markovbling

Totally - I used to work as quant for a boutique asset manager and the whole business was running on spreadsheets.

Insane.

Ended up putting it all in a database and developing an excel add-in to pull it from the database as array formulae. Used a great library called Excel DNA to develop the add-in using C# if anyone is interested.

I think there should be SAAS pricing style "packages" curricula like 1 week version learn ABC by reading XYZ, 4 week version, 6 month version, etc

People generally think in terms of how much time they have - these lists are great but hard to act on - I'd like to know if I spend x hours, I'll learn y skills

Protip: use karabiner to bind command to control when you're using a windows VM ;)

Very cool but wouldn't say simple haha!

I implemented a Hough circle detector in JavaScript that might be of interest - it's got UI to tweak the accumulator threshold and is all first principles without any calls to black box functions:

https://github.com/alexanderconway/Hough-Circle-Detection

Great, thanks!

I find the problem is actually how to select what content to study given a time constraint like if you had 5 hours or 20 hours or 200 hours - what should you read?

Like an exhaustive list is great but it's an optimization problem - how do I maximise understanding subject to a time constraint - which resources do I select to maximise learning in x hours/days/years?

Machine learning is subset of statistics.

The standard text in ML, "The Elements of Statistical Learning" is authored by statistics Professors.

Statistics is the new statistics. The rest is marketing bullshit.

Awesome!

Should make them take a video and pull a frame of the video - much harder to fake

Yes! The approach I took was to build a product that solves a broad problem (in my case geographic visualisation) and showed that to get clients who wanted something similar but tweaked for their use case. Of course they then ask 'is it possible to do this other thing' and all of a sudden you're doing data vis consulting.

As an aside, I'd suggest selling the data science too if you have the skills since data viz isn't just about showing data but figuring out how to show it to answer the relevant business questions - they go hand-in-hand and it's both a more compelling sell and adds more value to deliver the full stack solution instead of just the viz piece.

Good luck!

Yes, I got a data science job advertised as masters only by showing past work and painting a picture of how I could do similar for their industry specific problems. I think credentials are often more a factor in getting the interview which can be hacked through connections and good networking but once you're in the interview, you can often communicate compentency through past work rather than a stamped piece of paper. That said, I've since done a MSc in Statistics and it's definitely easier. Think it helps but not strictly required in all cases (some companies will insist but think especially now with such a shortage of skills, I think you can get away with just a solid portfolio and good networking skills)

Awesome! The GIF is great but would really love a link to a live demo to play with :)

Didn't know could do heatmap via leaflet - does it respond to crossfilter filtering?

there are a lot of gotchas that don't seem like errors but completely invalidate analysis when done without a thorough understanding of a technique.

For example you can learn regression from a stats textbook but unless you've gone through a thorough (and painful) graduate-level stats course, you probably haven't seen the edge cases that invalidate assumptions and necessitate a more complex regression e.g. your regression may suggest there is no effect but when you look at the residuals, you may find systematic bias that you can model using a subject-specific random effect or some transformation as a generalized linear model...

That isn't to say you need a graduate level stats degree but applying statistics without understanding the pitfalls can lead to seriously wrong conclusions.

That is normal statistics though to be fair

There is learning - the model is learnt from the data

It's lambda - basically the amount of regularization (simplification or 'overfit penalty') to impose. Usually chosen by cross-validation - try every lambda between say 0 and 10 in 0.5 increments and choose lambda that gives the model the lowest error cross validation measure...

Awesome - great work!

A tutorial on how to link it to a mysql database would be greatly appreciated :)

awesome!

Awesome! Think you should convert the currencies at current spot rates instead of just removing the currency sign though :)

Interesting idea!

<3 Nujabes RIP

For the hip hop heads check out 9th Wonder: https://www.youtube.com/watch?v=DBX_ADoO-7E

Very cool! I came across Peter Beshai's work a couple of weeks ago - he has a great d3.js visualisation of basketball shots that might be of interest :)

http://peterbeshai.com/buckets

Would love to hear replies on this - just because it's against their terms of use, does that mean it's illegal? Just because someone says they don't want you to look at them and write down what you see doesn't make it illegal. Of course it might be stalking / harassment in the legal sense but where's that line for spidering websites?

Would love to see a blog post on how your hosted service works :)

Sad to see Africa so empty!

Great job! :)

R is a lot like javascript - seems like voodoo until you 'get it' (get comfortable using it) ... and then it's the obvious choice for many common tasks.

In the abstract, people buy what makes them feel good.

This can be "direct purchase" like nice clothes, good food, vacations, property, furniture, life insurance, etc. Or it can be an "indirect purchase" in the form of attention which equals money in aggregate since some percent of youtube viewers will spend money downstream as a result of either paying the public speaker or clicking the ad before the youtube vid and paying an advertiser (public speaker gets a cut), etc.

A million writers/public speakers have cropped up because there is a market - people are willing to pay indirectly with their attention - for 'feel good' snake-oil style self-help advice.

The trouble is filtering the signal from the noise where a good filter tends to be whether the person providing you with productivity advice needs your money.

Interestingly, the value of productivity / 'self-help' advice (applied to personal or entrepreneurial life) is usually proportional to the person providing it's net worth. Peter Thiel's 'Zero to One', for example, is excellent.

Most of the time self-help is just a circle jerk though

Pre-workout = caffeine + marketing

I'm sure they could process exported gmail data dump: https://www.google.com/settings/takeout

An analogy would be this social graph network analysis tutorial which walks you through exporting your Facebook social graph data and creating cluster graph on it using a tool called Gephi: https://www.youtube.com/watch?v=kbLFMObmLNQ

South Africa represent! Be cool to meet up if you're in Cape Town some time :)

If you want a thorough explanation of concepts (as opposed to the 'black box' 'use this library' approach taken by many tutorials I've seen) then I highly recommend John Foreman's (Data Scientist at MailChimp) book "Data Smart".

He implements algorithms on practical problems in Excel. I know that may sound off-putting but it's so much easier to visualise what's going on when you can see vectors of data change in a spreadsheet compared to trying to debug arrays in a more general programming language.

What led me to John - something I heard about here actually - was his blog called 'Analytics Made Skeezy' where he creates fictional 'drug dealer' GTA style examples such as "Forecasting Made Skeezy: Projecting Meth Demand using Exponential Smoothing"

http://analyticsmadeskeezy.com/blog/

I have no affiliation with John, just liked the blog enough to order his book and the book was excellent too

Good luck! :)