Top 8 Learning Sources for Understanding Diffusion Models

NVIDIA's blog compares diffusion models to GANs, explaining their operation.

[{"selector":"#anim-2bc415ec-311f-46b0-8af7-b7b082f19cd4","keyframes":{"transform":["translate3d(-115.14196%, 0px, 0)","translate3d(0px, 0px, 0)"]},"delay":0,"duration":600,"easing":"cubic-bezier(0.4, 0.4, 0.0, 1)","fill":"both"}] [{"selector":"#anim-3e697574-f5bf-446e-aa16-46584453c43a","keyframes":{"opacity":[0,1]},"delay":0,"duration":600,"easing":"cubic-bezier(0.4, 0.4, 0.0, 1)","fill":"both"}] [{"selector":"#anim-577ecd66-ecf4-47ab-93ae-1438cb61acbb","keyframes":{"transform":["scale(0.15)","scale(1)"]},"delay":0,"duration":600,"easing":"cubic-bezier(0.4, 0.4, 0.0, 1)","fill":"forwards"}]

Assembly AI: Unveils DALL-E 2, guides Python-based diffusion model creation.

[{"selector":"#anim-ad09f515-75a4-49a6-9a24-f2fb4ad4b549","keyframes":{"transform":["translate3d(-114.8265%, 0px, 0)","translate3d(0px, 0px, 0)"]},"delay":0,"duration":600,"easing":"cubic-bezier(0.4, 0.4, 0.0, 1)","fill":"both"}] [{"selector":"#anim-5a38a63a-f3b2-4263-a08f-5839264555d8","keyframes":{"opacity":[0,1]},"delay":0,"duration":600,"easing":"cubic-bezier(0.4, 0.4, 0.0, 1)","fill":"both"}] [{"selector":"#anim-bca2d0c4-c192-4b74-a3f0-514cf8b78f90","keyframes":{"transform":["scale(0.15)","scale(1)"]},"delay":0,"duration":600,"easing":"cubic-bezier(0.4, 0.4, 0.0, 1)","fill":"forwards"}]

Yannic Kilcher's video dissects OpenAI's diffusion model research with examples.

[{"selector":"#anim-90404356-8410-4347-a030-a6fa0d542476","keyframes":{"transform":["translate3d(-114.8265%, 0px, 0)","translate3d(0px, 0px, 0)"]},"delay":0,"duration":600,"easing":"cubic-bezier(0.4, 0.4, 0.0, 1)","fill":"both"}] [{"selector":"#anim-9e86e2d8-ee4e-41d8-b312-16da188cf9b8","keyframes":{"opacity":[0,1]},"delay":0,"duration":600,"easing":"cubic-bezier(0.4, 0.4, 0.0, 1)","fill":"both"}] [{"selector":"#anim-14876aa1-0ec5-4f49-9b8b-16d62429b94f","keyframes":{"transform":["scale(0.15)","scale(1)"]},"delay":0,"duration":600,"easing":"cubic-bezier(0.4, 0.4, 0.0, 1)","fill":"forwards"}]

HuggingFace's course: Theory, training, pipelines for diffusion models - Python, PyTorch basics needed.

[{"selector":"#anim-432071f3-13b4-417f-8595-04ce5d6bf63e","keyframes":{"transform":["translate3d(-114.8265%, 0px, 0)","translate3d(0px, 0px, 0)"]},"delay":0,"duration":600,"easing":"cubic-bezier(0.4, 0.4, 0.0, 1)","fill":"both"}] [{"selector":"#anim-f23b9753-b862-43bf-b61c-8d9c8ac8afce","keyframes":{"opacity":[0,1]},"delay":0,"duration":600,"easing":"cubic-bezier(0.4, 0.4, 0.0, 1)","fill":"both"}] [{"selector":"#anim-a339d938-4644-4cf9-9a8e-0e1f1adfa980","keyframes":{"transform":["scale(0.15)","scale(1)"]},"delay":0,"duration":600,"easing":"cubic-bezier(0.4, 0.4, 0.0, 1)","fill":"forwards"}]

Lilian Weng's GitHub paper: Diffusion models vs. others, mathematical insight for ML enthusiasts.

[{"selector":"#anim-23a32e56-ca08-46ec-bea0-2be4d8f7d2e5","keyframes":{"transform":["translate3d(-114.8265%, 0px, 0)","translate3d(0px, 0px, 0)"]},"delay":0,"duration":600,"easing":"cubic-bezier(0.4, 0.4, 0.0, 1)","fill":"both"}] [{"selector":"#anim-070c99a2-2804-467a-929e-32a3131bff48","keyframes":{"opacity":[0,1]},"delay":0,"duration":600,"easing":"cubic-bezier(0.4, 0.4, 0.0, 1)","fill":"both"}] [{"selector":"#anim-31e73323-b6a1-4d22-896a-b66b2d93a00f","keyframes":{"transform":["scale(0.15)","scale(1)"]},"delay":0,"duration":600,"easing":"cubic-bezier(0.4, 0.4, 0.0, 1)","fill":"forwards"}]

Ayan Das' blog: Origins, use cases, math behind diffusion models, ties to score-based generative models.

[{"selector":"#anim-a32d2a2a-71ab-4f23-b1e2-0eedcb644757","keyframes":{"transform":["translate3d(-114.8265%, 0px, 0)","translate3d(0px, 0px, 0)"]},"delay":0,"duration":600,"easing":"cubic-bezier(0.4, 0.4, 0.0, 1)","fill":"both"}] [{"selector":"#anim-cc3d8524-98c3-403d-a82a-a620888f6e00","keyframes":{"opacity":[0,1]},"delay":0,"duration":600,"easing":"cubic-bezier(0.4, 0.4, 0.0, 1)","fill":"both"}] [{"selector":"#anim-5aca23ec-8f30-4f19-9559-87c0976ca8a0","keyframes":{"transform":["scale(0.15)","scale(1)"]},"delay":0,"duration":600,"easing":"cubic-bezier(0.4, 0.4, 0.0, 1)","fill":"forwards"}]

Yang Song's blog: Diffusion models' rise due to denoising, pivotal in generative AI evolution.

[{"selector":"#anim-f165c940-546d-442f-b76e-9ab0396d8fed","keyframes":{"transform":["translate3d(-114.8265%, 0px, 0)","translate3d(0px, 0px, 0)"]},"delay":0,"duration":600,"easing":"cubic-bezier(0.4, 0.4, 0.0, 1)","fill":"both"}] [{"selector":"#anim-03891e2f-2520-4b09-b5a5-364e10e748ed","keyframes":{"opacity":[0,1]},"delay":0,"duration":600,"easing":"cubic-bezier(0.4, 0.4, 0.0, 1)","fill":"both"}] [{"selector":"#anim-c5a258e2-5f7f-4012-b902-d3a1a18fb693","keyframes":{"transform":["scale(0.15)","scale(1)"]},"delay":0,"duration":600,"easing":"cubic-bezier(0.4, 0.4, 0.0, 1)","fill":"forwards"}]