【生成モデル】拡散モデル - 非平衡統計力学にアイデアを得た、ノイズ除去の生成モデル!【Diffusion Models】GM vol. 12 #232 #VRアカデミア

Modern generative models are mostly based on diffusion models! The theory behind them is deep, interesting, and difficult, but... The most important thing to understand for mastering them is that "it's ultimately just noise-reducing least squares"! Let's grasp this first! We'll cover the deep (and crazy) calculations next time, so please watch this video carefully at least once! [Playlist]    • 生成モデルに潜入   [References] Original Paper [1503.03585] Deep Unsupervised Learning using Nonequilibrium Thermodynamics https://arxiv.org/abs/1503.03585 To be honest, it's a bit difficult. But it has the feel of the original paper, which is great! For a standard learning approach, I would recommend: Okanohara's "Diffusion Models" https://amzn.to/3RHM6Gd My "Introduction to Calculus + Linear Algebra" https://amzn.to/42O8LDj [Table of Contents] 00:00 Opening 02:01 Introduction to the Paper 03:14 Historical Context 06:07 Results at the Time 10:54 Diffusion Process 16:45 Model Overview 24:35 Initial Ideas 29:05 Proliferation of Learning Logics 32:26 Why Does It Work? 34:29 How to Use It Effectively 36:54 Ending [Advertisement: Please Buy My Books!] I've written in-depth articles covering all areas of analytical models frequently used in data science! An Introduction to Analytical Models for Data Analysis that Captures the Essence https://amzn.to/3Ng0nC7 This book explains the mathematics necessary for data analysis, from the basics to the cutting edge, comprehensively and thoroughly, including the underlying principles of the formulas! An Introduction to Calculus and Linear Algebra for Uncompromising Data Analysis https://amzn.to/3B4y8G4 [Advertisement 2: We're running a crowdfunding campaign!] We're asking for your support to continue our activities. Details here! Crowdfunding (monthly) here → https://community.camp-fire.jp/projec... (Video explaining why we started crowdfunding →    • 【クラウドファンディング】始めるよ!【ご支援よろしくお願いします】#167 #VRア...  ) [In Closing] Thank you for watching! If you enjoyed it, please give it a thumbs up and subscribe to the channel. For questions and comments about the video, please leave a comment or tweet! For work inquiries, collaboration requests, and internship applications, please contact us via our official website or Twitter DM. AIcia Solid Project - Official Website - https://sites.google.com/view/aicia-o... Video Generation: AIcia Solid (Twitter:   / aicia_solid  ) Video Editing: Mai. (Twitter: https://x.com/nnql_nnql2pnt) ======= Logo: ChikakoHorio (   / chikakohorio   ) Model Design by: Bancho (   / k_ban_   ) Mama! 3D Model by: Kitsunetsuki (   / _kitsune_tsuki_   ) Papa!

【深層学習】Transformer - Multi-Head Attentionを理解してやろうじゃないの【ディープラーニングの世界vol.28】#106 #VRアカデミア #DeepLearning
▶︎

【深層学習】Transformer - Multi-Head Attentionを理解してやろうじゃないの【ディープラーニングの世界vol.28】#106 #VRアカデミア #DeepLearning

オードリー・タン × 東浩紀「プルラリティ・AI・民主主義 —喧騒と幻想を巡って— 」【日本財団HUMAIプログラム Spring Camp 2026】
▶︎

オードリー・タン × 東浩紀「プルラリティ・AI・民主主義 —喧騒と幻想を巡って— 」【日本財団HUMAIプログラム Spring Camp 2026】

[Generative Model] Boltzmann Machine - Overview - Embed information in the energy function! [Expl...
▶︎

[Generative Model] Boltzmann Machine - Overview - Embed information in the energy function! [Expl...

【宇宙の秘密】なぜ「素数」というたった一つの数学の問題が、万物を支配するのか? リーマン予想の真実
▶︎

【宇宙の秘密】なぜ「素数」というたった一つの数学の問題が、万物を支配するのか? リーマン予想の真実

The Power of a Single Neuron and a Path to Simulating the Brain | Dr. Konrad Kording
▶︎

The Power of a Single Neuron and a Path to Simulating the Brain | Dr. Konrad Kording

Billionaire's WARNING: I'm SELLING. The Crash Is Already Here!
▶︎

Billionaire's WARNING: I'm SELLING. The Crash Is Already Here!

bounce + bounce = no bounce
▶︎

bounce + bounce = no bounce

【ミュトスが止まっても「Fugu」は動く】Sakana AI CEOデイビッド・ハ「AIのアベンジャーズを作った」/AIの自己進化が始まる/「sakana.ai」2000円で購入【1on1 Tech】
▶︎

【ミュトスが止まっても「Fugu」は動く】Sakana AI CEOデイビッド・ハ「AIのアベンジャーズを作った」/AIの自己進化が始まる/「sakana.ai」2000円で購入【1on1 Tech】

You’ll stop using ChatGPT after listening to this | Jonathan Pageau [ARC 2026]
▶︎

You’ll stop using ChatGPT after listening to this | Jonathan Pageau [ARC 2026]

[Reinforcement Learning] Double DQN - Addressing Overestimation Bias and Stabilizing Learning! [D...
▶︎

[Reinforcement Learning] Double DQN - Addressing Overestimation Bias and Stabilizing Learning! [D...

But how do AI images and videos actually work? | Guest video by Welch Labs
▶︎

But how do AI images and videos actually work? | Guest video by Welch Labs

【フルスピーチ】2026年ハーバード大学 コナンの卒業式スピーチ『肩書を捨てる勇気』【世界情勢理解】【英語スピーチ】 リスニング 日本語字幕
▶︎

【フルスピーチ】2026年ハーバード大学 コナンの卒業式スピーチ『肩書を捨てる勇気』【世界情勢理解】【英語スピーチ】 リスニング 日本語字幕

[Official] Sandwichman Comedy [Beehive Removal] 2016
▶︎

[Official] Sandwichman Comedy [Beehive Removal] 2016

Yann LeCun: World Models: Enabling the next AI revolution
▶︎

Yann LeCun: World Models: Enabling the next AI revolution

Linus Torvalds Just EXPOSED Microsoft's Biggest Problem Yet
▶︎

Linus Torvalds Just EXPOSED Microsoft's Biggest Problem Yet

Yann LeCun's $1B Bet Against LLMs [Part 1]
▶︎

Yann LeCun's $1B Bet Against LLMs [Part 1]

Trumps Pool-Fiasko | Peinliche Videos von Politikern zur WM | extra 3 vom 25.06.2026 · Teil 2/2
▶︎

Trumps Pool-Fiasko | Peinliche Videos von Politikern zur WM | extra 3 vom 25.06.2026 · Teil 2/2

Bertrand's Paradox (with 3blue1brown) - Numberphile
▶︎

Bertrand's Paradox (with 3blue1brown) - Numberphile

We're 99.9% sure this pattern is true, but no one can prove it
▶︎

We're 99.9% sure this pattern is true, but no one can prove it

5 AI Agent Terms You Need to Know
▶︎

5 AI Agent Terms You Need to Know