Computed·Blg

Via Smithsonian

-----

Via MIT Technology Review

-----

A startup called Algorithmia has a new twist on online matchmaking. Its website is a place for businesses with piles of data to find researchers with a dreamboat algorithm that could extract insights–and profits–from it all.

The aim is to make better use of the many algorithms that are developed in academia but then languish after being published in research papers, says cofounder Diego Oppenheimer. Many have the potential to help companies sort through and make sense of the data they collect from customers or on the Web at large. If Algorithmia makes a fruitful match, a researcher is paid a fee for the algorithm’s use, and the matchmaker takes a small cut. The site is currently in a private beta test with users including academics, students, and some businesses, but Oppenheimer says it already has some paying customers and should open to more users in a public test by the end of the year.

“Algorithms solve a problem. So when you have a collection of algorithms, you essentially have a collection of problem-solving things,” says Oppenheimer, who previously worked on data-analysis features for the Excel team at Microsoft.

Oppenheimer and cofounder Kenny Daniel, a former graduate student at USC who studied artificial intelligence, began working on the site full time late last year. The company raised $2.4 million in seed funding earlier this month from Madrona Venture Group and others, including angel investor Oren Etzioni, the CEO of the Allen Institute for Artificial Intelligence and a computer science professor at the University of Washington.

Etzioni says that many good ideas are essentially wasted in papers presented at computer science conferences and in journals. “Most of them have an algorithm and software associated with them, and the problem is very few people will find them and almost nobody will use them,” he says.

One reason is that academic papers are written for other academics, so people from industry can’t easily discover their ideas, says Etzioni. Even if a company does find an idea it likes, it takes time and money to interpret the academic write-up and turn it into something testable.

To change this, Algorithmia requires algorithms submitted to its site to use a standardized application programming interface that makes them easier to use and compare. Oppenheimer says some of the algorithms currently looking for love could be used for machine learning, extracting meaning from text, and planning routes within things like maps and video games.

Early users of the site have found algorithms to do jobs such as extracting data from receipts so they can be automatically categorized. Over time the company expects around 10 percent of users to contribute their own algorithms. Developers can decide whether they want to offer their algorithms free or set a price.

All algorithms on Algorithmia’s platform are live, Oppenheimer says, so users can immediately use them, see results, and try out other algorithms at the same time.

The site lets users vote and comment on the utility of different algorithms and shows how many times each has been used. Algorithmia encourages developers to let others see the code behind their algorithms so they can spot errors or ways to improve on their efficiency.

One potential challenge is that it’s not always clear who owns the intellectual property for an algorithm developed by a professor or graduate student at a university. Oppenheimer says it varies from school to school, though he notes that several make theirs open source. Algorithmia itself takes no ownership stake in the algorithms posted on the site.

Eventually, Etzioni believes, Algorithmia can go further than just matching up buyers and sellers as its collection of algorithms grows. He envisions it leading to a new, faster way to compose software, in which developers join together many different algorithms from the selection on offer.

Via PCMag

-----

IBM on Thursday announced a new computer programming framework that draws inspiration from the way the human brain receives data, processes it, and instructs the body to act upon it while requiring relatively tiny amounts of energy to do so.

"Dramatically different from traditional software, IBM's new programming model breaks the mold of sequential operation underlying today's von Neumann architectures and computers. It is instead tailored for a new class of distributed, highly interconnected, asynchronous, parallel, large-scale cognitive computing architectures," IBM said in a statement introducing recent advances made by its Systems of Neuromorphic Adaptive Plastic Scalable Electronics (SyNAPSE) project.

IBM and research partners Cornell University and iniLabs have completed the second phase of the approximately $53 million project. With $12 million in new funding from the Defense Advanced Research Projects Agency (DARPA), IBM said work is set to commence on Phase 3, which will involve an ambitious plan to develop intelligent sensor networks built on a "brain-inspired chip architecture" using a "scalable, interconnected, configurable network of 'neurosynaptic cores'."

"Architectures and programs are closely intertwined and a new architecture necessitates a new programming paradigm," Dr. Dharmendra Modha, principal investigator and senior manager, IBM Research, said in a statement. "We are working to create a FORTRAN for synaptic computing chips. While complementing today's computers, this will bring forth a fundamentally new technological capability in terms of programming and applying emerging learning systems."

