Test Driven Neural Networks with Ruby
summary
Uses Ruby to build a feedforward neural network trained on letter frequency from the Bible (Matthew and Acts) to classify six languages, encoding 54 characters as 54 input nodes with roughly forty hidden units and six output bits. Tests enforce input vectors summing to one and catch odd utf‑8 spaces, training uses cross validation targeting 5% error rate, and a demo web app shows instant O(1) predictions from the learned weighted sums.
AI-written summary · about
chapters (23)
- 0:00Intro
- 0:17Gold Sponsors
- 1:49The Challenge
- 3:34Ruby is not complex Math
- 5:08Today we'll cover
- 6:47Input layer
- 7:00Hidden layer
- 7:21Output layer
- 8:02Digital Logic
- 8:27Fuzzy Logic
- 9:48Activation Functions
- 10:47Training Algorithms
- 11:18Visually What they do
- 12:36Just the tip of the iceberg
- 14:27Specifically
- 15:14Data Collection
- 16:49Character Distribution
- 18:00TDD Neural Nets
- 19:00The Model
- 19:31Test the Seams
- 21:02The Case for Cross Validation
- 22:46Ockham's Razor
- 27:54Conclusion
related talks
42mDōmo arigatō, Mr. Roboto: Machine Learning with Ruby
28mNeural Networks with RubyFANN
51mMachine Learning with Ruby
33mSyntax Isn't Everything: NLP for Rubyists
39mNatural Language Processing with Ruby
42mLearning Machine Learning
54mIntroduction to machine learning in Node.js
22mUsing Machine Learning to Improve UI/UX