deepbox/nn
Normalization & Regularization
Normalization layers and dropout-style regularization modules for stable training.
Normalization
AlphaDropout
Applies Alpha Dropout during training, designed for SELU-activated networks.
Dropout
Applies Dropout regularization during training.
Dropout2d
Applies 2D channel-wise Dropout during training.
BatchNorm1d
Batch Normalization layer.
BatchNorm2d
2D Batch Normalization.
BatchNorm3d
3D Batch Normalization.
GroupNorm
Group Normalization.
InstanceNorm
Instance Normalization.
InstanceNorm1d
1D Instance Normalization.
InstanceNorm2d
2D Instance Normalization.
InstanceNorm3d
3D Instance Normalization.
LayerNorm
Layer Normalization.
LocalResponseNorm
Local Response Normalization.
RMSNorm
Root Mean Square Layer Normalization (RMSNorm).
nn-normalization.ts
import { AlphaDropout, BatchNorm1d, Dropout, LayerNorm, RMSNorm,} from "deepbox/nn";console.log(new BatchNorm1d(32));console.log(new LayerNorm([32]));console.log(new RMSNorm(32));console.log(new Dropout(0.2), new AlphaDropout(0.1));