09 March 2018

Introducing the iNaturalist 2018 Challenge




Thanks to recent advances in deep learning, the visual recognition abilities of machines have improved dramatically, permitting the practical application of computer vision to tasks ranging from pedestrian detection for self-driving cars to expression recognition in virtual reality. One area that remains challenging for computers, however, is fine-grained and instance-level recognition. Earlier this month, we posted an instance-level landmark recognition challenge for identifying individual landmarks. Here we focus on fine-grained visual recognition, which is to distinguish species of animals and plants, car and motorcycle models, architectural styles, etc. For computers, discriminating fine-grained categories is challenging because many categories have relatively few training examples (i.e., the long tail problem), the examples that do exist often lack authoritative training labels, and there is variability in illumination, viewing angle and object occlusion.

To help confront these hurdles, we are excited to announce the 2018 iNaturalist Challenge (iNat-2018), a species classification competition offered in partnership with iNaturalist and Visipedia (short for Visual Encyclopedia), a project for which Caltech and Cornell Tech received a Google Focused Research Award. This is a flagship challenge for the 5th International Workshop on Fine Grained Visual Categorization (FGVC5) at CVPR 2018. Building upon the first iNaturalist challenge, iNat-2017, iNat-2018 spans over 8000 categories of plants, animals, and fungi, with a total of more than 450,000 training images. We invite participants to enter the competition on Kaggle, with final submissions due in early June. Training data, annotations, and links to pretrained models can be found on our GitHub repo.

iNaturalist has emerged as a world leader for citizen scientists to share observations of species and connect with nature since its founding in 2008. It hosts research-grade photos and annotations submitted by a thriving, engaged community of users. Consider the following photo from iNaturalist:
The map on the right shows where the photo was taken. Image credit: Serge Belongie.
You may notice that the photo on the left contains a turtle. But did you also know this is a Trachemys scripta, common name “Pond Slider?” If you knew the latter, you possess knowledge of fine-grained or subordinate categories.

In contrast to other image classification datasets such as ImageNet, the dataset in the iNaturalist challenge exhibits a long-tailed distribution, with many species having relatively few images. It is important to enable machine learning models to handle categories in the long-tail, as the natural world is heavily imbalanced – some species are more abundant and easier to photograph than others. The iNaturalist challenge will encourage progress because the training distribution of iNat-2018 has an even longer tail than iNat-2017.
Distribution of training images per species for iNat-2017 and iNat-2018, plotted on a log-log scale, illustrating the long-tail behavior typical of fine-grained classification problems. Image Credit: Grant Van Horn and Oisin Mac Aodha.
Along with iNat-2018, FGVC5 will also host the iMaterialist 2018 challenge (including a furniture categorization challenge and a fashion attributes challenge for product images) and a set of “FGVCx” challenges representing smaller scale – but still significant – challenges, featuring content such as food and modern art.

FGVC5 will be showcased on the main stage at CVPR 2018, thereby ensuring broad exposure for the top performing teams. This project will advance the state-of-the-art in automatic image classification for real world, fine-grained categories, with heavy class imbalances, and large numbers of classes. We cordially invite you to participate in these competitions and help move the field forward!

Acknowledgements
We’d like to thank our colleagues and friends at iNaturalist, Visipedia, and FGVC5 for working together to advance this important area. At Google we would like to thank Hartwig Adam, Weijun Wang, Nathan Frey, Andrew Howard, Alessandro Fin, Yuning Chai, Xiao Zhang, Jack Sim, Yuan Li, Grant Van Horn, Yin Cui, Chen Sun, Yanan Qian, Grace Vesom, Tanya Birch, Wendy Kan, and Maggie Demkin.

