Models¶
tinytopics.models
¶
NeuralPoissonNMF
¶
Bases: Module
__init__(n, m, k, device=None)
¶
Neural Poisson NMF model with sum-to-one constraints on document-topic and topic-term distributions.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
n
|
int
|
Number of documents. |
required |
m
|
int
|
Number of terms (vocabulary size). |
required |
k
|
int
|
Number of topics. |
required |
device
|
device | None
|
Device to run the model on. Defaults to CPU. |
None
|
forward(doc_indices)
¶
Forward pass of the neural Poisson NMF model.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
doc_indices
|
Tensor
|
Indices of documents in the batch. |
required |
Returns:
| Type | Description |
|---|---|
Tensor
|
Reconstructed document-term matrix for the batch. |
get_normalized_F()
¶
Get the learned, normalized topic-term distribution matrix (F).
Returns:
| Type | Description |
|---|---|
Tensor
|
Normalized F matrix on the CPU. |
get_normalized_L()
¶
Get the learned, normalized document-topic distribution matrix (L).
Returns:
| Type | Description |
|---|---|
Tensor
|
Normalized L matrix on the CPU. |