Soran Ghaderi
@soranghaderi.bsky.social
Looking for a PhD position | Researching generalisation and reasoning 🧠💻 (EBMs) | MSc in AI
@uni_of_essex
GitHub: github.com/soran-ghaderi | TransformerX/torchEBM libraries
Website: https://soran-ghaderi.github.io/
@uni_of_essex
GitHub: github.com/soran-ghaderi | TransformerX/torchEBM libraries
Website: https://soran-ghaderi.github.io/
Reposted by Soran Ghaderi
New video on the details of diffusion models: youtu.be/iv-5mZ_9CPY
Produced by Welch Labs, this is the first in a short series of 3b1b this summer. I enjoyed providing editorial feedback throughout the last several months, and couldn't be happier with the result.
Produced by Welch Labs, this is the first in a short series of 3b1b this summer. I enjoyed providing editorial feedback throughout the last several months, and couldn't be happier with the result.
But how do AI videos actually work? | Guest video by @WelchLabsVideo
YouTube video by 3Blue1Brown
youtu.be
July 25, 2025 at 12:27 PM
New video on the details of diffusion models: youtu.be/iv-5mZ_9CPY
Produced by Welch Labs, this is the first in a short series of 3b1b this summer. I enjoyed providing editorial feedback throughout the last several months, and couldn't be happier with the result.
Produced by Welch Labs, this is the first in a short series of 3b1b this summer. I enjoyed providing editorial feedback throughout the last several months, and couldn't be happier with the result.
Reposted by Soran Ghaderi
AAAI'26 (abs): 29 days.
WACV'26 (R1,reg): 15 days.
WACV'26 (R1,paper): 22 days.
AAAI'26 (paper): 36 days.
3DV'26: 52 days.
WACV'26 (R2,reg): 78 days.
WACV'26 (R2,paper): 85 days.
ICLR'26 (abs): 85 days.
ICLR'26 (paper): 90 days.
WACV'26 (R1,reg): 15 days.
WACV'26 (R1,paper): 22 days.
AAAI'26 (paper): 36 days.
3DV'26: 52 days.
WACV'26 (R2,reg): 78 days.
WACV'26 (R2,paper): 85 days.
ICLR'26 (abs): 85 days.
ICLR'26 (paper): 90 days.
June 27, 2025 at 9:00 AM
AAAI'26 (abs): 29 days.
WACV'26 (R1,reg): 15 days.
WACV'26 (R1,paper): 22 days.
AAAI'26 (paper): 36 days.
3DV'26: 52 days.
WACV'26 (R2,reg): 78 days.
WACV'26 (R2,paper): 85 days.
ICLR'26 (abs): 85 days.
ICLR'26 (paper): 90 days.
WACV'26 (R1,reg): 15 days.
WACV'26 (R1,paper): 22 days.
AAAI'26 (paper): 36 days.
3DV'26: 52 days.
WACV'26 (R2,reg): 78 days.
WACV'26 (R2,paper): 85 days.
ICLR'26 (abs): 85 days.
ICLR'26 (paper): 90 days.
Reposted by Soran Ghaderi
I asked "on the other platform" what were the most important improvements to the original 2017 transformer.
That was quite popular and here is a synthesis of the responses:
That was quite popular and here is a synthesis of the responses:
April 28, 2025 at 6:47 AM
I asked "on the other platform" what were the most important improvements to the original 2017 transformer.
That was quite popular and here is a synthesis of the responses:
That was quite popular and here is a synthesis of the responses:
📌 The process of learning the energy landscape (potential field) of an EBM.
Trained using TorchEBM v0.3.0!
I will be uploading more experimental animations for different configs and datasets—Gaussian Mixture with 4 components.
Trained using TorchEBM v0.3.0!
I will be uploading more experimental animations for different configs and datasets—Gaussian Mixture with 4 components.
April 21, 2025 at 3:58 AM
📌 The process of learning the energy landscape (potential field) of an EBM.
Trained using TorchEBM v0.3.0!
I will be uploading more experimental animations for different configs and datasets—Gaussian Mixture with 4 components.
Trained using TorchEBM v0.3.0!
I will be uploading more experimental animations for different configs and datasets—Gaussian Mixture with 4 components.
Reposted by Soran Ghaderi
The NeurIPS Call for Workshops is now live. Proposals are due May 30 AoE, with acceptance notification on July 4 AoE. neurips.cc/Conferences/...
If you plan to submit a proposal for a workshop, please read our detailed guidance in our new blog post: blog.neurips.cc/2025/04/12/g...
