| blog
2014 Fall Hackathon Outcome
On October 18-19 of 2014, we held our fourth NuPIC Hackathon in San Jose at Pinger, Inc. It was a 36-hour event, starting at 10AM on Saturday, and ending with hack demos at 4PM Sunday afternoon. We started with a Kickoff presentation to go over hackathon protocol, then jumped right into hacking. There were some interesting themes this time. A lot more people tried using cortical.io’s word fingerprinting service for natural language processing hacks. We also had a lot of interest in EEG analysis with NuPIC, which seems to be a hot topic within the community. And we even had another robotics demo!
I think I say this after every NuPIC hackathon we host, but this hackathon was the best one yet! We had well over 50 participants, 16 demos, and a ton of fun and community bonding. It was really nice to see some new faces at this hackathon. I love meeting the people of our community!
Watch Video on YouTube.
Once again, we had attendees from all over the world: Amsterdam, China, Austria, Ireland, not to mention from all over the United States. I was really impressed how with each hackathon our community gets a better understanding of NuPIC and how it can be used to tackle many different kinds of problems. Even though we had a high percentage of new hackers this time, I could really see the light bulbs turning on as I talked to people about their hackathon ideas.
I’ve separated this blog post into sections:
-
Photos: Taken at the hackathon (see them all on Flickr)
-
Testimonials: What people have been saying about the hackathon
-
Sessions: Videos of informational talks given by Numenta and guests
-
Demos: Hackathon demos presented at the end
Photos
Everyone seems excited as the hackathon demos get started.
Another big thanks to Pinger, who let us use their offices to host the event once again.
Jeff Hawkins explains HTM concepts on a whiteboard.
Felix Andrews talks about HTM implemented in Clojure.
Chetan and Yuwei's robot from the Whole New World demo.
Marion Le Borgne and Artem Avdacev watching hackathon demos.
Scott Purdy sharing his knowledge with Daniel McDonald and Mark Whelan.
Yuwei explains the representations displayed from the Whole New World demo.
Jeff Hawkins and a guest on the way to the hackathon demos.
Testimonials
More people than ever, and a more diverse community than ever. We had pro hackers, students, software professionals, a neurologist, and even some new to programming in general.
-
"Had an awesome time!! Thanks again to Matt and Scott for helping me so much with getting started with nupic."
-
"Thanks Matt and the whole Numenta crew for a wonderful weekend. It was a good chance to prototype how NuPIC can be used for wearables."
-
"Thank you Matt for making this an unforgettable event. I am already looking forward to the next Hack-a-thon!"
-
"Thanks For hosting the workshop and hackathon Nupic team . Had a great time!"
-
"I really had a good time. I learned so much and loved seeing the incredible variety of hacks."
-
"This was inspiring!"
-
"A fantastic event. It underscores Numenta’s approach of being totally open with their work and supportive of the community. It feels like we are at the cusp of a revolution, where a few more good ideas will really make this thing fly."
Sessions
We only gave a few presentations from Numenta and friends for hackathon participants to attend in order to give more time for attendees to work on their projects. You can watch all the recordings we took at the hackathon in this playlist.
Hackathon Kickoff
Matt Taylor & Jeff Hawkins
Where hackers are welcomed to the hackathon, protocol is reviewed, and ideas are brainstormed.
EEGs and NuPIC
Dr. Richard Pantera
We had a neurologist attending the hackathon, who was kind enough to sit in front of a camera for us and let us pick his brain about EEGs.
On the Way to Language Intelligence
Francisco Webber, cortical.io
Francisco talks about language, intelligence, and cortical.io's new REST API version 2.0 and all its capabilities.
Applications of Hierarchical Temporal Memory (HTM)
Chetan Surpur, Numenta Engineer
This talk is from the Numenta Workshop that preceeded the hackathon, but Chetan gave the same talk at the hackathon. Since the recording at the workshop was a better quality than the hackathon presentation, I've included the first one.
Q&A with Jeff Hawkins
Jeff Hawkins
Jeff Hawkins takes questions about HTM theory from hackathon attendees and explains the answers in front of a whiteboard.
Demos
Games, physics, robotics, geology, natural language, geospatial analysis… demos at this hackathon spanned a large breadth of topics. We were really impressed with the types of things hackers attempted to perform with NuPIC. When we have hackathons, it’s a chance for people interested in NuPIC and HTM to push the boundaries of HTM technology and really see what it might be capable of. I always come away inspired and excited about the future of NuPIC and HTM, and this hackathon was no exception at all.
