SNetwork class

A sequential neural network architecture that linearly stacks and executes Layer instances. Manages sequential forward propagation, parameter aggregation across constituent layers, training loops via fit, weight serialization to JSON via save and load, and autograd graph visualization.

Inheritance

• Object
• Layer
• SNetwork

Constructors

SNetwork(List<Layer> layers, {String name = 'snetwork'})

Creates an SNetwork with the given layers sequence and optional name.

Properties

hashCode → int

The hash code for this object.

no setter, inherited

layers ↔ List<Layer>

Ordered list of Layer instances composing the sequential pipeline.

getter/setter pair

name ↔ String

The assigned name of this network architecture for debugging and inspection.

getter/setter pair, override-getter

optimizer ↔ Optimizer

The Optimizer assigned via compile to update network parameters during training.

getter/setter pair

parameters → List<Tensor>

Returns all trainable parameters aggregated across every layer in layers.

no setter, override

runtimeType → Type

A representation of the runtime type of the object.

no setter, inherited

Methods

build(Tensor input) → void

Allocates and initializes weights/biases based on the incoming input shape. The Tensor input will be used for the setup and therefore should match the exact Tensor shape used in forward.

inherited

call(Tensor input) → Tensor

Helper function that, if not overwritten, checks whether build has already been called; if not, it calls build and only then forward.

inherited

compile({required Optimizer configuredOptimizer}) → void

Configures the network by binding the specified configuredOptimizer for subsequent training.

evaluate(List<List<double>> inputs, List<List<double>> targets) → double

Computes binary classification accuracy on inputs and targets by thresholding predictions at 0.5.

fit(List<List<double>> inputs, List<List<double>> targets, {int epochs = 100, bool averageWeight = false, bool debug = true}) → void

Trains the network parameters on provided inputs and targets over epochs using MSE loss and the assigned optimizer.

forward(Tensor input) → Tensor

Sequentially executes the forward pass across each layer in layers, routing each output to the subsequent layer.

override

getWeights() → Map<String, dynamic>

Serializes the weights of all parameterized layers into an indexed map.

override

inspectGraph(List<double> inputData, List<double> targetData) → void

Executes a forward pass with inputData and targetData to compute MSE loss and prints the autograd compute graph.

load(String filePath) → Future<void>

Reads serialized network weights from a JSON file at filePath and applies them to the network layers.

noSuchMethod(Invocation invocation) → dynamic

Invoked when a nonexistent method or property is accessed.

inherited

predict(Tensor input) → Tensor

Executes an inference pass on the given input tensor by invoking the layer call pipeline.

save(String filePath) → Future<void>

Serializes network weights into formatted JSON and saves the output to filePath.

setWeights(Map<String, dynamic> networkWeights) → void

Deserializes and loads weights into each parameterized layer from the provided networkWeights map.

override

toString() → String

A string representation of this object.

inherited

Operators

operator ==(Object other) → bool

The equality operator.

inherited