AG1LE has set up a Kaggle competition whose goal is to build a machine that learns how to decode audio files containing Morse Code. The Kaggle Morse Challenge was approved a couple of days ago.
Kaggle is actually a very interesting website. According to the website, the Kaggle community includes tens of thousands of PhDs from quantitative fields such as computer science, statistics, econometrics, maths and physics, and industries such as insurance, finance, science, and technology. They come from over 100 countries and 200 universities. In addition to the prize money and data, they use Kaggle to meet, learn, network and collaborate with experts from related fields.
According to AG1LE:
During the competition, the participants build a learning system capable of decoding Morse code. To that end, they get development data consisting of 200 .WAV audio files containing short sequences of randomized Morse code. The data labels are provided for a training set so the participants can self-evaluate their systems. To evaluate their progress and compare themselves with others, they can submit their prediction results on-line to get immediate feedback. A real-time leaderboard shows participants their current standing based on their validation set predictions.
I have also provided sample Python Morse decoder to make it easier too get started. While this software is purely experimental version it has some features of the FLDIGI Morse decoder but implemented using Python instead of C++.
You can of course leverage the experimental multichannel CW decoder I recently implemented on FLDIGI or the standalone version of Bayesian decoder written in C++. There is also some new tools I posted to Github.
The competition ends on December 27, which seems kind of short to me, but this is only phase 1. If this competition is successful, a more difficult competition will be set up. This second competition will distortions introduced by normal radio paths and hand-sent code, which can also be more difficult to answer.
Mauri Niininen AG1LE says
Hi Dan
Thanks for the story.
This is a great opportunity for people who are interested in gaining knowledge about software development, data science, statistical data analysis, machine learning algorithms etc. Kaggle community has been collaborating and solving some very hard problems – like Higgs Boson machine learning challenge – see the list of top challenges on http://www.kaggle.com/ front page.
During the first 3 days of this Morse Learning Machine challenge there has been almost 100 downloads of the provided open source Morse decoder (Python based, there is also C++ code available for free), audio material and sample submission file. We have also the first participants started posting results and my benchmark result was broken in less than 12 hours. For up-to-date status of the challenge you can check the leaderboard: https://inclass.kaggle.com/c/morse-challenge/leaderboard/public
I am doing my best to make it really easy to join this challenge and build a community and knowledge sharing among the participants. You have nothing to lose, but a lot knowledge to gain by joining this challenge.
The link to join is here: https://inclass.kaggle.com/c/morse-challenge
73
Mauri AG1LE – See more at: http://ag1le.blogspot.com/2014/09/morse-learning-machine-challenge.html
Elwood Downey, WB0OEW says
I’ve been reading his blog for a few years now. This could bring it up to a whole new level, I look forward to reading about the new techniques.
I just wish PSK31 would get similar study. Both modes use a binary encoding, neither use any error detection or correction and both would benefit greatly from use of Bayesian maximum likelihood patterns.
Dan KB6NU says
Maybe you should compose a Kaggle contest for PSK. It sounds like that would be a worthwhile effort.
Mauri Niininen AG1LE says
Morse Learning Machine Challenge is on ARRL headline news today!
http://www.arrl.org/news/morse-learning-machine-challenge-catching-on-with-hams