NeuroViz - Neural Network Visualiser

Choose a dataset to get started

Select one above, then click Start Training to see the model learn.

Training history will appear here

Samples: 200 More samples = better training, but slower Noise: 0.1 Class Balance: 50% Preprocessing Output Classes

CSV

Learning Rate How big each training step is Optimizer

Loss Function Batch Norm

Hidden Layers Neurons per layer; more = more complex boundaries

Activation

Create the neural network with these settings

Batch Size Max Epochs

Batch Normalization Speed (FPS): 60

Advanced training settings ▼

Performance Mode LR Schedule Validation Split

Regularization ▼

L1 L2 Dropout

Network Diagram ▼

Weight Distribution ▼

Confusion Matrix ▼

—

Prec

—

Rec

—

Layer Activations ▼

Model Complexity ▼

Parameters

—

FLOPs/Forward

—

Memory (Params) —

Memory (Activations) —

Total Memory —

Adversarial Examples ▼

Generate points that fool the model - small perturbations that cause misclassification.

Attack Method:

Spiral search pattern, fast but less accurate

Sample Size: 20

5 100

Training Speed ▼

Current: —