symproj 0.1.1

Symbolic projection and embeddings
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40

# symproj

[![crates.io](https://img.shields.io/crates/v/symproj.svg)](https://crates.io/crates/symproj)
[![Documentation](https://docs.rs/symproj/badge.svg)](https://docs.rs/symproj)
[![CI](https://github.com/arclabs561/symproj/actions/workflows/ci.yml/badge.svg)](https://github.com/arclabs561/symproj/actions/workflows/ci.yml)

Symbolic projection (embeddings).
Maps discrete symbols to continuous vectors using a Codebook and pooling.

## Intuition

Imagine a library where every book has a call number. The call number isn't just a label; it tells you where the book sits in a 3D space. `symproj` is the system that maps "book names" (tokens) to "library coordinates" (vectors).

## Usage

```rust
use symproj::{Codebook, Projection};
use textprep::BpeTokenizer;

// 1. Load a Codebook (matrix + dimension)
let matrix = vec![...]; // flattened [vocab_size * dim]
let codebook = Codebook::new(matrix, 384).unwrap();

// 2. Create a Projection (tokenizer + codebook)
let tokenizer = BpeTokenizer::from_file("tokenizer.json")?;
let proj = Projection::new(tokenizer, codebook);

// 3. Encode text -> vector (mean pooling)
let vec = proj.encode("Hello world").unwrap();
```

## Features

- **Codebook**: Dense embedding matrix lookup.
- **Pooling**: Mean, weighted mean (SIF), and sequence output.
- **Normalization**: L2 normalization and component removal (PCA-based denoising).

## License

MIT OR Apache-2.0