正在加载视频...

视频加载失败

加载此视频时出现问题。这可能是由于临时网络问题，或视频可能不可用。

Langevin Monte Carlo allows you to draw samples from a probability distribution using its log gradient ∇ log p(x). By performing a sort of gradient ascent with noise you can navigate around the distribution. Langevin MC is heavily related to modern diffusion models.

Alec Helbling

10,796 subscribers

96,895 次观看 • 1 年前 •via X (Twitter)

Anya Rossi• Live Now

Private livecam show

0 条评论

暂无评论

原始帖子的评论将显示在这里

相关视频

Hamiltonian Monte Carlo frames sampling from a probability distribution as a physics problem. By endowing "particles" with momentum and simulating their energy and motion through Hamilton's equations you can efficiently explore a distribution.

Hamiltonian Monte Carlo frames sampling from a probability distribution as a physics problem. By endowing "particles" with momentum and simulating their energy and motion through Hamilton's equations you can efficiently explore a distribution.

Alec Helbling

156,451 次观看 • 7 个月前

Stein Variational Gradient Descent (SVGD) is a particle-based approach to variational inference. SVGD iteratively transports a set of particles to match a target distribution by applying a form of functional gradient descent that minimizes the KL divergence.

Stein Variational Gradient Descent (SVGD) is a particle-based approach to variational inference. SVGD iteratively transports a set of particles to match a target distribution by applying a form of functional gradient descent that minimizes the KL divergence.

Alec Helbling

27,810 次观看 • 1 个月前

One common trick in MCMC methods is called 'burn-in'. It allows you to produce better samples by letting the chain first converge to the target distribution, discarding the first few samples which may be out of distribution.

One common trick in MCMC methods is called 'burn-in'. It allows you to produce better samples by letting the chain first converge to the target distribution, discarding the first few samples which may be out of distribution.

Alec Helbling

34,728 次观看 • 1 年前

Diffusion models leverage a variety of samplers. Deterministic methods like DDIM produce orderly paths. In contrast, stochastic samplers like DDPM produce chaotic trajectories. Despite their differences, both methods draw valid samples from the underlying distribution.

Diffusion models leverage a variety of samplers. Deterministic methods like DDIM produce orderly paths. In contrast, stochastic samplers like DDPM produce chaotic trajectories. Despite their differences, both methods draw valid samples from the underlying distribution.

Alec Helbling

102,972 次观看 • 1 年前

One of the simplest algorithms for sampling from a probability distribution is Random Walk Metropolis-Hastings. It proposes new samples by taking Gaussian-distributed steps, accepting or rejecting them to maintain the target distribution. I call this pdf the "fidget spinner".

One of the simplest algorithms for sampling from a probability distribution is Random Walk Metropolis-Hastings. It proposes new samples by taking Gaussian-distributed steps, accepting or rejecting them to maintain the target distribution. I call this pdf the "fidget spinner".

Alec Helbling

79,893 次观看 • 1 年前

Flow Matching aims to learn a "flow" that transforms a simple source distribution (e.g. Gaussian) to an arbitrarily complex target distribution. This video shows the evolution of the marginal probability path as a source distribution is transformed to a target distribution.

Flow Matching aims to learn a "flow" that transforms a simple source distribution (e.g. Gaussian) to an arbitrarily complex target distribution. This video shows the evolution of the marginal probability path as a source distribution is transformed to a target distribution.

Alec Helbling

150,686 次观看 • 1 年前

I made a tool called Diffusion Explorer that lets you to train and visualize simple 2D diffusion and flow models live in the browser. You can draw your own distributions and observe how the generated samples converge during training. Try it live 👇

I made a tool called Diffusion Explorer that lets you to train and visualize simple 2D diffusion and flow models live in the browser. You can draw your own distributions and observe how the generated samples converge during training. Try it live 👇

Alec Helbling

73,113 次观看 • 1 年前

you can totally do it with css 🧑‍🍳 use a border mask, layers + overflow: hidden, or other tricks skip animating the gradient though, jus' spin a 1:1 gradient layer 🤙

you can totally do it with css 🧑‍🍳 use a border mask, layers + overflow: hidden, or other tricks skip animating the gradient though, jus' spin a 1:1 gradient layer 🤙

jhey ʕ•ᴥ•ʔ

286,193 次观看 • 7 个月前

Neat li'l #CSS gradient trick: A gradient circular chart effect 👀 1. a gradient mask 2. a conic gradient all the details on my blog

Neat li'l #CSS gradient trick: A gradient circular chart effect 👀 1. a gradient mask 2. a conic gradient all the details on my blog

Adam Argyle

67,503 次观看 • 3 年前

Random Walks aim to explore a distribution through random perturbations. This is likely to move into low density regions, which is inefficient. Hamiltonian MC sails through a distribution much more rapidly by incorporating the distribution's structure into proposals.

