pilooch

Amazed at the very poor quality of their search! I ve setup a search engine for a service that then links to spotify when the song is available. With use of proper phonetic filters etc... we were able to get excellent relevance. I do not understand how spotify works internally to let this unattended. Other than that, good music ;)

There's the open source path: while you're looking for a job, get involved in an open source project that solves a problem you like, really can be anything. Get involved, participate, share ideas, code stuff up, commit. This may take longer, but this is a good way to get a job in something you like. Because contributors are employees somewhere, and they will recognize your talent, and know where to apply it. Good luck!

Since they were supposed to be the first to leave Earth, this sounds like another bad signal for us...

OK this is my life too. But there are clever ways of unit testing statistical systems. Hey, not that I do all I say, just saying ;) Sometimes it saves many days.

you would not buy surgery using this currency if you can avoid it since it would gain value in a possibly near future. So while people would still buy when prices fall, but with a preference given to other currencies.

Note: I am no expert.

it should be like in war, bombing civilians is a war crime, spying on even foreign civilians should never be legally accepted. ever. otherwise we re on the fast track toward another catastrophe for the human race and one of its unique features, freedom.

A very good source of information, and tutorials with source code is http://deeplearning.net/tutorial/contents.html

hyperparameter search is a solution. See http://jmlr.org/papers/v13/bergstra12a.html and http://jaberg.github.io/hyperopt/

put it on fdroid

Hopefully self driving car are on the way ;)

interesting but why is this a remote service that needs to parse in-browser content? if filtering is keyword based (as opposed to ml based), why not filter locally instead ?

FYI I use Solr cloud for that purpose.

Interestingly, Linux is moving to win on several fronts. Economy and debt crisis, in addition to rising Ubuntu-like UI (easy to use, LibreOffice included, etc...) increase the benefits for governments and agencies. That's one front.

On another front, Linux clearly being the top platform for software development, it can naturally then be a top OS for game development, and become a top gaming platform itself, typically with SteamOS and Android.

At Ubuntu install parties, it's been a pleasure to see older French citizens coming in with already good information, knowledge and know-how of the Linux platform, for some because of their daily job at government agencies. And they want the OS at home! What else ;)

I love rien, I am a Parisien ;)

The title is a bit confusing as probabilistic programming is a research field itself that the book seems to not touch upon. See http://probabilistic-programming.org/wiki/Home

yeah right, and my bill looks like this: http://i.imgur.com/WxZCFrH.png

Thanks to Google maths I presume!!

hi, thanks for your interest, the source code is here: https://sourceforge.net/p/seeks/code/ci/experimental/tree/

Right, one way is to share one instance with someone you trust (and who runs a linux server), and to use it remotely.

Probably none yet, unfortunately. Seeks is usable because it relies on Google or any other engine in the background, when no results are available among peers.

Yacy is more advanced, uses a DHT, but does not really work for daily use, the accuracy being too low.

Unhosted is distributed architecture on top of the Web really. I've mentioned it because it may be useful and/or an inspiration to whoever interested in these matters.

DDG is nice initiative. The problem is that it does not solve any problem related to privacy because it is based on a matter of trust.

Distributed search, similar to bittorrent, DHT-based designs and the like are notoriously difficult.

I've participated in such efforts, like the Seeks Project [1], Yacy [2], and related initiatives like Unhosted [3], and it takes a certain amount of dedication (and suffering ;) ).

However, I believe it is not entirely impossible that we see a true alternative sometimes. From what experience, what is needed is a slightly better set of distributed algorithms, a business model with the ability to sustain such a technical effort, and a range of features that no search engine can yet offer (because centralized).

[1] http://www.seeks-project.info/ [2] http://www.yacy.net/ [3] https://www.unhosted.org/

So, actually one I've discovered, and tried with great success, is Mumble, http://mumble.sourceforge.net/

Someone here said 'Champagne holder' ;)

many thanks! It may well be the one.

right, we already use an IRC server over VPN. Definitely the best for chat and logs I agree!

haha, everyone would like to kill URSAFF, but they are your life after your biological CPU has turned into a good old Pentium :)

And yes Limousin, may not be your startup hub of choice, but after all, it is probably cheaper than everywhere else. Probably good enough while you are in the coding phase, not that good afterwards for sure.

The language problem is real though, and I understand your move much better now.

We're in Toulouse, some other startups I know are in Paris. Ping me if you ever come down South (everything above Bordeaux is Great far 'beyond the Wall' North for us :) ).

For the past month, I've been checking on G+ before Twitter and other media or sources. Here is how I was dragged into it.

I had never used Facebook, and I had never used Google+. I guess I was too busy with Emacs.

Then recently I had to hit Google+ to follow some community thread on the Deep Learning channel, https://plus.google.com/communities/112866381580457264725

Somehow, the researchers from the field did settle there. Next I started reading the Machine Learning community threads, then the C++ one, etc... Eventually I did post to the Machine Learning community, and I saw the hits on our website, a few hundreds every day, for a few days, not much, but I could tell the right people were visiting, because they were spending way more time than visitors from other sources.

So my current understanding is that G+ is worth for its communities, and the targeted traffic it offers... my 2cents :)

Interested in the reasons behind the 'no brainer' move. We're in France, and the regime for startups, with innovation tax cuts and other security nets is pretty favorable. The engineers are pretty good too!

Check this out maybe, http://www.reddit.com/r/MachineLearning/comments/1dzoim/a_su...

With others, I recently played around with a custom built dataset, using unsupervised machine learning to cluster the subreddits, based on content.

You can play with it here:

http://demo.xplr.com/xplr/umbreddit/

The system can be queried:

http://demo.xplr.com/xplr/umbreddit/?search=cyclists

http://demo.xplr.com/xplr/umbreddit/?search=species

For the curious, more technical information is available here: https://xplr.com/xplr-umbrella-dataviz-on-top-of-unsupervise... and here: https://xplr.com/a-subreddit-recommender-with-xplr/

This analysis is slightly different than the study of cross-posed links. Here the content of posted URLs (not yet comments)is analyzed and subreddits are put into clusters within a search engine.

This allows the easy building of a subreddit recommender for Reddit (Chrome only for now): https://chrome.google.com/webstore/detail/preddit/epicmjpmnm...

I'm pretty sure this would be useful to me, but at install it does ask for 'access your data on all website'. Now, what does this mean exactly ? I do understand this is very much probably not this extension's fault, but Chrome being very vague.