If you plan to submit a proposal for a workshop, please read our detailed guidance in our new blog post: blog.neurips.cc/2025/04/12/g...
Call For Workshops 2025
Sat Dec 6 and Sun Dec 7, 2025
neurips.cc
April 13, 2025 at 7:52 PM
The NeurIPS Call for Workshops is now live. Proposals are due May 30 AoE, with acceptance notification on July 4 AoE. neurips.cc/Conferences/...
If you plan to submit a proposal for a workshop, please read our detailed guidance in our new blog post: blog.neurips.cc/2025/04/12/g...
If you plan to submit a proposal for a workshop, please read our detailed guidance in our new blog post: blog.neurips.cc/2025/04/12/g...
📌 Training an EBM on a Gaussian Mixture
Read this blog post on how to train an energy-based model (EBM) on a 2D Gaussian mixture distribution using the TorchEBM library. 👇
#EBM #OpenSource #DeepLearning #GenerativeAI
Read this blog post on how to train an energy-based model (EBM) on a 2D Gaussian mixture distribution using the TorchEBM library. 👇
#EBM #OpenSource #DeepLearning #GenerativeAI
April 13, 2025 at 9:13 PM
📌 Training an EBM on a Gaussian Mixture
Read this blog post on how to train an energy-based model (EBM) on a 2D Gaussian mixture distribution using the TorchEBM library. 👇
#EBM #OpenSource #DeepLearning #GenerativeAI
Read this blog post on how to train an energy-based model (EBM) on a 2D Gaussian mixture distribution using the TorchEBM library. 👇
#EBM #OpenSource #DeepLearning #GenerativeAI
Watch this mini energy-based model learning a Gaussian mixture distribution! ✨
📌 This is the first model trained using #TorchEBM
- The probability density (blue colors) learned by the EBM sharpens around the white circles.
- Red dots get closer to the GM dist.
Link 👇
#EBM #GenerativeAI
📌 This is the first model trained using #TorchEBM
- The probability density (blue colors) learned by the EBM sharpens around the white circles.
- Red dots get closer to the GM dist.
Link 👇
#EBM #GenerativeAI
April 13, 2025 at 9:11 PM
Watch this mini energy-based model learning a Gaussian mixture distribution! ✨
📌 This is the first model trained using #TorchEBM
- The probability density (blue colors) learned by the EBM sharpens around the white circles.
- Red dots get closer to the GM dist.
Link 👇
#EBM #GenerativeAI
📌 This is the first model trained using #TorchEBM
- The probability density (blue colors) learned by the EBM sharpens around the white circles.
- Red dots get closer to the GM dist.
Link 👇
#EBM #GenerativeAI
The new version of the 🟢 #TorechEBM docs is now out. There are some tutorials on #EBMs there. I'm thinking about adding support for #diffusion- and #flow-based methods.
📌Collaboration's welcome!
Link: ...
📌Collaboration's welcome!
Link: ...
April 6, 2025 at 2:17 PM
The new version of the 🟢 #TorechEBM docs is now out. There are some tutorials on #EBMs there. I'm thinking about adding support for #diffusion- and #flow-based methods.
📌Collaboration's welcome!
Link: ...
📌Collaboration's welcome!
Link: ...
My new post on Hamiltonian mechanics and HMC. Breaks down the math to with bits of visualization.
How can Hamiltonian mechanics apply to generative modeling and Markov chain Monte Carlo? Soran Ghaderi presents a thoughtful deep dive on a fascinating topic.
From Physics to Probability: Hamiltonian Mechanics for Generative Modeling and MCMC | Towards Data Science
Hamiltonian mechanics is a way to describe how physical systems, like planets or pendulums, move over time, focusing on energy rather than just forces. By reframing complex dynamics through energy…
towardsdatascience.com
March 29, 2025 at 5:41 PM
My new post on Hamiltonian mechanics and HMC. Breaks down the math to with bits of visualization.
📌 Different quadratic forms in two variables show how different covariance matrices influence the shape of these forms.
I'll publish an intro to Hamiltonian mechanics and Monte Carlo on the #TorchEBM lib blog soon. #Hamiltonian #Sampling #EnergyBasedModels #MCMC #GenerativeAI
I'll publish an intro to Hamiltonian mechanics and Monte Carlo on the #TorchEBM lib blog soon. #Hamiltonian #Sampling #EnergyBasedModels #MCMC #GenerativeAI
March 15, 2025 at 9:51 PM
📌 Different quadratic forms in two variables show how different covariance matrices influence the shape of these forms.