A Guide to FilmStruck, the Netflix for Classic Movies


flimstruck-classic-movies

Back in 2017, we introduced you to Primephonic, the Spotify for classical music. And it turns out that there’s a movie equivalent of Primephonic. It’s called FilmStruck, and this streaming service specializes in classic movies. With Netflix focusing more on original content, FilmStruck may just be a viable alternative to Netflix for those of you who enjoy old cinema. In this article, we’re going to explain everything you need to know about FilmStruck. What films are available? How much does it cost? What are its pros and cons? And what platforms is it available on? Let’s find out. What Is...

Read the full article: A Guide to FilmStruck, the Netflix for Classic Movies


Read Full Article

Keyo modernizes housing with rent auto-pay that boosts your credit


 There’s nothing we spend more money on for a worse experience than housing. Keyo wants to fix all of it. The audacious startup envisions a world where a building’s other tenants get $20 to show you an available apartment on your schedule. Where you auto-pay your rent online and it improves your credit score. Where you get local business perks and can communicate maintenance… Read More

Read Full Article

I Quit Ubuntu Linux and I’m Happier Than Ever


quit-ubuntu

My Linux journey began in the Ubuntu ecosystem. I installed Xubuntu on an old computer without an internet connection and played around with it. A year or two later, after suffering from a Windows crash that took all of my data with it, I transitioned entirely to Ubuntu 8.10. Back then I distro hopped quite a bit, but Ubuntu was my anchor. Over the next few years, that changed. The more I used Linux, the more I valued the free and open source ethos along with the culture of collaboration. I became increasingly put off by the direction Canonical was...

Read the full article: I Quit Ubuntu Linux and I’m Happier Than Ever


Read Full Article

Facebook has signed a deal to stream 25 afternoon MLB games


 Filed under moves that are potentially groundbreaking with big implications but with quite small numbers, Facebook has signed an exclusive deal with the MLB to stream 25 afternoon games, according to Bloomberg. The deal is a bit reminiscent of the one Twitter signed for Thursday Night Football back in 2016 to stream games that are outside of the range of primetime football (usually reserved… Read More
Read Full Article

Facebook and Warner Music ink recorded and published music deal for videos and messages


 Just weeks after signing a large licensing deal with leading indy label representative ICE, Facebook has continued its march into the world of music with the news that it has now added Warner Music Group into the mix, the last major label that was not yet working with the social network. Facebook has signed a wide-ranging licensing deal that covers all of Warner Music’s recorded and… Read More
Read Full Article

Mario can now guide your route in Google Maps


 It’s Mario Time! Beloved gaming character Mario is coming to Google Maps, thanks to a partnership between Google and Nintendo ahead of Mario Day, March 10. (MAR10…get it, Mario?) In addition to the various deals and sales on Nintendo games, the mustachioed plumber will also make an appearance in Google Maps’ navigation, if you opt in to have Mario accompany you on your… Read More
Read Full Article

Facebook has signed a deal to stream 25 afternoon MLB games


 Filed under moves that are potentially groundbreaking with big implications but with quite small numbers, Facebook has signed an exclusive deal with the MLB to stream 25 afternoon games, according to Bloomberg. The deal is a bit reminiscent of the one Twitter signed for Thursday Night Football back in 2016 to stream games that are outside of the range of primetime football (usually reserved… Read More

Read Full Article

Fortnite Battle Royale Is Coming to iOS and Android


Fortnite Battle Royale is an absolute phenomenon. It is, as the name suggests, a battle royale game in the vein of PlayerUnknown’s Battlegrounds and H1Z1. However, Fortnite also boasts Minecraft-style elements and Overwatch-style visuals. After proving to be a huge success on Windows, Mac, PS4, and Xbox One, Epic Games has announced that Fortnite Battle Royale is coming to a smartphone or tablet near you. Starting with iOS, and with support for Android to follow in the next few months. Play Fortnite Battle Royale on Your Smartphone In its blog post announcing the mobile version of Fortnite Battle Royale, Epic...

