梯度下降:類神經網路的學習機制
神經網路如何透過梯度下降法進行學習,並以辨識手寫數字的 MNIST 數據庫作為核心範例。網路透過調整上萬個權重與偏差值,旨在最小化衡量預測誤差的代價函數。這個過程就像球體滾下山谷尋找局部最小值,利用微積分計算出的梯度方向來優化效能。雖然初階網路未必能如人類般理解構圖邏輯,甚至可能只是記憶數據,但其展現出的數學規律奠定了現代人工智慧的基礎。作者強調,理解這種函數優化的本質,能讓看似科幻的機器學習轉化為清晰的微積分實作。

▶︎
But what is a neural network? | Deep learning chapter 1

▶︎
Gradient descent, how neural networks learn | Deep Learning Chapter 2

▶︎
Transformers, the tech behind LLMs | Deep Learning Chapter 5

▶︎
How might LLMs store facts | Deep Learning Chapter 7

▶︎
Harvard Professor: CS50, What Matters More Than Programming Now, Lecturing Well | David J Malan

▶︎
AlphaFold - The Most Useful Thing AI Has Ever Done

▶︎
Learn RAG From Scratch – Python AI Tutorial from a LangChain Engineer

▶︎
Attention in transformers, step-by-step | Deep Learning Chapter 6

▶︎
The Most Important Algorithm in Machine Learning

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

▶︎
How reading changes the way your brain works - BBC World Service

▶︎
全面解析“世界模型”:定义、路线、实践与AGI的更近一步
![Yann LeCun's $1B Bet Against LLMs [Part 1]](https://i.ytimg.com/vi/kYkIdXwW2AE/hqdefault.jpg?sqp=-oaymwEjCNACELwBSFryq4qpAxUIARUAAAAAGAElAADIQj0AgKJDeAE=&rs=AOn4CLDbV4izF3i-wxevCVIn7FJjoy1vlA)
▶︎
Yann LeCun's $1B Bet Against LLMs [Part 1]

▶︎
But what is a convolution?

▶︎
孫以瀚研究員 - 運用AI輔助撰寫研究計畫與論文:心態、技巧、學倫

▶︎
Why colliding blocks compute pi

▶︎
The FASTEST introduction to Reinforcement Learning on the internet

▶︎
AI教父承认自己当年预测错了 | 杰弗里·辛顿MIT讲座 | AI的危险 | 智能的两种范式 | 符号主义 | 幻觉问题 | 硅基智能 | 控制超级智能 | AI到底有没有意识 | 如何让AI更善良

▶︎
【蔡康永对谈水哥】智商碾压!见识中国最可怕的天才,这跟超能力有什么区别?#蔡康永#水哥#最强大脑#天才#智商测试#超能力#观察力#心理学#思维能力#窦文涛#周轶君#马未都#许子东#尹烨#圆桌派

▶︎
