loki2.models.classifier.linear_classifier

Cell Classification Module for Loki2.

This module provides a linear classifier for cell type classification based on cell embeddings.

Module Contents

class loki2.models.classifier.linear_classifier.LinearClassifier(embed_dim: int, hidden_dim: int = 100, num_classes: int = 2, drop_rate: float = 0)

Bases: torch.nn.Module

Linear Classifier for cell type classification.

A two-layer fully connected network with ReLU activation and optional dropout for classifying cells based on their embeddings.

Parameters:

• embed_dim – Embedding dimension (input dimension).

• hidden_dim – Hidden layer dimension. Defaults to 100.

• num_classes – Number of output classes. Defaults to 2.

• drop_rate – Dropout rate. Defaults to 0.

fc1

fc2

activation

dropout

forward(x: torch.Tensor) → torch.Tensor

Forward pass through the classifier.

Parameters:

x – Input tensor of shape (batch_size, embed_dim).

Returns:

Output logits of shape (batch_size, num_classes).

Return type:

torch.Tensor