Machine Learning with Node.js
summary
Builds a probabilistic football match predictor in Node.js using synaptic and a perceptron-style network (2 inputs, two hidden layers of six nodes, 3-way output). Trains with gradient descent and a chosen learning rate for ~100,000 iterations and logs mean squared (data) error ≈ 18% while the classification error ≈ 45% and cross‑validation error is higher, indicating overfit; adding more features barely helps and predictions can be worse than a coin toss.
AI-written summary · about
chapters (17)
- 0:00Start
- 1:49Understand Neural Networks
- 5:30Activation Function
- 5:59Output Layer
- 7:59Supervised Learning
- 8:40Unsupervised Learning
- 10:56Gradient Descent
- 13:02Learning Rate
- 15:49Perceptron
- 17:01Predictions
- 18:26Error Rates
- 18:41Synaptic Data Error
- 19:58The Classification Error
- 20:28Predict from Probability
- 21:45The Cross-Validation Error
- 23:11The Mean Squared Error
- 26:09Regularization
related talks
30mApplying The Magic Of Neural Networks
42mLearning Machine Learning
54mIntroduction to machine learning in Node.js
01h 05mIntro to machine learning in Node- 42mLearning Machine Learning
23mReal Time Machine Learning With Node js
22mUsing Machine Learning to Improve UI/UX
26mMachine Learning in the Browser with Deep Neural Networks