Affordable and Accessible research and innovation in an unseen olympics where many Indian companies win 🥇 and make the country proud. US enterprises are way ahead in cutting-edge Lab R&D, but they come to 🇮🇳 for scale-up and affordable manufacturing (CDMO model).

Go 🇮🇳 Go 👏👏

How to get rich trading (without getting lucky)

Complete beginner’s guide to wealth of Rs 10 Cr in 10 years

Step 1: start trading live

1/n

Nothing can substitute experience of live trading

No amount of back-test, studies or indicators can substitute training your mind with screen time

As a beginner, In live trading, you will make losses almost everyday

2/n

So here are the conditions which you can’t violate in first year
If you can’t follow these rules, this document is not for you

Start with trading capital of Rs 1 lakhs

Max loss per trade 0.25%, i.e. 250 Rs

Max loss for the day, 1%, i.e. Rs 1000

Max 6 trades in a day

3/n

I know, these conditions will cripple you in terms of what all you can trade and how

But just like any start up, trading is also about making it through initial gruelling constraints of life

You will also learn, how to turn these adverse rules to your advantage

4/n

Btw, you can sell directional banknifty options, obviously you can buy options, can do cash stock trading and much more under these conditions

Think about it, you will be do all kinds of instruments, just that PNL will not be roller coaster ride or exciting like a casino

5/n

I want to explain what happened to tumblr overnight. And why Twitter as reached it's dream (or have come ever so close to it)

Yahoo, who bought Tumblr years ago, used to have a huge adult presence on the early net. They allowed adult groups and what not.

However, people and bots (just like now) misused the service, and Yahoo were forced to make a choice. They made private the groups (and later closed them down and sold some of it to other companies) and then ended their chatrooms on yahoo messenger...

after a incident with one of the chatrooms vid cams. The damage was done, Yahoo Messenger lost a lot of people - and with the closing of the groups - backpage and Craigslist came more important.

Now backpage is no more and Craigslist is slowly passing away. Tumblr had a semi strong community, but once 2014 came around and both porn, and political bots exploded the quality started to go down,

ugh i wanted to do a thread about the wars in Star Wars and how the two outlying wars (the clone wars in the prequels and the cold war/civil war in the sequels) are really weird but i don't feel like it any more.

my first argument is that the Clone Wars were a really complex morally gray moment in Star Wars canon in like, 5 different directions and I'd kind of like to know why Lucas did that.

especially with the direct line that's drawn in the OT with the allegory of WW2 and Vietnam with the rebels being the Vietnamese and the Empire being America. That...doesn't exist in the Clone Wars.

for starters: it's weird it's called the Clone Wars. I get why, because that's what Kenobi calls it in STAR WARS, but the clones are a small facet in the larger story of the war.

the clones basically only exist as a macguffin in AOTC that is...extremely confusing ultimately. Because if you accept that Sidious has this grand master plan to use this war to destabilize the Republic and form an Empire...it's way too complex.

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.