nlinear_module
NLinear network: one linear layer with last-value normalization (NF-faithful).
NLinearNet
nlinear_module.NLinearNet
forecast = (y - y_last) @ W.T + b + y_last. I/O [B, L, 1] -> [B, h, 1].
__init__(self, h: int, input_size: int, *, rngs: nnx.Rngs)
(No prose summary available.)
| Parameter | Type | Default | Description |
|---|---|---|---|
h |
int |
- | (undocumented) |
input_size |
int |
- | (undocumented) |
rngs |
nnx.Rngs |
- | (undocumented) |
__call__(self, x: jnp.ndarray) -> jnp.ndarray
(No prose summary available.)
| Parameter | Type | Default | Description |
|---|---|---|---|
x |
jnp.ndarray |
- | (undocumented) |
Returns: jnp.ndarray (Output data, shape [B, h, 1]).