbits that results when we take a circle and apply all the symmetries of the 600-cell:

For more details, read this:

• Greg Egan, Kepler orbits with the symmetries of the 600-cell.

Postscript

To do this really right, one should learn a bit about ‘old quantum theory’. I believe people have been getting it a bit wrong for quite a while—starting with Bohr and Sommerfeld!

If you look at the ℓ = 0 orbit in the picture above, it’s a long skinny ellipse. But I believe it really should be a line segment straight through the proton: that’s what’s an orbit with no angular momentum looks like.

There’s a paper about this:

• Manfred Bucher, Rise and fall of the old quantum theory.

Matt McIrvin had some comments on this:

This paper from 2008 is a kind of thing I really like: an exploration of an old, incomplete theory that takes it further than anyone actually did at the time.

It has to do with the Bohr-Sommerfeld “old quantum theory”, in which electrons followed definite orbits in the atom, but these were quantized–not all orbits were permitted. Bohr managed to derive the hydrogen spectrum by assuming circular orbits, then Sommerfeld did much more by extending the theory to elliptical orbits with various shapes and orientations. But there were some problems that proved maddeningly intractable with this analysis, and it eventually led to the abandonment of the “orbit paradigm” in favor of Heisenberg’s matrix mechanics and Schrödinger’s wave mechanics, what we know as modern quantum theory.

The paper argues that the old quantum theory was abandoned prematurely. Many of the problems Bohr and Sommerfeld had came not from the orbit paradigm per se, but from a much simpler bug in the theory: namely, their rejection of orbits in which the electron moves entirely radially and goes right through the nucleus! Sommerfeld called these orbits “unphysical”, but they actually correspond to the s orbital states in the full quantum theory, with zero angular momentum. And, of course, in the full theory the electron in these states does have some probability of being inside the nucleus.

So Sommerfeld’s orbital angular momenta were always off by one unit. The hydrogen spectrum came out right anyway because of the happy accident of the energy degeneracy of certain orbits in the Coulomb potential.

I guess the states they really should have been rejecting as “unphysical” were Bohr’s circular orbits: no radial motion would correspond to a certain zero radial momentum in the full theory, and we can’t have that for a confined electron because of the uncertainty principle.

Synchronization in networks of coupled oscillators is known to be largely determined by the spectral and symmetry properties of the interaction network. Here we leverage this relation to study a class of networks for which the threshold coupling strength for global synchronization is the lowest among all networks with the same number of nodes and links. These networks, defined as being uniform, complete, and multi-partite (UCM), appear at each of an infinite sequence of network-complement transitions in a larger class of networks characterized by having near-optimal thresholds for global synchronization. We show that the distinct symmetry structure of the UCM networks, which by design are optimized for global synchronizability, often leads to formation of clusters of synchronous oscillators, and that such states can coexist with the state of global synchronization.

The post The future of crypto-assets? (from the comments) appeared first on Marginal REVOLUTION.

The post Were U.S. nuclear tests more harmful than we had thought? appeared first on Marginal REVOLUTION.

It is demonstrated is this letter that linear multistep methods for integrating ordinary differential equations can be used to develop a family of fast forward scattering algorithms with higher orders of convergence. Excluding the cost of computing the discrete eigenvalues, the nonlinear Fourier transform (NFT) algorithm thus obtained has a complexity of $O{KN+C_pN\log^2N}$ such that the error vanishes as $O{N^{-p}}$ where $p\in\{1,2,3,4\}$ and $K$ is the number of eigenvalues. Such an algorithm can be potentially useful for the recently proposed NFT based modulation methodology for optical fiber communication. The exposition considers the particular case of the backward differentiation formula ($C_p=p^3$) and the implicit Adams method ($C_p=(p-1)^3$) of which the latter proves to be the most accurate family of methods for fast NFT.

The post Switzerland is Prepared for Civilizational Collapse appeared first on Marginal REVOLUTION.

Chemistries exhibiting complex dynamics—from inorganic oscillators to gene regulatory networks—have been long known but either cannot be reprogrammed at will or rely on the sophisticated enzyme chemistry underlying the central dogma. Can simpler molecular mechanisms, designed from scratch, exhibit the same range of behaviors? Abstract chemical reaction networks have been proposed as a programming language for complex dynamics, along with their systematic implementation using short synthetic DNA molecules. We developed this technology for dynamical systems by identifying critical design principles and codifying them into a compiler automating the design process. Using this approach, we built an oscillator containing only DNA components, establishing that Watson-Crick base-pairing interactions alone suffice for complex chemical dynamics and that autonomous molecular systems can be designed via molecular programming languages.

Axo-myelinic neurotransmission: a novel mode of cell signalling in the central nervous system

Axo-myelinic neurotransmission: a novel mode of cell signalling in the central nervous system, Published online: 14 December 2017; doi:10.1038/nrn.2017.166

Axo-myelinic neurotransmission: a novel mode of cell signalling in the central nervous system

Author(s): C. Marletto and V. Vedral

Two proposals describe how to test whether gravity is inherently quantum by measuring the entanglement between two masses.