Read the full article: Fortnite Battle Royale Is Coming to iOS and Android


Read Full Article

How to Manage Your Chrome Extensions Without Any Bother


manage-chrome-extensions

Extensions make Chrome a great tool for any task. But if you don’t know how to manage them, you can end up with a cluttered, slow browser. And no one wants that. So let’s take a look at a few ways to manage your Chrome extensions. We’ll go through the built-in extension manager, then look at a powerful extension that can replace it. After that, we’ll move on to a few tips and tricks that will help you get the most out of your extensions. How to Manage Your Chrome Extensions You can see a lot of your extensions in...

Read the full article: How to Manage Your Chrome Extensions Without Any Bother


Read Full Article

Porsche’s electrification to lead to most powerful 911 to date, says CEO


 The storied 911 will not be immune from Porsche’s electrification plans and that could be a good thing. The car company’s CEO recently told Autocar that the hybrid 911 “will be the most powerful 911 we’ve ever had; 700 bhp might be possible.” Count me in (as long as someone buys it for me). Porsche already makes 911 models above the 700 mark so it’s likely… Read More

Read Full Article

YC-backed Playbook wants students to make plans online, hang out offline


 Social planning apps have yet to take the world by storm. A handful of these apps tried and failed years ago, and even Swarm decided to ditch that methodology and focus on location logging. But a new Y Combinator-backed company seems to be picking up traction. And shockingly enough, the winning ingredient seems to be location! Location! Location! The startup is called Playbook, and its main… Read More
Read Full Article

Can iPhones Get Viruses? Here Are the Facts


iphone-get-viruses

It’s an age-old idea that Windows and Android devices are insecure messes just waiting to pick up a virus, while Macs and iPhones are immune to such threats. And while Android can indeed pick up malware, smart habits will protect the majority of users. On the Apple side of the fence, you have to act quite foolishly to infect your Mac. But what about iOS? Can your iPhone really get a virus? Let’s look at the facts. Viruses and Malware Defined Before we discuss viruses on iOS—the operating system that powers iPhones, iPads, and iPod Touchs—it’s important to note what...

Read the full article: Can iPhones Get Viruses? Here Are the Facts


Read Full Article

10 Ways to Supercharge Your Chrome Browser on Android


supercharge-chrome-android

Chrome for Android is fast and easy to use. Even though you can’t run extensions in Chrome’s mobile app, you can tweak a lot of settings to make it work just the way you want. Today we’ll cover some more advanced, lesser-known tips for using Chrome on Android. 1. Access Experimental Features in Chrome Just like on Windows, Mac, and Linux, there are experimental features available in Chrome on Android that are not enabled by default. These features, known as flags, may not work perfectly and Google might remove them at any time. They could cause problems in Chrome and...

Read the full article: 10 Ways to Supercharge Your Chrome Browser on Android


Read Full Article

Waymo and Google launch a self-driving truck pilot in Atlanta


 Waymo’s autonomous trucking program is coming along – though we haven’t heard much about it since discovering it was a real thing last year, Waymo today announced that it’s launching a pilot program in Atlanta to focus specifically on self-driving trucks and automated logistics. The pilot is being done in partnership with Google, another Alphabet company and… Read More
Read Full Article

What Apple Missed: 8 Awesome Apps That Should Be Preloaded on Mac


awesome-mac-apps

Your Mac comes with a ton of great tools that let you accomplish most common tasks without having to install more software. Some of them aren’t great, but they do the job. Then there are other areas where Apple hasn’t gone far enough, or has avoided entirely. You could make a good case for Apple never including any of these tools (or copies thereof), since they all do the job just fine. But for the purposes of the title that landed you on this page, here are 8 apps we’d love to see built into Mac. 1. KnockKnock You can...

