12 releases (5 breaking)
0.7.0 | Jun 25, 2024 |
---|---|
0.6.0 | Feb 7, 2024 |
0.5.0 | Jan 23, 2024 |
0.3.0 | Nov 21, 2023 |
0.1.4 | Jul 26, 2023 |
#867 in Machine learning
Used in 4 crates
(3 directly)
350KB
9K
SLoC
Lace codebook
Contains the lace codebook specification as well as utilities for generating defaults.
If you design a new type, implement FromStr
in lace_utils
, and decide its
precident for the codebook in this crate.
lib.rs
:
The Codebook
is a YAML file used to associate metadata with the dataset.
The user can set the priors on the structure of each state, can identify
the model for each columns, and set hyper priors.
Often the data has too many columns to write a codebook manually, so there are functions to guess at a default codebook given a dataset. The user can then edit the default file.
Example
An Example codebook for a two-column dataset.
use indoc::indoc;
let codebook_str = indoc!("
---
table_name: two column dataset
state_prior_process:
!dirichlet
alpha_prior:
shape: 1.0
rate: 1.0
view_prior_process:
!pitman_yor
alpha_prior:
shape: 1.0
rate: 1.0
d_prior:
alpha: 1.0
beta: 2.0
col_metadata:
- name: col_1
notes: first column with all fields filled in
coltype:
!Categorical
k: 3
hyper:
pr_alpha:
shape: 1.0
scale: 1.0
prior:
k: 3
alpha: 0.5
value_map: !string
0: red
1: green
2: blue
- name: col_2
notes: A binary column with optional fields left out
coltype:
!Categorical
k: 2
value_map: !u8 2
comments: An example codebook
row_names:
- A
- B
- C
- D
- E");
let codebook: Codebook = serde_yaml::from_str(&codebook_str).unwrap();
assert_eq!(codebook.col_metadata.len(), 2);
Dependencies
~29–61MB
~1M SLoC