What is Deep Learning, how does it fit within the field of Machine Learning, what can you do with it, and how?
From smartphones' predictive apps to driverless cars, Artificial Intelligence (AI) is everywhere. It is also used by organizations to make sense of their data (big and small). For that, they apply Machine Learning techniques that find patterns in data, use them to make predictions, prescribe actions, and ultimately to automate decisions. PAPIs Connect is Europe's 1st AI conference for business and IT decision makers, managers and app developers. It brings together international experts who are exploiting AI’s opportunities in various domains. The conference also hosts the world's first AI Startup Battle where selection is made by an AI — powered by Telefonica Open Future.
Seldon is a company that has been operating a "black box" predictive API for three years and that recently open sourced its entire predictive stack. Alex Housley, Seldon's CEO, is one of the speakers at PAPIs '15 where he'll talk about the journey from closed to open: the challenges and pitfalls, architectural considerations, case studies, changes to business models, and new opportunities for partnership across the full stack — between both open and closed technology providers. In this post, Alex gives us a preview of his talk and a first look at the reasons behind this important change...
Big Data can only deliver an impact when associated with intelligence that extracts knowledge from it. This is what Machine Learning does by automatically finding patterns in data and using them to make predictions. These predictions can then be integrated in real-world applications and at large scale thanks to predictive APIs (Application Programming Interfaces). PAPIs '15 brings together experts from all over the world who are defining the future of predictive technology and who are exploiting its opportunities in various domains.
The first talks will be aimed at providing a gentle introduction to predictive applications. I’ll start with a keynote aimed at demystifying Machine Learning and showing its possibilities. Then, Keiran Thompson will use an example in real-estate to demonstrate the importance of data when building predictive models, while showing how simple it can be to create a predictive app. David Jones will show how new open source software allows to easily deploy predictive features on your website while focusing on your domain (and not on technical aspects). He will show how even small data holds value, with an example of how integrating product recommendations in a modest e-commerce website increased revenue.
After that, Athmane Hamel will introduce some problematics of big data problems, which we’ll also discuss in our conversation on stage with Claude Riwan of Orange. Claude will tell us about the use of predictive models for marketing and customer relationship management purposes at Orange, and we'll chat about what has changed for them in the last couple of years.
We’ll have another conversation on stage with Rand Hindi on consumer-facing mobile apps and their predictive future. Rand will tell us about the importance of open data and context-awareness, he will tell us about the apps his company has built for big French brands and he will give us a sneak peak at what they are currently building. This will be followed by showcases of tools and solutions that make it easier to create value from data with predictive technology: ChurnSpotter, NP6, BigML and Dataiku. You’ll be able to chat more with these companies at their booths during the coffee break that’s right after the showcases.
In the second part of the afternoon we'll have Lars Trieloff tell us about how predictive apps are even used for automating business decision making, and why it’s such a big deal. Then, Florian Douettau will discuss practical considerations in deploying, maintaining and improving predictive apps in production. He’ll share his vision on what’s to come by drawing a parallel with the deployment of web sites. We’ll end the afternoon with some lightning talks and a relaxed conversion with Yves Denneulin on ML in Education, before heading for drinks and canapés (sponsored by Blue Yonder).
Many thanks to all the speakers and companies I mentioned here! You'll find more information about our speakers on Lanyrd.
See you all very soon in Paris!
Big Data can only deliver an impact when associated with intelligence that extracts knowledge from it. This is what Machine Learning does by automatically finding patterns in data and using them to make predictions. PAPIs Connect brings together experts from all over the world who are exploiting predictive technology’s opportunities in various domains. It already ambitions to become the equivalent of the “Le Web” conference in the world of data...
In the PAPIs team, we share the vision that a predictive world will be a much better world. We are beginning to enter this new world where we can do things such as predicting demand to make better usage of our resources, where we can anticipate breakdowns and other issues (including medical ones) so we can take action to prevent them. In this world, information systems also give us access to the right content at the right time, businesses serve customers better by predicting their needs, and tedious tasks that require intelligence are automated. For some people, this can lead to terrible things. For others like us, it also opens up amazing possibilities to transform our world for the better.