I'll publish an intro to Hamiltonian mechanics and Monte Carlo on the #TorchEBM lib blog soon. #Hamiltonian #Sampling #EnergyBasedModels #MCMC #GenerativeAI
I'll publish an intro to Hamiltonian mechanics and Monte Carlo on the #TorchEBM lib blog soon. #Hamiltonian #Sampling #EnergyBasedModels #MCMC #GenerativeAI
Symmetries write particles into existence.
March 15, 2025 at 3:08 AM
Symmetries write particles into existence.
The phase space is a mathematical space where we can represent a physical system's set of possible states.
March 14, 2025 at 11:58 PM
The phase space is a mathematical space where we can represent a physical system's set of possible states.
Reposted by Soran Ghaderi
I think the “hammer in search of a nail” style of science is perpetually underrated. From Leeuwenhoek’s lens to DeepMind’s AlphaFold, powerful tools chasing problems spark revolutions. If your hammer is Mjölnir, why not swing for the grandest problem?
January 15, 2025 at 2:53 PM
I think the “hammer in search of a nail” style of science is perpetually underrated. From Leeuwenhoek’s lens to DeepMind’s AlphaFold, powerful tools chasing problems spark revolutions. If your hammer is Mjölnir, why not swing for the grandest problem?
Reposted by Soran Ghaderi
Anne Gagneux, Ségolène Martin, @quentinbertrand.bsky.social Remi Emonet and I wrote a tutorial blog post on flow matching: dl.heeere.com/conditional-... with lots of illustrations and intuition!
We got this idea after their cool work on improving Plug and Play with FM: arxiv.org/abs/2410.02423
We got this idea after their cool work on improving Plug and Play with FM: arxiv.org/abs/2410.02423
November 27, 2024 at 9:00 AM
Anne Gagneux, Ségolène Martin, @quentinbertrand.bsky.social Remi Emonet and I wrote a tutorial blog post on flow matching: dl.heeere.com/conditional-... with lots of illustrations and intuition!
We got this idea after their cool work on improving Plug and Play with FM: arxiv.org/abs/2410.02423
We got this idea after their cool work on improving Plug and Play with FM: arxiv.org/abs/2410.02423
Reposted by Soran Ghaderi
Hello, world! We are a community-developed toolkit that performs Bayesian inference for simulators. We support a broad range of methods (NPE, NLE, NRE, amortized and sequential), neural network architectures (flows, diffusion models), samplers, and diagnostics. Join us!
November 18, 2024 at 7:26 AM
Hello, world! We are a community-developed toolkit that performs Bayesian inference for simulators. We support a broad range of methods (NPE, NLE, NRE, amortized and sequential), neural network architectures (flows, diffusion models), samplers, and diagnostics. Join us!
Reposted by Soran Ghaderi
Tomorrow (Monday) Akhil Premkumar will present his "Neural Entropy" arxiv.org/abs/2409.03817
Join us on zoom at 9am PT / 12pm ET / 6pm CET: portal.valencelabs.com/logg
Join us on zoom at 9am PT / 12pm ET / 6pm CET: portal.valencelabs.com/logg
November 24, 2024 at 8:27 PM
Tomorrow (Monday) Akhil Premkumar will present his "Neural Entropy" arxiv.org/abs/2409.03817
Join us on zoom at 9am PT / 12pm ET / 6pm CET: portal.valencelabs.com/logg
Join us on zoom at 9am PT / 12pm ET / 6pm CET: portal.valencelabs.com/logg
Reposted by Soran Ghaderi
SIGGRAPH'25 (form): 53 days.
RSS'25 (abs): 54 days.
SIGGRAPH'25 (paper-md5): 60 days.
RSS'25 (paper): 61 days.
ICML'25: 67 days.
ICCV'25: 102 days.
RSS'25 (abs): 54 days.
SIGGRAPH'25 (paper-md5): 60 days.
RSS'25 (paper): 61 days.
ICML'25: 67 days.
ICCV'25: 102 days.
November 24, 2024 at 10:00 AM
SIGGRAPH'25 (form): 53 days.
RSS'25 (abs): 54 days.
SIGGRAPH'25 (paper-md5): 60 days.
RSS'25 (paper): 61 days.
ICML'25: 67 days.
ICCV'25: 102 days.
RSS'25 (abs): 54 days.
SIGGRAPH'25 (paper-md5): 60 days.
RSS'25 (paper): 61 days.
ICML'25: 67 days.
ICCV'25: 102 days.
"Hello Bluesky"
November 24, 2024 at 11:34 PM
"Hello Bluesky"