Not all hacks are success stories, but each one is a learning experience. We’ll never know what can be achieved with cortically-inspired machine intelligence unless we try to solve hard problems.
Pendulum
Alejandro Schuler, Andrew Morrison, Shashwat Kandadai
An attempt to balance an inverted pendulum using predicted data.
MineHack
Matt Taylor
I created a Minecraft mod that exports player X,Y,Z coordinates into NuPIC using the CoordinateEncoder to get anomaly indications for a live player.
Shake Hack
Austin Marshall
According to recent news, the Bay Area is overdue another "Big One". I'll be evaluating encoding schemes for data that comprises the last 10+ years of magnitude 2.5 or greater earthquake activity for the 1000 km radius centered on the Pinger headquarters in San Jose, CA. I'm hoping to identify earthquake swarms leading up to larger events in the same region.
A Whole New World
Chetan Surpur, Yuwei Cui
A demonstration of sensorimotor inference in simple robotics.
Semantic Fingerprint Sentence Generation
Daniel McDonald, Mark Whelan
In this hack, we will attempt to generate semantic fingerprints from WordNet semantic relationships and train the HTM to recognize sequences of meaning from training texts. The trained HTM will be used to generate English sentences by using the predicted sequence of SDRs from the HTM to select words from the training set to fill in blanks in the sentences generated according to a limited English grammar.
Smart Harbor
Daniel Ducro, Egbert Wietses
- Video
- Source code (coming soon?)
Tracking geolocations of cargo ships in the port of Rotterdam.
Temporal Memory in Racket
Rian Shams
A Racket-based implementation of the current temporal memory algorithm. Amazingly, this was Rian's first real program. It's very impressive that he chose to implement HTM in a Lisp!
The Jinglinator 4000
Sergey Alexashenko
I am training NuPIC on a dataset of 500 jingles and generating new jingles based on input vectors of a few notes.
Corti-Yelp
Jim Bridgewater, Artem Avdacev
Using cortical.io to analyze Yelp's academic dataset. This might not have been the most successful hack at the hackathon, but it was one of the most entertaining.
NuPIC EEG
Nicolas Thiebaud
- Video (coming soon)
- Source code (coming soon)
Nicolas used his hack to work on a Kaggle competition, and the rules of the competition prevent us from displaying his video or source code until the competition is over.
AutoDJ
George London
- Video
- Source code (coming soon?)
I'm extracting my track-by-track music listening history from Last.fm and then seeing if NUPIC can predict what artist I'm going to listen to next given a sequence of my previous listens.
Kaggle Bike Share Demand Prediction
Chandan Maruthi
- Video
- Source code (coming soon?)
Chandan tried using NuPIC to solve a Kaggle Bike Sharing competition, where one week of demand data is missing from each month.
We'll Always Have Paris
Pablo Gonzalvez, Soren Madsen, Erik Graf of cortical.io.
Feed a network of htms with articles about different topics. Get the network to extract and learn interesting facts from the input. Then query the network's knowledge.
Heartbeat
Ben Morrow
Apple Watch app for telling if there is an anomaly detected in your heartbeat. You can view the rhythm strip and share the information with your doctor.
EEG Data Classification
Marion Le Borgne
EEG data is classified by NuPIC based upon the thoughts of the subject. EEG data was collected by an OpenBCI board.
As the NuPIC community grows, I continue to be awed and inspired by its passion and perseverance. A sincere thank you to all who have participated on our mailing lists, our code repositories, and our hackathons. You folks are the reason we’ve gone open source, and the reason we continue to thrive. I see a bright future for us, and I truly believe we will lead the path forward to truly intelligent machines built on neocortical principles.
It’s not an easy road, but you are the pioneers helping us pave it for the masses that will follow. I had a wonderful time interacting with all of you, and I look forward to even more ground-breaking work at the next hackathon in Spring 2015.
Matt Taylor
Open Source Community Flag-Bearer
Numenta, Inc.
Thanks to Jeff Fohl for the excellent logo design!
By the way! You can see all the videos and photos taken at every hackathon on our YouTube channel and Flickr page.
Back to Blog