I just interviewed the man who invented Amazon Prime.

He started a business worth $40 billion today and served on Amazon's board for 15 yrs.

Here are lessons from Bing Gordon that'll save you years of mistakes:

Sri Maha Bhairavar Rudhra alayam, Chengalpet, Tamilnadu

This temple is being constructed merely on the instructions from Bhairava by his devotee Shri Bhairava Siddhantham Swamy. The Temple is fully enshrinedin a transparent glass enclosure..1/n

with Rudrakshas on both the sides of the entrance This temple has a circular Sanctum Sanctorum which has twelve stepseach representing a sign of zodiac and after that twelve steps we find the supreme deity of this temple Sri Kshetra Paata Bhairava. 2/n

Beneath the Bhairava, in the lower level are installed the Ashta Bhairavas, SapthaRishis in the form of stones, and Saptha Kannis in the form of brass lamps. Thereare a number of little blue and white lights in the ceiling that is shaped like a dome.. 3/n

which give an ethereal atmosphere when switched on. The steps representthe twelve signs of the zodiac and when one steps over them to reach thesanctum sanctorum all the doshas pertaining to the horoscope are cleared onceand for all..4/n

The Sthala Vriksham ”Peepal Tree” which is grown on a pedestal, a trident which is considered as Bhairavi, Pancha Mukha Anjaneya, Prathyankara Devi, Subramanya, Rudra Vinayaga, Vaishnavi Devi, Naga Rudra, Eswara, Vana Durga etc. 5/5

#IncredibleBharat
#DekhoApnaDesh

Transform your body in next 6 months.

People will charge INR 27,000+ for this course.

I am giving away the entire course for FREE:

Module 1:
Build your Fat Loss Diet Plan

· Step by Step Tutorial
· Length: 26

Module 2:
Learn how to easily track your calories/macros.

· Step by Step Tutorial
· Length: 17

Module 3:
Calculate how many weeks it will take to reach six pack.

. Math behind building six pack
· Length: 12

Module 4:
Learn how to Burn Belly Fat in 3 steps

· Science behind belly fat
· 100% Works
· Length: 15

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.