Read the full article: What Apple Missed: 8 Awesome Apps That Should Be Preloaded on Mac


Read Full Article

GeChic On-Lap 1305H Portable Monitor Review: The Best Portable Display That Doesn’t Need Drivers


Our verdict of the GeChic On-Lap 1305H:The best quality image in its class. With power from any USB port and a standard HDMI input, it can be used with any source, and doesn't require drivers installed. Portrait mode is also possible, but it's a little delicate.710Portable monitors lend themselves to many tasks. Productivity hounds enjoy the added workspace of second screen when they travel. Some of us just enjoy their sheer versatility, requiring no more than a meagre USB port for power. Even home users may find some benefit from having a second screen for videos, while tackling emails on...

Read the full article: GeChic On-Lap 1305H Portable Monitor Review: The Best Portable Display That Doesn’t Need Drivers


Read Full Article

How to Clear Recent Documents History in Microsoft Word


It’s not enough if you only learn the basics of Microsoft Word and its productivity secrets. You have to practice a few security habits to keep your work private too! And one of those habits is to clear (or even disable) the Recent Documents list in Microsoft Word. Not only is it a good way to declutter the launch screen on Word, it also staves off any prying eyes. How to Clear the Recent Documents List in Word The recent documents list appears on the left side of the screen. You can remove all signs of the latest opened files or disable the list...

Read the full article: How to Clear Recent Documents History in Microsoft Word


Read Full Article

How to Exclude Specific Domains in Google Search Results


make-google-search-great-again

There’s no denying the power of Google search operators. With just a little added effort, you can make sure that you’re finding exactly the content you’re looking for—and one way to do that is to exclude results from sites that you don’t care about. How to Exclude Keywords in Google Search Results When it comes to excluding specific keywords or items from your search results, all you need to do is place a hyphen in front of the word without a space. So if you wanted to search for articles about macro lenses for photography but didn’t want anything about Nikon,...

Read the full article: How to Exclude Specific Domains in Google Search Results


Read Full Article

How to Change Folder Colors on Mac


Want to add some personality to your Mac? A great way to do so is by changing the color of a few folders. Whether you do this to dress up your desktop a bit, make Finder easier to use, or want to use color to easily identify certain types of folders, you can do it in just a few moments. Here’s how. How to Change Folder Colors on Mac Create a new folder if you don’t want to change an existing one. Right-click on the folder you want to change and choose Get Info. At the top of this dialog,...

Read the full article: How to Change Folder Colors on Mac


Read Full Article

You Can Now Visit Disney Parks Using Google Street View


Love it or hate it, Disney is a phenomenal brand. And as Disney now owns Star Wars that brand is reaching out to geeks in a big way. It is, therefore, rather fitting that you can now take a virtual tour of a number of Disney Parks using Google Street View. There are some really geeky virtual tours you can take with Street View, and Google are adding more all the time. Google Street View has even made it into space, allowing us Earth-dwelling proles the chance to walk around the International Space Station. Visit Disney Parks Without Leaving Home...

Read the full article: You Can Now Visit Disney Parks Using Google Street View


Read Full Article

Facebook had ‘a negative role’ in politics says co-founder Chris Hughes


 “Facebook has played at times a negative role in the political discourse,” according to Facebook co-founder and Mark Zuckerberg’s Harvard roommate Chris Hughes. “The algorithms are not neutral,” he said today at a Bloomberg Beta event promoting his new book Fair Shot: Rethinking Inequality and How We Earn. Mentioning “the filter bubbles” and… Read More
Read Full Article

Mine Bitcoin on Linux With These Best Free Apps


mine-bitcoin-linux

Cryptocurrency mining, especially Bitcoin mining, rose in popularity as the price of Bitcoins skyrocketed. Though Bitcoins are available for purchase, a rewarding means of obtaining cryptocurrencies is through mining. On Windows and Mac, many miners are available. Similarly, for Linux users, Bitcoin mining options are plentiful. Learn how to mine Bitcoin on Linux with these best free apps! Before You Mine Bitcoin With Free Apps Primarily, when mining Bitcoin with free apps for Linux, think about your software and hardware. Most Ubuntu and Debian derivatives should foster a suitable environment for installing these free Bitcoin miners. Still, it’s best to...

