Layer 2: BatchNorm | Physics-Informed CNN-BiLSTM

📊 Statistics

0.00

Mean (μ)

σ²

1.00

Variance (σ²)

1.00

Scale (γ)

0.00

Shift (β)

🔄 Mode

Training

Batch statistics

Inference

Running statistics

Normalization Formula

x̂ = (x - μ) / √(σ² + ε)

y = γ × x̂ + β

BatchNorm stabilizes training by normalizing layer inputs. It reduces internal covariate shift and allows higher learning rates.

Effect: Notice how the "wild" orange data is centered and standardized into the blue distribution.

🎮 Controls

Scale (γ)1.00

Shift (β)0.00

Animation Speed1.0x

📐 Running Stats

0.12

Running Mean

0.95

Running Var

Uses batch mean/variance. Updates running statistics with momentum for inference.

نظرة تفصيلية: تسوية أو تطبيع الدفعة (BatchNorm)