Nathan Strong @naughtyphoton - Twitter Profile

As a proud American company, we're thrilled to continue to make significant investments in the US. Today, we’re announcing a $500 billion commitment to support American innovation, advanced manufacturing, and high tech job creation. https://t.co/xoNy7zGqZG

5K

69K

8K

2K

23M

Nathan Strong @naughtyphoton

over 1 year ago

@threepointone I follow you and @IlkayGuendogan and I always have to check the @

0

1

0

47

Nathan Strong @naughtyphoton

over 1 year ago

@pants Love this one! I'd love to buy this one signed and frame it. I didn't see it in the shop, hope you make some prints :D

0

121

Nathan Strong @naughtyphoton

over 1 year ago

Joel, you are reading too much into the problem. You can easily solve this with a neural network. First, to encode all relevant possibilities, we’ll treat each child as a 14-dimensional one-hot vector: 7 days of the week × 2 possible genders = 14 combos. we’ll need binary encoding for each child’s (gender, day-of-week) for optimal deep. That’s ⌈log₂(14)⌉ = 4 bits, but we can double that to 8 bits just to be safe. We generate all possible pairs of children, each with an 8-bit encoding. Then label them with: • 1 if “both children are boys” • 0 otherwise But we only train on the subset that includes “at least one boy who is a Tuesday baby.” From there, we feed everything into a neural network with a hidden layer—64 hidden units (or 640, or 6400, if you really want to be sure). We train for 2,000 epochs, because 27 data points definitely call for thousands of epochs. Then we predict which pairs are “both boys” and check the ratio of those pairs to the total subset. If it’s not close to 13/27, we can always crank up the layers until the network memorizes every possible scenario. The answer is 0.4815 or 13/27.

naughtyphoton's tweet photo. Joel, you are reading too much into the problem. You can easily solve this with a neural network.

First, to encode all relevant possibilities, we’ll treat each child as a 14-dimensional one-hot vector:
7 days of the week × 2 possible genders = 14 combos.

we’ll need binary encoding for each child’s (gender, day-of-week) for optimal deep. That’s ⌈log₂(14)⌉ = 4 bits, but we can double that to 8 bits just to be safe.

We generate all possible pairs of children, each with an 8-bit encoding. Then label them with:
• 1 if “both children are boys”
• 0 otherwise

But we only train on the subset that includes “at least one boy who is a Tuesday baby.”

From there, we feed everything into a neural network with a hidden layer—64 hidden units (or 640, or 6400, if you really want to be sure). We train for 2,000 epochs, because 27 data points definitely call for thousands of epochs. Then we predict which pairs are “both boys” and check the ratio of those pairs to the total subset. If it’s not close to 13/27, we can always crank up the layers until the network memorizes every possible scenario.

The answer is 0.4815 or 13/27.

0

3

0

2

191

naughtyphoton retweeted

Zongyu (Johnson) Lin

@zy27962986

over 1 year ago

🚀🚀🚀Want to develop a cutting-edge video generation model towards Sora? Please dive into Apple’s latest recipe and studies for scalable video generation models🔥🔥🔥. In this work, we aim at providing a transparent and detailed recipe 📖 for model architecture, training strategy and data for scalable text-image conditioned video generation. Arxiv: https://t.co/VDZ2o3X2bz Hugging Face link: https://t.co/iL8QmdpRtA