Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                
Scribd Logo
Upload

Welcome to Scribd!

  • 8 views

    Classes

    Uploaded by

    heat mass
    The document describes a PyTorch implementation of a Siamese network for metric learning. It defines classes for a simple feedforward network, a Siamese network using two embedding networks, a triplet dataset, a binary classifier, and a Siamese binary classifier combining the Siamese network and binary classifier.

    Copyright:

    © All Rights Reserved
    Download as TXT, PDF, TXT or read online from Scribd

    Classes

    Uploaded by

    heat mass
    8 views2 pages
    The document describes a PyTorch implementation of a Siamese network for metric learning. It defines classes for a simple feedforward network, a Siamese network using two embedding networks, a triplet dataset, a binary classifier, and a Siamese binary classifier combining the Siamese network and binary classifier.

    Original Description:

    The classes used for the Siamese networks

    Original Title

    classes

    Copyright

    © © All Rights Reserved

    Available Formats

    TXT, PDF, TXT or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    The document describes a PyTorch implementation of a Siamese network for metric learning. It defines classes for a simple feedforward network, a Siamese network using two embedding networks, a triplet dataset, a binary classifier, and a Siamese binary classifier combining the Siamese network and binary classifier.

    Copyright:

    © All Rights Reserved
    Download as TXT, PDF, TXT or read online from Scribd
    Download as txt, pdf, or txt
    8 views2 pages

    Classes

    Uploaded by

    heat mass
    The document describes a PyTorch implementation of a Siamese network for metric learning. It defines classes for a simple feedforward network, a Siamese network using two embedding networks, a triplet dataset, a binary classifier, and a Siamese binary classifier combining the Siamese network and binary classifier.

    Copyright:

    © All Rights Reserved
    Download as TXT, PDF, TXT or read online from Scribd
    Download as txt, pdf, or txt
    of 2

    # %%

    import torch
    import torch.nn as nn
    import torch.optim as optim
    from torch.utils.data import Dataset, DataLoader
    import numpy as np

    # %%
    class SimpleFFNetwork(nn.Module):
    def __init__(self, input_length, hidden_sizes, output_size):
    super(SimpleFFNetwork, self).__init__()
    self.input_length = input_length
    self.hidden_sizes = hidden_sizes
    self.output_size = output_size

    self.layers = nn.ModuleList()
    prev_size = input_length
    for hidden_size in hidden_sizes:
    self.layers.append(nn.Linear(prev_size, hidden_size))
    self.layers.append(nn.ReLU())
    prev_size = hidden_size
    self.layers.append(nn.Linear(hidden_size, output_size))

    def forward(self, x):


    norms = torch.norm(x, dim=1, keepdim=True)
    x = x / norms
    for layer in self.layers:
    x = layer(x)
    return x

    # %%
    class SiameseNetwork(nn.Module):

    def __init__(self, input_lengths, hidden_sizes_list, output_size):


    super(SiameseNetwork, self).__init__()
    self.embedding_network1 = SimpleFFNetwork(input_lengths[0],
    hidden_sizes_list[0], output_size)
    self.embedding_network2 = SimpleFFNetwork(input_lengths[1],
    hidden_sizes_list[1], output_size)

    def forward(self, x1, x2):


    output1 = self.embedding_network1(x1)
    output2 = self.embedding_network2(x2)
    distance = torch.norm(output1 - output2, dim=1)
    return distance

    # %%
    class TripletDataset(Dataset):

    def __init__(self, dataframe):


    dataframe['anchor'] = dataframe['anchor'].apply(lambda x:
    x/np.linalg.norm(x))
    dataframe['positive'] = dataframe['positive'].apply(lambda x:
    x/np.linalg.norm(x))
    dataframe['negative'] = dataframe['negative'].apply(lambda x:
    x/np.linalg.norm(x))
    self.dataframe = dataframe

    def __len__(self):
    return len(self.dataframe)

    def __getitem__(self, idx):


    anchor = torch.tensor(self.dataframe.iloc[idx]['anchor'],
    dtype=torch.float32)
    positive = torch.tensor(self.dataframe.iloc[idx]['positive'],
    dtype=torch.float32)
    negative = torch.tensor(self.dataframe.iloc[idx]['negative'],
    dtype=torch.float32)
    return anchor, positive, negative

    # %%

    class BinaryClassifier(nn.Module):
    def __init__(self, input_size, hidden_size):
    super(BinaryClassifier, self).__init__()
    self.fc1 = nn.Linear(input_size, hidden_size)
    self.relu = nn.ReLU()
    self.fc2 = nn.Linear(hidden_size, 1) # Output size is 1 for binary
    classification

    def forward(self, x):


    x = self.fc1(x)
    x = self.relu(x)
    x = self.fc2(x)
    return x

    class SiameseBinaryClassifier(nn.Module):
    def __init__(self, siamese_network):
    super(SiameseBinaryClassifier, self).__init__()
    self.siamese_network = siamese_network
    self.fc = nn.Linear(1, 1)

    def forward(self, x1, x2):


    distance = self.siamese_network(x1, x2)
    output = torch.sigmoid(self.fc(distance))
    return output

    # %%

    You might also like