Struct NeuralNet 

pub struct NeuralNet<T>
where
    T: Tensor<NeuralNetDataType> + TensorMath<NeuralNetDataType, MathOutput = T> + 'static,
{
    pub layers: Vec<Box<dyn Layer<T>>>,
    pub loss_fn: Box<dyn LossFunction<NeuralNetDataType, T>>,
    pub parameter_count: u64,
    pub label: String,
    pub name: String,
    /* private fields */
}

Feed-forward neural network container.

Holds an ordered list of Layer trait objects, the configured loss function, and training metadata (parameter count, name/label, and learning state). Instances are used to predict, fit, and save_model.

Fields

layers: Vec<Box<dyn Layer<T>>>

loss_fn: Box<dyn LossFunction<NeuralNetDataType, T>>

parameter_count: u64

label: String

name: String

Implementations

impl<T> NeuralNet<T>

where T: Tensor<NeuralNetDataType> + TensorMath<NeuralNetDataType, MathOutput = T> + 'static,

pub fn new( layers: Vec<Box<dyn Layer<T>>>, loss_fn: Box<dyn LossFunction<NeuralNetDataType, T>>, param_count: u64, label: String, name: String, current_epoch: usize, current_lr: NeuralNetDataType, ) -> Self

pub fn add(&mut self, layer: Box<dyn Layer<T>>)

Append a layer to the network.

The provided box must implement the Layer trait for the network’s tensor type T.

pub fn predict(&mut self, input: &T) -> Result<T, String>

Run a forward pass and return the network output for input.

pub fn fit<F>( &mut self, x_train: &T, y_train: &T, epochs: usize, epoch_offset: usize, base_lr: NeuralNetDataType, lr_adjustment: bool, hook: F, hook_interval: usize, ) -> Result<(), String>

where F: FnMut(usize, NeuralNetDataType, NeuralNetDataType, &mut Self),

pub fn save_model(&self, filepath: &str)

Serialize and write the model weights and metadata to filepath.