Learning from data and making predictions, at large scale
Currently, our world is mostly one where we measure everything we can — situations and outcomes — by collecting data. The Internet of Things is one example of that. Sometimes, it isn’t so clear why we collect this data for. It turns out that machines can learn from data and "understand" how to relate situations to outcomes. They can generalise from their learnings to predict future outcomes when encountering new situations (e.g. maintenance issues, demand, actions…). As such, predictive technology powered by machine learning techniques is the main way that we can create value from (big) data. This has been referred to as “Big Data 2.0” in the Data Science for Business book by Foster Provost and Tom Fawcett.
Bret Victor, a designer behind the 1st iPad, wrote in his 2006 essay Magic Ink: “Until machine learning is as accessible and effortless as typing the word “learn,” it will never become widespread”. 5 years later, the McKinsey Big Data report pointed at a shortage of talent necessary for organizations to take advantage of big data in the coming years. Today, there’s a new wave of data analysis tools that provide an important part of the solution by removing many barriers to entry to machine learning. They make it easier to create predictive APIs that will be consumed by applications to make them starter. Some of these tools were presented last year at PAPIs ’14, the 1st International Conference on Predictive APIs and Apps, and I believe that they are key to exploiting the value of data at large scale, in all domains. But for this technology to be useful, it has to be connected to the right data sources, and predictions should be used in ways that solve real-world problems.
Announcing PAPIs Connect...
There were two things that stood out from the original description of PAPIs: 1. providing a "forum for the presentation of new machine learning APIs, techniques, architectures, and tools to build predictive applications" and 2. bringing together the makers of these APIs and tools, their users, and practicioners from industry, government and academia. APIs in general — and Predictive APIs in particular — impact everybody, including non technical people. PAPIs ’14 was proof of that with 25% of its audience being non technical. We think it’s important to address this part of our audience more, in order to connect predictive technology to business and to domains where it will be beneficial.
Today, we are announcing PAPIs Connect, a new series of events that are applications and business-driven, aimed at increasing awareness of predictive technology. We want to educate decision makers and enable them to instil a data-oriented, predictive culture into their organizations — whether they are startups, SMBs or large enterprises. For that, we intend to keep the practical mindset of PAPIs '14. The first PAPIs Connect will take place in Paris on 21 May 2015 (don’t confuse PAPIs and Paris!) and we just opened registration! We’re also planning a PAPIs Connect in the US towards the end of the year.
Coming soon to a city near you!
Of course, the annual PAPIs conferences will continue to exist; they will keep a technical bias by featuring more technical talks and they will be more focused on the presentation of new use cases, latest advancements and challenges in building predictive APIs and applications. The next edition, PAPIs ’15, will take place on 6-7 August 2015 in Sydney (right before KDD). By the way, our Call for Proposals is still open until 3 April...
PAPIs Connect will travel around the world and events will take place more frequently, which will increase the chances that we come near where you live! They will complement PAPIs and help achieve its objective: to expand adoption of machine learning and predictive technology by showing everyone how to use them and how to benefit from them.
We hope you will be a part of our journey.
General Chair of PAPIs.io
As you may already know, PAPIs will be coming to Sydney this year (and to other places as well, but more on that later). We’ll be collocated with KDD, the ACM conference on knowledge discovery and data mining which attracts 2000+ big data practitioners and researchers. While we haven’t confirmed the Sydney venue yet, we’ve decided on a date which we’re announcing today — so you can already put it in your calendar and start making plans...
As we're preparing for PAPIs.io '15, it's interesting to have a look at who came to the previous edition a few months ago in Barcelona. For this, we have two main sources of data: the registration data provided by Eventbrite and the responses to our post-conference survey. With it we were able to learn about the "profiles" of attendees, which we classified into 4 types...
The following is a guest post by the PAPIs.io '14 Hack Night winning team, Marian Moldovan (@marianmoldovan) and Enrique Otero (@meteotester) of the Innovation Department at BEEVA. The Hack Night took place from 6pm to 11pm on tutorials day and a prize was kindly offered by Import.io. In this post, Marian goes over the hack they did using tools they had learnt about earlier that day during tutorials, and how they were able to analyze images of buildings in Barcelona in an attempt to automatically detect different architectural styles.
