Marketplaces for algorithms – Computer Week

The development started even before one to two years in Analytics / Data science sector: startups, for example Algorithmia and RapidMiner but also from leading companies like Alteryx, FICO, IBM, Microsoft and Teradata develop still quite rudimentary marketplaces for software components , The primary goal is to provide customers with more flexibility and less monolithic compromises. But the broader implications are also considerably

marketplaces for algorithms
Photo violetkaipa – shutterstock.com

The App Economy reaches Big Data Science

The essential idea is very similar to the App Store or Play store of smartphone vendors: About marketplaces are all sorts of software and service components are distributed. In contrast to these mobile apps software apps or analytic apps offer only partial functionalities such as algorithms, models for predictive analytics, data sets and Datenwrapper. But this part of functionalities make the strength of the approach from. They are intended as building blocks that are to be linked to almost any data processing chains

However, should the marketplaces such as smartphones to be open and app providers are their building blocks ( such as algorithms) effortlessly “publish” can. By using the modules cash flows are triggered on the App providers and platform operators. So the interesting thing is the basic idea of ​​the App Economy to distribute software through an open yet powerful monetization and distribution infrastructure. In other words: We are talking about a Spotify, Uber or eBay for algorithms or software. The comparison limps a little, so the following is an example.

An illustration of the basic mechanics

Imagine an airport or shopping mall, where you want to automatically capture the mood of the passers-by. To illustrate how the algorithm marketplaces work and offer, in fact, novel solutions to complex business issues, we have a problem from the selected “Emotion Detection” (see Figure 1).

Figure. 1: How can algorithm marketplaces function using the example of “Emotion Detection” App
. Photo: Alex Linden / Gartner

In a first step (1), the two engineers Anton and Karl each its software components for emotion recognition and face extraction in the marketplace “published”. The important thing is: The code is not published in the strict sense, but merely asked about the runtime environment of the marketplace run available

In the second step, the software engineer who is responsible for the overall solution discovered these two. . components and integrating them – after registering the marketplace (3) – in its Code (4)

With each API call, controlled by an authentication code or token to pass essentially two things:

>

The corresponding algorithm is executed in the runtime environment of the marketplace and returned the result in the runtime environment of the user.

• There will be three cash flows:. two each to the developers and to the operator platform

Anton and Karl can therefore sit back comfortably and watch as the money arrives to your account.

algorithms like “a dime a dozen”

What does the principle of algorithm marketplaces so so interesting?

“3”

• The distribution costs of software are negligible: If this is so, why should not “publish” their implementations through algorithm-marketplaces just the engineers and scientists of the world? One can never know if the algorithm might not be in any industrial context of great benefit and potentially a lot of money could bring yes !? As with Spotify or via the setting will be fairly easy. Not least because Spotify now has over 30 million tracks

• decreases the granularity of commercialization. “Functions or features can not commercialize it This is only possible with complete. products. ” Or so was the decades-old doctrine which is now thrown through algorithm-markets into disarray.

In Figure 2, we illustrate the possible range of devices. One thing is clear: The term “algorithm” is much too narrow

Figure. 2: The spectrum of components comprising connectors, records, Datenwrapper, visualization, different classes of algorithms to partial solutions and “Human-as-apps”
Photo:.. Alex Linden / Gartner

• The result is a massive eco-system wiederbenutzbarer components: A very important aspect of the Algorithm marketplaces are the mechanisms of reusability of components. The function “allEmotions” could be published as a new composite algorithm in the market again. Any use of the “allEmotions” app would now cause four cash flows:. Three as usual to the developers of “recognizeEmotion” and “extractFaces” components as well as the platform operator and an additional to the developers of “allEmotions”

In practical terms, however, that algorithm marketplaces give software developers world wide incentives to develop reusable components. This development will give the standardization of IT in new push – something that can not be underestimated. More huge potential just for Big Data and Advanced Analytics algorithm marketplaces unfold in combination with PaaS (Platform-as-a-Service): every three Paas / cloud Mega providers (Amazon, Microsoft Azure and IBM Bluemix) invest massively <. /> p> Newsletter ‘News morning’ order!