Datascience Chronicles

Burning Out

box_mugI’ve done the first decision tree
It looks all wrong – perhaps it’s me
I’ve tried all week – but something’s lacking
I’m boosting, bagging – even stacking

I’m analysing corporate greed
While trying to mine this Twitter feed
I haven’t found a pattern yet
So next I’ll try a neural net

Tableau won’t help – that’s your obsession
And frankly I prefer regression
Cos SVMs and C four-five
Are faster than Hadoop and Hive

And fuzzy logic is ok
Like R, logit and PCA
But if Bayes himself was here he’d say
Try out the subjective way

We need a pipe that’s built to last
The data’s coming in too fast
The whole team has become frenetic
It’s almost time to get genetic

The data sets all turned up late
There’s nothing to associate
You’re gonna have to learn to wait
Cos nothing seems to correlate

And if you think I’m being dark
You haven’t used Apache Spark
Perhaps you think I’m being grim
A little slow. A little dim?



I’ll tell you what I told the boys
I can’t find a signal in the noise
Even with k nearest neighbours
There’s squat to show for all my labours

And training on paired GPUs
Will leave you wanting more than two
It’s better if you have a few
But there’s still so much you’ll need to do


And mostly I can’t get to sleep
And every paper starts with “Deep”
Forcing yet another leap
The bleating of electric sheep

There’s so much I still need to know
Athena looks just like Presto
There’s Scala, python, R and Go
Theano, Torch and TensorFlow


Tomorrow is another day
I’ll mine the sets another way

And maybe I’ll reach even higher
And find the perfect classifier
And find out where the pattern is
And expose it all through data viz

I’ve given this my heart and soul
I’ve fallen down the rabbit hole
So we’re gonna need alot of luck
Cos I no longer give a fuck

Leave a Reply