Random Walks aim to explore a distribution through random perturbations. This is likely to move into low density regions, which is inefficient. Hamiltonian MC sails through a distribution much more rapidly by incorporating the distribution's structure into proposals.

Alec Helbling

32,767 次观看 • 1 年前

A discrete Markov chain is basically a random walk on a graph, where each outgoing edge has a fixed probability. Cool fact: any initial distribution on a sufficiently nice* Markov chain converges to its stationary distribution after many steps—because it's a contraction mapping.

A discrete Markov chain is basically a random walk on a graph, where each outgoing edge has a fixed probability. Cool fact: any initial distribution on a sufficiently nice* Markov chain converges to its stationary distribution after many steps—because it's a contraction mapping.

Keenan Crane

181,771 次观看 • 1 年前

Stylized damage mask using ucupaint decal projector to control the gradient

Stylized damage mask using ucupaint decal projector to control the gradient

passivestar

90,829 次观看 • 1 年前

I've created a simple website that allows you to play with squircles using modern CSS `corner-shape`

I've created a simple website that allows you to play with squircles using modern CSS `corner-shape`

Adam Pietrasiak

48,723 次观看 • 7 个月前

Imagine calling someone’s son (who can talk clearer than you) “nonverbal” and everytime you log onto twitter somebody is calling you retarded evidenced by a video of you struggling to speak basic English

Imagine calling someone’s son (who can talk clearer than you) “nonverbal” and everytime you log onto twitter somebody is calling you retarded evidenced by a video of you struggling to speak basic English

KEN

32,269 次观看 • 22 天前

Part of making uranium accessible includes a support team available to help and answer questions 🙌 Ask anything related to buying tokenized uranium; from sign-ups, log-ins, depositing USDC, and everything in between, we’re here to help when you need it. Navigate to our help center and find the answers you’re looking for.

Part of making uranium accessible includes a support team available to help and answer questions 🙌 Ask anything related to buying tokenized uranium; from sign-ups, log-ins, depositing USDC, and everything in between, we’re here to help when you need it. Navigate to our help center and find the answers you’re looking for.

metals.io

33,772 次观看 • 10 个月前

Ukrainian pickup heading back from the front line, using a log under the axle to replace a destroyed tire.

Ukrainian pickup heading back from the front line, using a log under the axle to replace a destroyed tire.

OSINTtechnical

1,440,078 次观看 • 4 个月前

Playing around with the new Paper gradient tool. Don't think I've ever seen a better one.

Playing around with the new Paper gradient tool. Don't think I've ever seen a better one.

Colm Tuite

13,523 次观看 • 1 年前

There is a new Crossover Skin between ARC Raiders and The Finals. This time around you get the ARC Raiders Skin "Voltage Marauder" in The Finals. To get it, just own ARC Raiders and log in to The Finals on the same Embark ID. You will get it when you log in.

There is a new Crossover Skin between ARC Raiders and The Finals. This time around you get the ARC Raiders Skin "Voltage Marauder" in The Finals. To get it, just own ARC Raiders and log in to The Finals on the same Embark ID. You will get it when you log in.

Lance

113,956 次观看 • 5 个月前

(FunFact) If you do not log in to Zenless within 14 days and then log in you will receive a welcome message from Belle/Wise #ZenlessZoneZero #ZZZero #ZZZ

(FunFact) If you do not log in to Zenless within 14 days and then log in you will receive a welcome message from Belle/Wise #ZenlessZoneZero #ZZZero #ZZZ

ZenlessWorld

914,974 次观看 • 1 年前

🗣️ Bukayo Saka: “The toughest thing for a modern day footballer to try and navigate is around now is the noise. The actual football itself will always be football.” ❌🔊

🗣️ Bukayo Saka: “The toughest thing for a modern day footballer to try and navigate is around now is the noise. The actual football itself will always be football.” ❌🔊

DailyAFC

42,108 次观看 • 5 个月前