NAVINFLUOR

Double Top Buy triggered above 3704.66 daily close on 3% Box Size chart. https://t.co/b5KkTmtjl5

I'm gonna do my other star wars thread tomorrow, which will primarily be me yelling at a @RowanKaiser-shaped strawman for calling the First Order and The Resistance "cosplayers who the rest of the galaxy has to deal with"

the deal w/the New Republic, the First Order, and the Resistance is very weird and complex and I actually take issue with how Story Group frames the actual conflict.

pre-TFA stuff is presented as a "cold war" between the New Republic and the remnants of the Empire that form the First Order, which 100% makes sense.

But they frame the First Order's attack on the Hosnian system and the subsequent battle that takes place across TFA and TLJ as a full on war when it's very clearly a coup. A war implies at least equal standing, but the First Order completely wrecks shop and takes over the galaxy.

ok never mind I'm fucking doing this now.

The First Order fleet we see in TLJ is *not* the entire fleet of the First Order (I don't even think Starkiller Base was their main base of operations honestly, it couldn't have been considering they were operating out of Wild Space)

I got a Toshiba T1200, which has a bad screen, and a dead CMOS battery, and I thought the hard drive was broken too...

but I rebooted it and heard a KTHUNK and a really loud spin-up noise and holy shit, the hard drive works!

and it still has files on it. the newest date is January 1994.

It's a 21mb hard drive, with ~37k of bad sectors.

This fucker is never going to spin up again, so I better go find some serial cables and copy shit off now.

The backlight is fucked so I'm shining a flashlight on it.
without it, it looks like this.

(1/3)

I am using only one indicator in the chart.
i.e DC(20) = Donchain Channel
I use 3 min chart and the DC period is 20, so DC high is basically the last 1 hour high & DC low is the last 1 hour low.
The middle line is the median of high and low.

(2/3)

Donchian channels are mainly used to identify the breakout of nifty/BNF.
We can take directional bias based on the indicator.
If nifty is breaking DC high, can take bullish view.
if it is breaking DC down, can take bearish view.
if it is in between, can trade range-bound.

(3/3)

We can monitor nifty/banknifty futures with volume along with DC.
If the volume is 1.5 to 2 times the average volume and if DC high/low is taken out then it can be a strong bullish/bearish signal.

Important paper from Google on large batch optimization. They do impressively careful experiments measuring # iterations needed to achieve target validation error at various batch sizes. The main "surprise" is the lack of surprises. [thread]

https://t.co/7QIx5CFdfJ

The paper is a good example of lots of elements of good experimental design. They validate their metric by showing lots of variants give consistent results. They tune hyperparamters separately for each condition, check that optimum isn't at the endpoints, and measure sensitivity.

They have separate experiments where the hold fixed # iterations and # epochs, which (as they explain) measure very different things. They avoid confounds, such as batch norm's artificial dependence between batch size and regularization strength.

When the experiments are done carefully enough, the results are remarkably consistent between different datasets and architectures. Qualitatively, MNIST behaves just like ImageNet.

Importantly, they don't find any evidence for a "sharp/flat optima" effect whereby better optimization leads to worse final results. They have a good discussion of experimental artifacts/confounds in past papers where such effects were reported.