import torch
import torch.nn as nn
import torch.nn.functional as F
[docs]
class CNN1DModel(nn.Module):
[docs]
def __init__(self, input_length=1000, num_classes=5):
"""
1D CNN for Exoplanet Spectrum Classification
"""
super(CNN1DModel, self).__init__()
self.conv1 = nn.Conv1d(
in_channels=1,
out_channels=16,
kernel_size=11,
padding=5)
self.bn1 = nn.BatchNorm1d(16)
self.pool1 = nn.MaxPool1d(kernel_size=4)
self.conv2 = nn.Conv1d(
in_channels=16,
out_channels=32,
kernel_size=7,
padding=3)
self.bn2 = nn.BatchNorm1d(32)
self.pool2 = nn.MaxPool1d(kernel_size=4)
self.conv3 = nn.Conv1d(
in_channels=32,
out_channels=64,
kernel_size=5,
padding=2)
self.bn3 = nn.BatchNorm1d(64)
self.pool3 = nn.MaxPool1d(kernel_size=4)
# Calculate flattened size
# length: 1000 -> 250 -> 62 -> 15
flat_length = input_length // 4 // 4 // 4
self.fc1 = nn.Linear(64 * flat_length, 128)
self.dropout = nn.Dropout(0.3)
self.fc2 = nn.Linear(128, num_classes)
def forward(self, x):
# x is (batch, channels=1, length)
x = self.pool1(F.relu(self.bn1(self.conv1(x))))
x = self.pool2(F.relu(self.bn2(self.conv2(x))))
x = self.pool3(F.relu(self.bn3(self.conv3(x))))
x = torch.flatten(x, 1)
x = F.relu(self.fc1(x))
x = self.dropout(x)
logits = self.fc2(x)
# Sigmoid for multi-label classification
probas = torch.sigmoid(logits)
return probas