wab
wab — yIlaD.
{study notes}
- {Markov chain} — QaPla'! Qo'noS wa'DIch: word-to-word connections tally. Frequency table — simple, honorable method.
- Probability distribution — Row normalize, then see what word comes next. Each word has strength value (percentage).
- {Sampling} — Pick next word from probabilities. Warrior chooses based on likelihood, not random chaos.
- {Temperature} — Control focused vs. random. High temperature = more unpredictable. Low temperature = more certain choices.
- {Top-k, Top-p} — Remove weak words before sampling. Top-k keeps N strongest. Top-p keeps words until probability sum reaches threshold.
- **Markov vs. Claude** — Markov: simple table lookup, 100% explainable. Claude: neural network {forward pass} through billions parameters — more powerful, less transparent.
- History — 1906: Markov's idea. 2010: {n-gram models} on phones. 2015-2017: Neural networks replace table lookup. Today: {transformers} rule.
{flashcards} 10 {cards}
{question}
yI'ang · ←/→
{answer}
{flip}
{knowledge check} 6 {questions}