Module gpu_nn 

GPU-accelerated neural network compute (CPU mock backend).

Provides layer-wise forward passes, backpropagation gradients, and an Adam optimizer — all running on the CPU as a mock GPU backend.

Structs

AdamOptimizer

Adam adaptive moment estimator.

BackpropGpu

Backpropagation state: stores per-layer gradient tensors.

GpuNeuralNet

A sequential neural network backed by a CPU mock GPU context.

GpuTrainer

Combines a network, a backprop context, and an optimiser for training.

NeuralLayer

A single layer in a neural network, carrying weights, biases and a type.

Enums

LayerType

The computational type of a single neural network layer.

OptimizerType

Optimiser selection for the GPU trainer.

Functions

mse_loss

Mean-squared error: mean((pred_i - target_i)^2).

relu

Rectified linear unit: max(0, x).

sigmoid

Logistic sigmoid: 1 / (1 + e^{-x}).

softmax

Softmax of a slice: exp(x_i) / sum(exp(x_j)).