Going forward, work on the project will focus on honing a programming language for the SyNAPSE chip architecture first shown by IBM in 2011, with an agenda of using the new framework to deal with "big data" problems more efficiently.

IBM listed the following tools and systems it has developed with its partners towards this end:

• Simulator: A multi-threaded, massively parallel and highly scalable functional software simulator of a cognitive computing architecture comprising a network of neurosynaptic cores.
• Neuron Model: A simple, digital, highly parameterized spiking neuron model that forms a fundamental information processing unit of brain-like computation and supports a wide range of deterministic and stochastic neural computations, codes, and behaviors. A network of such neurons can sense, remember, and act upon a variety of spatio-temporal, multi-modal environmental stimuli.
• Programming Model: A high-level description of a "program" that is based on composable, reusable building blocks called "corelets." Each corelet represents a complete blueprint of a network of neurosynaptic cores that specifies a based-level function. Inner workings of a corelet are hidden so that only its external inputs and outputs are exposed to other programmers, who can concentrate on what the corelet does rather than how it does it. Corelets can be combined to produce new corelets that are larger, more complex, or have added functionality.
• Library: A cognitive system store containing designs and implementations of consistent, parameterized, large-scale algorithms and applications that link massively parallel, multi-modal, spatio-temporal sensors and actuators together in real-time. In less than a year, the IBM researchers have designed and stored over 150 corelets in the program library.
• Laboratory: A novel teaching curriculum that spans the architecture, neuron specification, chip simulator, programming language, application library and prototype design models. It also includes an end-to-end software environment that can be used to create corelets, access the library, experiment with a variety of programs on the simulator, connect the simulator inputs/outputs to sensors/actuators, build systems, and visualize/debug the results.

 

Via Culture Digitally

By Tarleton Gillespie

-----

I’m really excited to share my new essay, “The Relevance of Algorithms,” with those of you who are interested in such things. It’s been a treat to get to think through the issues surrounding algorithms and their place in public culture and knowledge, with some of the  participants in Culture Digitally (here’s the full litany: Braun, Gillespie, Striphas, Thomas, the third CD podcast, and Anderson‘s post just last week), as well as with panelists and attendees at the recent 4S and AoIR conferences, with colleagues at Microsoft Research, and with all of you who are gravitating towards these issues in their scholarship right now.

The motivation of the essay was two-fold: first, in my research on online platforms and their efforts to manage what they deem to be “bad content,” I’m finding an emerging array of algorithmic techniques being deployed: for either locating and removing sex, violence, and other offenses, or (more troublingly) for quietly choreographing some users away from questionable materials while keeping it available for others. Second, I’ve been helping to shepherd along this anthology, and wanted my contribution to be in the spirit of the its aims: to take one step back from my research to articulate an emerging issue of concern or theoretical insight that (I hope) will be of value to my colleagues in communication, sociology, science & technology studies, and information science.

The anthology will ideally be out in Fall 2013. And we’re still finalizing the subtitle. So here’s the best citation I have.

Gillespie, Tarleton. “The Relevance of Algorithms. forthcoming, in Media Technologies, ed. Tarleton Gillespie, Pablo Boczkowski, and Kirsten Foot. Cambridge, MA: MIT Press.

Below is the introduction, to give you a taste.

Algorithms play an increasingly important role in selecting what information is considered most relevant to us, a crucial feature of our participation in public life. Search engines help us navigate massive databases of information, or the entire web. Recommendation algorithms map our preferences against others, suggesting new or forgotten bits of culture for us to encounter. Algorithms manage our interactions on social networking sites, highlighting the news of one friend while excluding another’s. Algorithms designed to calculate what is “hot” or “trending” or “most discussed” skim the cream from the seemingly boundless chatter that’s on offer. Together, these algorithms not only help us find information, they provide a means to know what there is to know and how to know it, to participate in social and political discourse, and to familiarize ourselves with the publics in which we participate. They are now a key logic governing the flows of information on which we depend, with the “power to enable and assign meaningfulness, managing how information is perceived by users, the ‘distribution of the sensible.’” (Langlois 2012)