The first day at PAPIs.io was our favorite. We assisted to many tutorials on interesting tools and we participated in the Hack Night: we wanted to dig deeper into some of the tools we'd seen during the day that impressed us most (Ŷhat, GraphLab and Indico.io). Gathering the things that impressed us most, our first idea was a computer vision system oriented towards recognizing buildings' architectural styles. Living in Barcelona, we found it particularly interesting to have a computer recognize the genius of Antoni Gaudi! Here's what we managed to do during the time we spent hacking...
As a starting point we got a dataset of 100 images using the Google Image Search API and the query “Barcelona building”. We used the Indico.io Image Features API to extract 2048-dimensional feature vectors from the images of Barcelona buildings that would enable us to compute similarities between them (the technology behind this is Deep Learning, and more specifically convolutional neural networks with imagenet pre-training). Then we used an R implementation of t-SNE algorithm in order to generate representations of images in 3D feature space and rendering the set of images with OpenGL using the rgl package. t-SNE is a machine learning algorithm for dimensionality reduction models which transforms high-dimensional objects into three-dimensional points in such a way that similar objects are modeled by nearby points and dissimilar objects are modeled by distant points. Proximity in 3D space would represent similarity between images, so we thought of using this visualization to find out clusters of buildings with similar style. In the animation below we rotate in this space in order to get an idea of where images are located.
Image proximity in feature space
If we look at points 77 and 84 in the representation above in 3D feature space, we can see they move together and they have almost same color, clearly distinct from their closest neighbors. It appears that the corresponding images are of the same building:
Moreover if we observe the group of images 38, 33, 90, 28, 67, 59, 19 (bottom-left quadrant, with the same color), we can verify they correspond to the same building too:
Gaudi’s “Casa Batlló” style has been clearly identified. Digging a little deeper into this we can see that a big part of images from Gaudi work lay out on the left-hand side of the visualization in feature space. Specifically the top-left corner gathers several pictures corresponding to “Casa Milá” (43, 70, 48) with some from “Casa Batlló (3) and “Sagrada Familia” (55,36,72,35) in the middle. While clearly separated, most pictures towards the left-hand side belong to modern building, mainly skyscrapers, like the Agbar tower, Media-Tic building, W-Barcelona hotel or Diagonal Zero-Zero tower.
Finding the Shortest path
Based on the image features generated by the previous algorithms, we were also able to figure out the "shortest path" from one building to another, made of transitions between similar buildings. Here's what we got with two buildings of very different styles: the Agbar tower (modern) and Gaudi’s “Casa Batlló”:
This makes sense, visually speaking! To compute this path we used an implementation provided by GraphLab — see our notebook in the GraphLab gallery (UPDATE: GraphLab changed their name to Dato). You'll find the rest of our code on Github.
With this hack we mainly wanted to play with computer vision and some of the technology presented at PAPIs.io. We plan to develop a mobile app, maybe with augmented reality in order to help tourists identify buildings. We could also help them create custom city tours, focused on an architectural style, or with transitions from one style to another. The 100 images we used here were downloaded from the Google Image Search API, which has restrictions, so we'll have a look at workarounds that allow us to get more images — maybe using Import.io?
We presented our work at the end of the Hack Night and we also gave a lightning talk on the second day of the conference (see slides below). To recharge batteries after the Hack Night we went for a drink and something to eat with people from BigML. Not only they told us interesting stuff but they kindly offered to pay the bill — thanks guys! Many thanks to Import.io too for the prize!
Import.io Magic is the ultimate in data extraction speed. It's the first data extraction tool to require absolutely no training on the part of the user, simply paste in a URL and their algorithms will do the rest. It represents a major step forward in import.io's greater mission of making web data accessible to everyone.
PAPIs.io '14 is over, but many other PAPIs will follow! It has been a real pleasure to see so many people participate and we'd like to thank you all for coming. We hope you enjoyed being part of the 1st International Conference on Predictive APIs and Apps! We'd very much like to learn more about you, to hear what you thought of the conference and to get suggestions on how we could improve it...