Struct Optimizer 

pub struct Optimizer { /* private fields */ }

Active optimization state for UMAP embedding.

This contains all the state needed to run and resume stochastic gradient descent optimization. It’s large and mutable, meant to be used during training and then converted to a lightweight FittedUmap when done.

The optimizer can be serialized mid-training to enable fault-tolerant training with checkpoints.

Implementations

impl Optimizer

pub fn new( manifold: LearnedManifold, init: Array2<f32>, total_epochs: usize, opt_params: &UmapConfig, metric_type: MetricType, ) -> Self

Create a new optimizer from a learned manifold.

This performs preprocessing:

• Filters weak edges from the graph
• Extracts head/tail edge lists
• Computes epoch sampling schedules
• Normalizes the initial embedding to [0, 10]

Arguments

• manifold - The learned manifold structure
• init - Initial embedding (will be normalized)
• total_epochs - Total number of epochs to run
• opt_params - Optimization parameters (learning rate, negative sampling, etc.)
• metric_type - Type of distance metric being used

pub fn step_epochs(&mut self, n: usize, output_metric: &dyn Metric)

Run n more epochs of stochastic gradient descent.

Panics

Panics if this would exceed total_epochs. Check remaining_epochs() first.

pub fn current_epoch(&self) -> usize

Get the current epoch number.

pub fn total_epochs(&self) -> usize

Get the total epochs this optimizer is configured for.

pub fn remaining_epochs(&self) -> usize

Get the number of remaining epochs.

pub fn embedding(&self) -> ArrayView2<'_, f32>

Get a view of the current embedding.

pub fn manifold(&self) -> &LearnedManifold

Get a reference to the learned manifold.

pub fn into_fitted(self, config: UmapConfig) -> FittedUmap

Consume the optimizer and return a lightweight fitted model.

This drops all the optimization state (epoch counters, preprocessed arrays) and keeps only the manifold and final embedding.

Trait Implementations

impl Debug for Optimizer

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl<'de> Deserialize<'de> for Optimizer

fn deserialize<__D>(__deserializer: __D) -> Result<Self, __D::Error>

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl Serialize for Optimizer

fn serialize<__S>(&self, __serializer: __S) -> Result<__S::Ok, __S::Error>

where __S: Serializer,

Serialize this value into the given Serde serializer.