Algorithms need not be software: in the broadest sense, they are encoded procedures for transforming input data into a desired output, based on specified calculations. The procedures name both a problem and the steps by which it should be solved. Instructions for navigation may be considered an algorithm, or the mathematical formulas required to predict the movement of a celestial body across the sky. “Algorithms do things, and their syntax embodies a command structure to enable this to happen” (Goffey 2008, 17). We might think of computers, then, fundamentally as algorithm machines — designed to store and read data, apply mathematical procedures to it in a controlled fashion, and offer new information as the output.

But as we have embraced computational tools as our primary media of expression, and have made not just mathematics but all information digital, we are subjecting human discourse and knowledge to these procedural logics that undergird all computation. And there are specific implications when we use algorithms to select what is most relevant from a corpus of data composed of traces of our activities, preferences, and expressions.

These algorithms, which I’ll call public relevance algorithms, are — by the very same mathematical procedures — producing and certifying knowledge. The algorithmic assessment of information, then, represents a particular knowledge logic, one built on specific presumptions about what knowledge is and how one should identify its most relevant components.  That we are now turning to algorithms to identify what we need to know is as momentous as having relied on credentialed experts, the scientific method, common sense, or the word of God.

What we need is an interrogation of algorithms as a key feature of our information ecosystem (Anderson 2011), and of the cultural forms emerging in their shadows (Striphas 2010), with a close attention to where and in what ways the introduction of algorithms into human knowledge practices may have political ramifications. This essay is a conceptual map to do just that. I will highlight six dimensions of public relevance algorithms that have political valence:

1. Patterns of inclusion: the choices behind what makes it into an index in the first place, what is excluded, and how data is made algorithm ready

2. Cycles of anticipation: the implications of algorithm providers’ attempts to thoroughly know and predict their users, and how the conclusions they draw can matter

3. The evaluation of relevance: the criteria by which algorithms determine what is relevant, how those criteria are obscured from us, and how they enact political choices about appropriate and legitimate knowledge

4. The promise of algorithmic objectivity: the way the technical character of the algorithm is positioned as an assurance of impartiality, and how that claim is maintained in the face of controversy

5. Entanglement with practice: how users reshape their practices to suit the algorithms they depend on, and how they can turn algorithms into terrains for political contest, sometimes even to interrogate the politics of the algorithm itself

6. The production of calculated publics: how the algorithmic presentation of publics back to themselves shape a public’s sense of itself, and who is best positioned to benefit from that knowledge.

Considering how fast these technologies and the uses to which they are put are changing, this list must be taken as provisional, not exhaustive. But as I see it, these are the most important lines of inquiry into understanding algorithms as emerging tools of public knowledge and discourse.

It would also be seductively easy to get this wrong. In attempting to say something of substance about the way algorithms are shifting our public discourse, we must firmly resist putting the technology in the explanatory driver’s seat. While recent sociological study of the Internet has labored to undo the simplistic technological determinism that plagued earlier work, that determinism remains an alluring analytical stance. A sociological analysis must not conceive of algorithms as abstract, technical achievements, but must unpack the warm human and institutional choices that lie behind these cold mechanisms. I suspect that a more fruitful approach will turn as much to the sociology of knowledge as to the sociology of technology — to see how these tools are called into being by, enlisted as part of, and negotiated around collective efforts to know and be known. This might help reveal that the seemingly solid algorithm is in fact a fragile accomplishment.

~ ~ ~

Here is the full article [PDF]. Please feel free to share it, or point people to this post.

Via PhysOrg

-----

The Fourier transform is one of the most fundamental concepts in the information sciences. It’s a method for representing an irregular signal — such as the voltage fluctuations in the wire that connects an MP3 player to a loudspeaker — as a combination of pure frequencies. It’s universal in signal processing, but it can also be used to compress image and audio files, solve differential equations and price stock options, among other things.

The reason the Fourier transform is so prevalent is an algorithm called the fast Fourier transform (FFT), devised in the mid-1960s, which made it practical to calculate Fourier transforms on the fly. Ever since the FFT was proposed, however, people have wondered whether an even faster algorithm could be found.

