Machine Learning Exposed
summary
A hands‑on exposition of what a neural net actually computes: shows feed‑forward math (weights like 8.54, bias → sigmoid → 0.99), training by backprop with gradient descent, batching and epochs, and a real result of about 500–1000 iterations reducing cost to ~0.2 and producing predictions like 0.67 probability for an iris class. Demonstrates convolutional nets, autoencoders + t‑SNE on MNIST/art, and a working visualization app on GitHub built with DeepLearning4j, Angular 2 + Spring and live updates over WebSocket.
AI-written summary · about
chapters (27)
- 0:00Start
- 1:06Pivotal
- 1:49Machine Learning Exposed
- 2:40What Is Machine Learning
- 3:18Robots To Tidy the House
- 4:37Machine Learning Applications
- 5:07Three Types of Machine Learning
- 7:10Unsupervised Learning
- 7:43Reinforcement Learning
- 8:18The Alphago Project by Google Deepmind
- 8:47The Iris Data Set
- 10:39Dimensionality Reduction
- 13:15Modeling that in an Artificial Neural Network
- 13:48Deep Learning Network
- 17:04Calculations
- 19:50Sigmoid Activation
- 24:04Gradient Descent
- 25:01Epoch
- 26:41Tensor Flow Lab
- 27:34Optimizing a Neural Network
- 27:53Neural Networks
- 30:242d Visualization
- 33:16Google Ai Experiments
- 34:02Demo
- 36:13Euclidean Distance
- 39:25Training a Neural Network
- 40:47Ai Control Problem
related talks
51mMachine Learning Exposed Deep and Reinforcement Learning
51mMachine Learning Exposed Introduction to Machine Learning
49mMachine Learning Exposed: Introduction to Machine Learning
50mMachine Learning Exposed: The Fundamentals
28mDemystifying Deep Neural Nets
26mMachine Learning in the Browser with Deep Neural Networks
27mDemystifying Deep Neural Networks
35mDeconstructing Deep Learning