Read the full article: Mine Bitcoin on Linux With These Best Free Apps


Read Full Article

Android beats iOS in smartphone loyalty, study finds


 Samsung’s new Galaxy S9 may not quite live up to the iPhone X when it comes to Samsung’s implementation of a Face ID-style system or its odd take on AR emoji. But that’s not going to matter much to Samsung device owners – not only because the S9 is a good smartphone overall, but because Android users just aren’t switching to iPhone anymore. In fact, Android users… Read More
Read Full Article

Android beats iOS in smartphone loyalty, study finds


 Samsung’s new Galaxy S9 may not quite live up to the iPhone X when it comes to Samsung’s implementation of a Face ID-style system or its odd take on AR emoji. But that’s not going to matter much to Samsung device owners – not only because the S9 is a good smartphone overall, but because Android users just aren’t switching to iPhone anymore. In fact, Android users… Read More

Read Full Article

Discover your own planets Google-style with AstroNet


 You may remember that in December, a pair of new planets was discovered via machine learning, using NASA data and Google Brain ingenuity. If you’re interested in doing the same, the team behind the system has just released the code they used to accomplish this astronomical achievement, which they call AstroNet. Read More
Read Full Article

5 Awesome DIY Lego Raspberry Pi Cases


raspi-lego-case

Perhaps the most important item you can buy for your Raspberry Pi is a case. Generally inexpensive, they serve to protect your little computer from dust, tiny fingers, and paws. In some cases, they’re rugged enough to protect your Pi from the elements. But what if you want a more DIY approach? Many options are available for DIY builds, from using the approved cardboard punnet to repurposing old hardware such as games cartridges. You might even 3D print a case, or simply bolt two pieces of wood around your Raspberry Pi. But there’s another option: Lego. The world’s foremost construction...

Read the full article: 5 Awesome DIY Lego Raspberry Pi Cases


Read Full Article

Lorde can now start posting on Google


 Over the last couple of years, Google has been slowly (very slowly!) opening up the ability for celebrities, politicians, sports teams, museums and local businesses to post social media-like updates directly to its search result pages. Today, after testing this in a few markets over the course of the last year, Google is also opening this feature up to musicians in the U.S. This means that… Read More
Read Full Article

False news spreads faster than truth online thanks to human nature


 The rapidity with which falsity travels is proverbial: “Falsehood flies, and the Truth comes limping after it,” wrote Swift in 1710. Yet evidence of this common wisdom has been scarce. A new study from MIT looks at a decade of tweets, and finds that not only is the truth slower to spread, but that the threat of bots and the natural effects of social media are no excuse:… Read More
Read Full Article

Samsung’s Galaxy S9 deserves better than Bixby


 Bixby is, by far, the most frustrating part of the S9. It’s true, of course, that no voice assistant is perfect, but a full year after Samsung’s smart assistant debuted on the Galaxy S8, it still feels unfinished. Of course, the company’s made some significant progress, and a slew of artificial intelligence startup acquisitions point to an offering that’s going to… Read More

Read Full Article

CodeStream wants to move developer chat straight to the source code


 There are tons of services out there from Slack to Jira that are designed to help developers communicate with one another about code issues, but there is a surprising dearth of tools that have been purpose-built to provide communication capabilities right in the IDE where developers work. CodeStream, a member of the Y Combinator Winter 2018, aims to fix that. “We are team chat in your… Read More
Read Full Article

Google promises publishers an alternative to AMP


 Google’s AMP project is not uncontroversial. Users often love it because it makes mobile sites load almost instantly. Publishers often hate it because they feel like they are giving Google too much control in return for better placement on its search pages. Now Google proposes to bring some of the lessons it learned from AMP to the web as a whole. Ideally, this means that users will… Read More
Read Full Article

Open Sourcing the Hunt for Exoplanets




Recently, we discovered two exoplanets by training a neural network to analyze data from NASA’s Kepler space telescope and accurately identify the most promising planet signals. And while this was only an initial analysis of ~700 stars, we consider this a successful proof-of-concept for using machine learning to discover exoplanets, and more generally another example of using machine learning to make meaningful gains in a variety of scientific disciplines (e.g. healthcare, quantum chemistry, and fusion research).

