t-sne 数据可视化网络中的部分参数+

       注：本代码主要实现对于网络中对于某个中间特征或计算得到的网络参数的可视化实现。如果仅可视化网络中的某个简单的参数，可以考虑使用 model.weight得到矩阵，然后放入分部代码中的降维可视化部分即可。TSNE参数说明

总体代码

import torch
import torch.nn as nn
import torch.nn.functional as F
from sklearn.manifold import TSNE
import matplotlib.pyplot as plt




class Network(nn.Module):  # extend nn.Module class of nn
    def __init__(self):
        super().__init__()  # super class constructor
        self.conv1 = nn.Conv2d(in_channels=1, out_channels=6, kernel_size=(5, 5))
        self.batchN1 = nn.BatchNorm2d(num_features=6)
        self.conv2 = nn.Conv2d(in_channels=6, out_channels=12, kernel_size=(5, 5))
        self.fc1 = nn.Linear(in_features=12 * 4 * 4, out_features=120)
        self.batchN2 = nn.BatchNorm1d(num_features=120)
        self.fc2 = nn.Linear(in_features=120, out_features=60)
        self.out = nn.Linear(in_features=60, out_features=10)

    def forward(self, t):  # implements the forward method (flow of tensors)

        # hidden conv layer
        t = self.conv1(t)
        t = F.max_pool2d(input=t, kernel_size=2, stride=2)
        t = F.relu(t)
        t = self.batchN1(t)

        # hidden conv layer
        t = self.conv2(t)
        t = F.max_pool2d(input=t, kernel_size=2, stride=2)
        t = F.relu(t)

        # flatten
        t = t.reshape(-1, 12 * 4 * 4)
        t = self.fc1(t)
        t = F.relu(t)
        t = self.batchN2(t)
        t = self.fc2(t)
        t = F.relu(t)

        # output
        t = self.out(t)

        return t
cnn_model = Network() # init model


pretrained_dict = cnn_model.state_dict()




class Identity(nn.Module):
    def __init__(self):
        super(Identity, self).__init__()
        self.conv1 = nn.Conv2d(in_channels=1, out_channels=6, kernel_size=(5, 5))

    def forward(self, x):
        return x


model = Identity()
model_dict = model.state_dict()
# model.load_state_dict(pretrained_dict) # RuntimeError: Error(s) in loading state_dict for Identity:	Unexpected key(s) in state_dict: "batchN1.weight", "batchN1.bias", "batchN1.running_mean",




# 1. filter out unnecessary keys
pretrained_dict = {k: v for k, v in pretrained_dict.items() if k in model_dict}
# 2. overwrite entries in the existing state dict
model_dict.update(pretrained_dict)
# 3. load the new state dict
model.load_state_dict(pretrained_dict)

vector = model.conv1.weight.detach().numpy()[0,0,:,:]

digits_final = TSNE(perplexity=30).fit_transform(vector)  #
plt.scatter(digits_final[:,0], digits_final[:,1])
plt.show()

功能分部代码

模型处理部分

import torch
import torch.nn as nn
import torch.nn.functional as F


class Network(nn.Module):  # extend nn.Module class of nn
    def __init__(self):
        super().__init__()  # super class constructor
        self.conv1 = nn.Conv2d(in_channels=1, out_channels=6, kernel_size=(5, 5))
        self.batchN1 = nn.BatchNorm2d(num_features=6)
        self.conv2 = nn.Conv2d(in_channels=6, out_channels=12, kernel_size=(5, 5))
        self.fc1 = nn.Linear(in_features=12 * 4 * 4, out_features=120)
        self.batchN2 = nn.BatchNorm1d(num_features=120)
        self.fc2 = nn.Linear(in_features=120, out_features=60)
        self.out = nn.Linear(in_features=60, out_features=10)

    def forward(self, t):  # implements the forward method (flow of tensors)

        # hidden conv layer
        t = self.conv1(t)
        t = F.max_pool2d(input=t, kernel_size=2, stride=2)
        t = F.relu(t)
        t = self.batchN1(t)

        # hidden conv layer
        t = self.conv2(t)
        t = F.max_pool2d(input=t, kernel_size=2, stride=2)
        t = F.relu(t)

        # flatten
        t = t.reshape(-1, 12 * 4 * 4)
        t = self.fc1(t)
        t = F.relu(t)
        t = self.batchN2(t)
        t = self.fc2(t)
        t = F.relu(t)