At the Association for Computing Machinery’s Symposium on Discrete Algorithms (SODA) this week, a group of MIT researchers will present a new algorithm that, in a large range of practically important cases, improves on the fast Fourier transform. Under some circumstances, the improvement can be dramatic — a tenfold increase in speed. The new algorithm could be particularly useful for image compression, enabling, say, smartphones to wirelessly transmit large video files without draining their batteries or consuming their monthly bandwidth allotments.

Like the FFT, the new algorithm works on digital signals. A digital signal is just a series of numbers — discrete samples of an analog signal, such as the sound of a musical instrument. The FFT takes a digital signal containing a certain number of samples and expresses it as the weighted sum of an equivalent number of frequencies.

“Weighted” means that some of those frequencies count more toward the total than others. Indeed, many of the frequencies may have such low weights that they can be safely disregarded. That’s why the Fourier transform is useful for compression. An eight-by-eight block of pixels can be thought of as a 64-sample signal, and thus as the sum of 64 different frequencies. But as the researchers point out in their new paper, empirical studies show that on average, 57 of those frequencies can be discarded with minimal loss of image quality.

Heavyweight division

Signals whose Fourier transforms include a relatively small number of heavily weighted frequencies are called “sparse.” The new algorithm determines the weights of a signal’s most heavily weighted frequencies; the sparser the signal, the greater the speedup the algorithm provides. Indeed, if the signal is sparse enough, the algorithm can simply sample it randomly rather than reading it in its entirety.

“In nature, most of the normal signals are sparse,” says Dina Katabi, one of the developers of the new algorithm. Consider, for instance, a recording of a piece of chamber music: The composite signal consists of only a few instruments each playing only one note at a time. A recording, on the other hand, of all possible instruments each playing all possible notes at once wouldn’t be sparse — but neither would it be a signal that anyone cares about.

The new algorithm — which associate professor Katabi and professor Piotr Indyk, both of MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL), developed together with their students Eric Price and Haitham Hassanieh — relies on two key ideas. The first is to divide a signal into narrower slices of bandwidth, sized so that a slice will generally contain only one frequency with a heavy weight. 

In signal processing, the basic tool for isolating particular frequencies is a filter. But filters tend to have blurry boundaries: One range of frequencies will pass through the filter more or less intact; frequencies just outside that range will be somewhat attenuated; frequencies outside that range will be attenuated still more; and so on, until you reach the frequencies that are filtered out almost perfectly.

If it so happens that the one frequency with a heavy weight is at the edge of the filter, however, it could end up so attenuated that it can’t be identified. So the researchers’ first contribution was to find a computationally efficient way to combine filters so that they overlap, ensuring that no frequencies inside the target range will be unduly attenuated, but that the boundaries between slices of spectrum are still fairly sharp.

Zeroing in

Once they’ve isolated a slice of spectrum, however, the researchers still have to identify the most heavily weighted frequency in that slice. In the SODA paper, they do this by repeatedly cutting the slice of spectrum into smaller pieces and keeping only those in which most of the signal power is concentrated. But in an as-yet-unpublished paper, they describe a much more efficient technique, which borrows a signal-processing strategy from 4G cellular networks. Frequencies are generally represented as up-and-down squiggles, but they can also be though of as oscillations; by sampling the same slice of bandwidth at different times, the researchers can determine where the dominant frequency is in its oscillatory cycle.

Two University of Michigan researchers — Anna Gilbert, a professor of mathematics, and Martin Strauss, an associate professor of mathematics and of electrical engineering and computer science — had previously proposed an algorithm that improved on the FFT for very sparse signals. “Some of the previous work, including my own with Anna Gilbert and so on, would improve upon the fast Fourier transform algorithm, but only if the sparsity k” — the number of heavily weighted frequencies — “was considerably smaller than the input size n,” Strauss says. The MIT researchers’ algorithm, however, “greatly expands the number of circumstances where one can beat the traditional FFT,” Strauss says. “Even if that number k is starting to get close to n — to all of them being important — this algorithm still gives some improvement over FFT.”