Today, we’re excited to release our code for processing the Kepler data, training our neural network model, and making predictions about new candidate signals. We hope this release will prove a useful starting point for developing similar models for other NASA missions, like K2 (Kepler’s second mission) and the upcoming Transiting Exoplanet Survey Satellite mission. As well as announcing the release of our code, we’d also like take this opportunity to dig a bit deeper into how our model works.

A Planet Hunting Primer
First, let’s consider how data collected by the Kepler telescope is used to detect the presence of a planet. The plot below is called a light curve, and it shows the brightness of the star (as measured by Kepler’s photometer) over time. When a planet passes in front of the star, it temporarily blocks some of the light, which causes the measured brightness to decrease and then increase again shortly thereafter, causing a “U-shaped” dip in the light curve.
A light curve from the Kepler space telescope with a “U-shaped” dip that indicates a transiting exoplanet.
However, other astronomical and instrumental phenomena can also cause the measured brightness of a star to decrease, including binary star systems, starspots, cosmic ray hits on Kepler’s photometer, and instrumental noise.
The first light curve has a “V-shaped” pattern that tells us that a very large object (i.e. another star) passed in front of the star that Kepler was observing. The second light curve contains two places where the brightness decreases, which indicates a binary system with one bright and one dim star: the larger dip is caused by the dimmer star passing in front of the brighter star, and vice versa. The third light curve is one example of the many other non-planet signals where the measured brightness of a star appears to decrease.
To search for planets in Kepler data, scientists use automated software (e.g. the Kepler data processing pipeline) to detect signals that might be caused by planets, and then manually follow up to decide whether each signal is a planet or a false positive. To avoid being overwhelmed with more signals than they can manage, the scientists apply a cutoff to the automated detections: those with signal-to-noise ratios above a fixed threshold are deemed worthy of follow-up analysis, while all detections below the threshold are discarded. Even with this cutoff, the number of detections is still formidable: to date, over 30,000 detected Kepler signals have been manually examined, and about 2,500 of those have been validated as actual planets!

Perhaps you’re wondering: does the signal-to-noise cutoff cause some real planet signals to be missed? The answer is, yes! However, if astronomers need to manually follow up on every detection, it’s not really worthwhile to lower the threshold, because as the threshold decreases the rate of false positive detections increases rapidly and actual planet detections become increasingly rare. However, there’s a tantalizing incentive: it’s possible that some potentially habitable planets like Earth, which are relatively small and orbit around relatively dim stars, might be hiding just below the traditional detection threshold — there might be hidden gems still undiscovered in the Kepler data!

A Machine Learning Approach
The Google Brain team applies machine learning to a diverse variety of data, from human genomes to sketches to formal mathematical logic. Considering the massive amount of data collected by the Kepler telescope, we wondered what we might find if we used machine learning to analyze some of the previously unexplored Kepler data. To find out, we teamed up with Andrew Vanderburg at UT Austin and developed a neural network to help search the low signal-to-noise detections for planets.
We trained a convolutional neural network (CNN) to predict the probability that a given Kepler signal is caused by a planet. We chose a CNN because they have been very successful in other problems with spatial and/or temporal structure, like audio generation and image classification.
Luckily, we had 30,000 Kepler signals that had already been manually examined and classified by humans. We used a subset of around 15,000 of these signals, of which around 3,500 were verified planets or strong planet candidates, to train our neural network to distinguish planets from false positives. The inputs to our network are two separate views of the same light curve: a wide view that allows the model to examine signals elsewhere on the light curve (e.g., a secondary signal caused by a binary star), and a zoomed-in view that enables the model to closely examine the shape of the detected signal (e.g., to distinguish “U-shaped” signals from “V-shaped” signals).