        # output
        t = self.out(t)

        return t
cnn_model = Network() # init model


pretrained_dict = cnn_model.state_dict()




class Identity(nn.Module):
    def __init__(self):
        super(Identity, self).__init__()
        self.conv1 = nn.Conv2d(in_channels=1, out_channels=6, kernel_size=(5, 5))

    def forward(self, x):
        return x


model = Identity()
model_dict = model.state_dict()
# model.load_state_dict(pretrained_dict) # RuntimeError: Error(s) in loading state_dict for Identity:	Unexpected key(s) in state_dict: "batchN1.weight", "batchN1.bias", "batchN1.running_mean",




# 1. filter out unnecessary keys
pretrained_dict = {k: v for k, v in pretrained_dict.items() if k in model_dict}
# 2. overwrite entries in the existing state dict
model_dict.update(pretrained_dict)
# 3. load the new state dict
model.load_state_dict(pretrained_dict)

降维可视化部分

• 参数解释

import numpy as np


import sklearn #Import scikitlearn for machine learning functionalities
from sklearn.manifold import TSNE
from sklearn.datasets import load_digits # For the UCI ML handwritten digits dataset

import matplotlib # Matplotlib 是 Python 中的一个库，它是 NumPy 库的数值数学扩展
import matplotlib.pyplot as plt
import matplotlib.patheffects as pe


import seaborn as sb


digits = load_digits()
print(digits.data.shape) # There are 10 classes (0 to 9) with alomst 180 images in each class
                         # The images are 8x8 and hence 64 pixels(dimensions)
plt.gray();
#Displaying what the standard images look like
for i in range(0,10):
    plt.matshow(digits.images[i])
    plt.show()


X = np.vstack([digits.data[digits.target==i] for i in range(10)]) # Place the arrays of data of each digit on top of each other and store in X
# X = np.random.random([1797, 64])

#Implementing the TSNE Function - ah Scikit learn makes it so easy!
digits_final = TSNE(perplexity=30).fit_transform(X) # plt.scatter(digits_final[0], digits_final[1])
#Play around with varying the parameters like perplexity, random_state to get different plots


# With the above line, our job is done. But why did we even reduce the dimensions in the first place?
# To visualise it on a graph.

# So, here is a utility function that helps to do a scatter plot of thee transformed data

def plot(x, colors):
    palette = np.array(sb.color_palette("hls", 10))  # Choosing color palette

    # Create a scatter plot.
    f = plt.figure(figsize=(8, 8))
    ax = plt.subplot(aspect='equal')
    sc = ax.scatter(x[:, 0], x[:, 1], lw=0, s=40, c=palette[colors.astype(np.int)])

    
    # 添加文本
    txts = []
    # for i in range(10):
    #     # Position of each label.
    #     xtext, ytext = np.median(x[colors == i, :], axis=0) # 返回数组元素的中位数。
    #     txt = ax.text(xtext, ytext, str(i), fontsize=24)  # Text(6.610861, 37.19979, '9')
    #     txt.set_path_effects([pe.Stroke(linewidth=5, foreground="w"), pe.Normal()])# 文本效果
    #     txts.append(txt)
    return f, ax, txts




Y = np.hstack([digits.target[digits.target==i] for i in range(10)]) # Place the arrays of data of each target digit by the side of each other continuosly and store in Y
plot(digits_final,Y)
plt.show()

t-sne 数据可视化的数学解释

参考与更多

pytorch官方保存加载模型教程(包含使用以 TorchScript 格式导出/加载模型)

模型字典的修改： https://discuss.pytorch.org/t/how-to-load-part-of-pre-trained-model/1113/2

scikit-learn.org

https://github.com/shivanichander/tSNE/blob/master/Code/tSNE%20Code.ipynb

t-SNE：最好的降维方法之一 - 知乎 (zhihu.com)

https://discuss.pytorch.org/t/changing-state-dict-value-is-not-changing-model/88695/2
model = nn.Linear(1, 1)
print(model.weight)

更多方法

# ISOMAP https://scikit-learn.org.cn/view/452.html
from sklearn.manifold import Isomap
digits_final = Isomap(n_components=2).fit_transform(res)

UMAP和TSNE：总述（持续更新）