The New York Times article about SlateStarCodex is finally out and it is...bad. There's a lot I could parse, but let me just walk you through one paragraph that is so misleading as to be

1/ Productivity tip: when I keep procrastinating on something I often trick myself by saying, "ok just work on it for the next 20 minutes" or some short period of time, "and then you can chill/relax guilt free" and give myself some treat (sugar, netflix, etc)

2/ Usually at the 20 minute mark (sometimes I set a timer) I'm so engrossed in what I'm doing that I don't want want to stop and end up working on it for the next hour or two. But getting started seems to be the hardest part.

3/ Key to this: you have to actually be ok with stopping after 20 minutes and being guilt free if that's how you feel. So it's not a trick, i have the option every time, I just often don't want to use it once i'm in the zone.

Did you know?

Rani Bhabani was the Queen of Natore, Bengal or 'Natorer Rani' as she was famously called. She made great contributions in Tarapith & renovated Benaras.

In Varanasi, Durga Kund Mandir was constructed by Rani Bhabani in the 18th century.

Born in Chhatimgram village, Bogra District,Her father's name was Atmaram Choudhury.Bhabani was married off to Raja Ramkanta Moitra,the then zamindar of Rajshahi.After his death, Bhabani became the zamindar, and started being referred to as Rani.She was known as Ardhabangeshwari.

A woman as a zamindar was extremely rare in those days, but Rani Bhabani managed the vast Rajshahi zamindari most efficiently and effectively for over four decades.
Maharani Bhavani was very pious in her personal life. She passed her life in a strict disciplined way.

She conducted the administrative work skillfully. As a Maharani she was an impartial judge. Rani Bhavani had a very hard time during the fag end of her life. The Zamindari (estate) incurred serious loss due to the change of the revenue policy of the company.

In spite of being the owner of a large Zamindari (estate), she had to depend on the stipend of the government in the later part of her life. The amount of this stipend gradually decreased and became Rs. 1000/- only in the end.

SRI SWETHA VINAYAGAR, THIRUVALANCHUZHI, KUMBAKONAM
(TN)
#bharatmandir

While main deity is Shiva as Kapartheeswarar, the kshetram is famous for Ganesha, known as Swetha/Vellai Vinayagar (white Vinayak)

During samudra manthan, Vasuki snake (used as rope) started spewing venom.

Unable to bear the intense heat, Devas rushed to Siva. He told them that this was because the task was started without getting Ganesha’s approval.
So the Devas created a Ganesha murti from white sea foam and worshipped it.
It was later taken by Indra to Indralok.
@GunduHuDuGa

But once Indra had to come to Earth to pray to Shiva to seek relief from a curse inflicted on him. So he brought the murti of Ganesha with him. When he came here, he wanted to take bath before puja. So he gave the murti to a boy to hold for sometime. The boy was actually Shiva.

He kept the murti down and vanished. Indra could not lift it again and a divine voice told him to let Vinayagar stay there and visit every Ganesh Chaturthi to pray.
Since Vinayagar is made of foam, no abhishekam is done here. Only punugu is applied n camphor powder is sprinkled

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.

Okay look - I didn’t grow up with comic books. But I read stacks of science fiction digests and I kit-bashed model cars into fantasy rods, and just tried to maintain a low profile until I could get out of the house. So you could say I’m a distant cousin of the comic book fan. /1

And sometimes one’s childhood passion follows them into adulthood, sometimes not, sometimes as an adult one finds a passion they never had as a child and embraces it. For me, that must include getting involved with a puppet show at the age of thirty. /2

So who’s to say whose passion’s legitimate and whose is bogus? Nobody. Someone might indict your passion but they have no ground to stand on and should be dismissed out of hand. /3

Maybe they’re envious that you have a passion, maybe they think that their passion is cooler than yours, or maybe they’re just insecure, who knows. Doesn’t matter. /4

What matters is that if something gives you joy, and in the process doesn’t hurt others, then by all means follow it. It’s how we explore our secret hopes, our shadow sides, our inner character. /5