Learn Before
Essay

Properties of Positional Cosine Vectors

A positional encoding method defines a 'cosine vector' for each position i using the formula: [cos(i*θ₁), ..., cos(i*θ_{d/2})], where θ is a fixed set of frequency parameters. Compare the expected similarity between the cosine vectors for two nearby positions (e.g., i=10 and j=11) versus two distant positions (e.g., i=10 and j=100). Justify your reasoning by explaining how the cosine function's properties and the position index i interact within the formula.

0

1

Updated 2025-10-08

Contributors are:

Who are from:

Tags

Ch.3 Prompting - Foundations of Large Language Models

Foundations of Large Language Models

Foundations of Large Language Models Course

Computing Sciences

Analysis in Bloom's Taxonomy

Cognitive Psychology

Psychology

Social Science

Empirical Science

Science