Once we had trained our model, we investigated the features it learned about light curves to see if they matched with our expectations. One technique we used (originally suggested in this paper) was to systematically occlude small regions of the input light curves to see whether the model’s output changed. Regions that are particularly important to the model’s decision will change the output prediction if they are occluded, but occluding unimportant regions will not have a significant effect. Below is a light curve from a binary star that our model correctly predicts is not a planet. The points highlighted in green are the points that most change the model’s output prediction when occluded, and they correspond exactly to the secondary “dip” indicative of a binary system. When those points are occluded, the model’s output prediction changes from ~0% probability of being a planet to ~40% probability of being a planet. So, those points are part of the reason the model rejects this light curve, but the model uses other evidence as well - for example, zooming in on the centred primary dip shows that it's actually “V-shaped”, which is also indicative of a binary system.
Searching for New Planets
Once we were confident with our model’s predictions, we tested its effectiveness by searching for new planets in a small set 670 stars. We chose these stars because they were already known to have multiple orbiting planets, and we believed that some of these stars might host additional planets that had not yet been detected. Importantly, we allowed our search to include signals that were below the signal-to-noise threshold that astronomers had previously considered. As expected, our neural network rejected most of these signals as spurious detections, but a handful of promising candidates rose to the top, including our two newly discovered planets: Kepler-90 i and Kepler-80 g.

Find your own Planet(s)!
Let’s take a look at how the code released today can help (re-)discover the planet Kepler-90 i. The first step is to train a model by following the instructions on the code’s home page. It takes a while to download and process the data from the Kepler telescope, but once that’s done, it’s relatively fast to train a model and make predictions about new signals. One way to find new signals to show the model is to use an algorithm called Box Least Squares (BLS), which searches for periodic “box shaped” dips in brightness (see below). The BLS algorithm will detect “U-shaped” planet signals, “V-shaped” binary star signals and many other types of false positive signals to show the model. There are various freely available software implementations of the BLS algorithm, including VARTOOLS and LcTools. Alternatively, you can even look for candidate planet transits by eye, like the Planet Hunters.
A low signal-to-noise detection in the light curve of the Kepler 90 star detected by the BLS algorithm. The detection has period 14.44912 days, duration 2.70408 hours (0.11267 days) beginning 2.2 days after 12:00 on 1/1/2009 (the year the Kepler telescope launched).
To run this detected signal though our trained model, we simply execute the following command:
python predict.py  --kepler_id=11442793 --period=14.44912 --t0=2.2
--duration=0.11267 --kepler_data_dir=$HOME/astronet/kepler
--output_image_file=$HOME/astronet/kepler-90i.png
--model_dir=$HOME/astronet/model
The output of the command is prediction = 0.94, which means the model is 94% certain that this signal is a real planet. Of course, this is only a small step in the overall process of discovering and validating an exoplanet: the model’s prediction is not proof one way or the other. The process of validating this signal as a real exoplanet requires significant follow-up work by an expert astronomer — see Sections 6.3 and 6.4 of our paper for the full details. In this particular case, our follow-up analysis validated this signal as a bona fide exoplanet, and it’s now called Kepler-90 i!
Our work here is far from done. We’ve only searched 670 stars out of 200,000 observed by Kepler — who knows what we might find when we turn our technique to the entire dataset. Before we do that, though, we have a few improvements we want to make to our model. As we discussed in our paper, our model is not yet as good at rejecting binary stars and instrumental false positives as some more mature computer heuristics. We’re hard at work improving our model, and now that it’s open sourced, we hope others will do the same!

If you’d like to learn more, Chris is featured on the latest episode of This Week In Machine Learning discussing his work.