[Phys. Rev. Lett. 119, 240402] Published Wed Dec 13, 2017

Humans use rules to organize their actions to achieve specific goals. Although simple rules that link a sensory stimulus to one response may suffice in some situations, often, the application of multiple, hierarchically organized rules is required. Recent theories suggest that progressively higher level rules are encoded along an anterior-to-posterior gradient within PFC. Although some evidence supports the existence of such a functional gradient, other studies argue for a lesser degree of specialization within PFC. We used fMRI to investigate whether rules at different hierarchical levels are represented at distinct locations in the brain or encoded by a single system. Thirty-seven male and female participants represented and applied hierarchical rule sets containing one lower-level stimulus–response rule and one higher-level selection rule. We used multivariate pattern analysis to investigate directly the representation of rules at each hierarchical level in absence of information about rules from other levels or other task-related information, thus providing a clear identification of low- and high-level rule representations. We could decode low- and high-level rules from local patterns of brain activity within a wide frontoparietal network. However, no significant difference existed between regions encoding representations of rules from both levels except for precentral gyrus, which represented only low-level rule information. Our findings show that the brain represents conditional rules regardless of their level in the explored hierarchy, so the human control system did not organize task representation according to this dimension. Our paradigm represents a promising approach to identifying critical principles that shape this control system.

SIGNIFICANCE STATEMENT Several recent studies investigating the organization of the human control system propose that rules at different control levels are organized along an anterior-to-posterior gradient within PFC. In this study, we used multivariate pattern analysis to explore independently the representation of formally identical conditional rules belonging to different levels of a cognitive hierarchy and provide for the first time a clear identification of low- and high-level rule representations. We found no major spatial differences between regions encoding rules from different hierarchical levels. This suggests that the human brain does not use levels in the investigated hierarchy as a topographical organization principle to represent rules controlling our behavior. Our paradigm represents a promising approach to identifying which principles are critical.

Not the "national debt," but the debts of individual Americans, as depicted on this map (my embed is a screencap - the interactive version here lets you zoom in on your state and your county for data).

2016 data derived from a random sample of deidentified, consumer-level records from a major credit bureau, as well as estimates from summary tables of the US Census Bureau’s American Community Survey (2015 or 2011–15)... Debt in collections includes past-due credit lines that have been closed and charged-off on their books as well as unpaid bills reported to the credit bureaus that the creditor is attempting to collect.

Links to various commentaries at Digg.  My attention was drawn to Minnesota...

A previous analysis by the Urban Institute focused on medical debt, and found one reason it was so concentrated in the South was because the uninsured rates tended to be higher. While that changed to some extent with the Affordable Care Act, many Southern states chose not to expand Medicaid. On the other hand, Minnesota — which has the lowest rates of debt — has one of the most generous Medicaid programs in the country, and a more inclusive and higher-quality health care system.

...only 3% of Minnesotans have medical debt in collections, and only one county (rural Clearwater County) has medical debt rates over 11%.

Compare that picture to the state of medical debt in the rest of the country. Nationwide, 18% of people have medical debt in collections, and, as CityLab noted, much of that debt is concentrated in states that chose not to expand Medicaid under the ACA.

Most Washington politicians are tone-deaf to the financial crises experienced by so many Americans.  They are busy waging their internecine battles, meeting with lobbyists, and pandering to their donors.

Jellyfish have been referred to as the "cockroaches of the sea," with reference to both species' ability to survive under the harshest conditions.  An article in the newest edition of The Atlantic reviews a new book about jellyfish:

Their delicacy notwithstanding, in recent decades jellyfish species have come to be thought of as the durable and opportunistic inheritors of our imperiled seas. Jellyfish blooms—the intermittent, and now widely reported, flourishing of vast swarms—are held by many to augur the depletion of marine biomes; they are seen as a signal that the oceans have been overwarmed, overfished, acidified, and befouled... The vision—hat tipped to science fiction—is of the planet’s oceans transformed into something like an aspic terrine. In waters thickened by the gummy mucus of living and dead jellyfish, other sea life will be smothered. Because jellyfish recall the capsules of single-celled protozoa, this eventuality invites portrayal as a devolution of the marine world—a reversion to the “primordial soup.”..

Perhaps the most complex issue Berwald takes on is jellyfish blackouts. Sweden, Scotland, the Philippines, Tokyo, California, and Israel have all suffered intermittent electrical outages caused by jellyfish sucked into the intake pipes and cooling systems of coal-fired and nuclear power stations... In Spineless, Berwald travels to Spain’s Murcia region and takes us to the Mar Menor lagoon, which had become so jellified in 2002 that “you couldn’t drive a boat through the water.” Here barrel and fried-egg jellyfish are pernicious—so much so that they’re removed from the sea by the bargeload and dumped into ditches near the airport.

More at that link. Then today I found a report of jellyfish menacing Chinese aircraft carriers:

In 2006, the aircraft carrier USS Ronald Reagan was incapacitated while visiting Brisbane, Australia due to blubber jellyfish swarms. Reportedly, cooling pipes for the ship’s nuclear reactor were clogged with the foot-wide jellies, necessitating an evacuation of the carrier...

Ironically, the jellyfish problem is partially of China’s own doing. As many as one hundred million sharks are killed each year, much of it in the form of bycatch in an attempt to catch other forms of seafood but also for shark’s fin soup, a delicacy in China. Although demand for the soup has declined in recent years, the shark population is still way down. Sharks are a major predator of jellyfish and scientists believe their absence is a major reason why jellyfish populations have exploded.

Related: Your children will eat jellyfish for dinner.

Photo credit: GettyLucia /Terui

Similarity search—for example, identifying similar images in a database or similar documents on the web—is a fundamental computing problem faced by large-scale information retrieval systems. We discovered that the fruit fly olfactory circuit solves this problem with a variant of a computer science algorithm (called locality-sensitive hashing). The fly circuit assigns similar neural activity patterns to similar odors, so that behaviors learned from one odor can be applied when a similar odor is experienced. The fly algorithm, however, uses three computational strategies that depart from traditional approaches. These strategies can be translated to improve the performance of computational similarity searches. This perspective helps illuminate the logic supporting an important sensory function and provides a conceptually new algorithm for solving a fundamental computational problem.