Niranjankumar-c

# 4. Remove punctuation

main_content_string = main_content_string.translate(str.maketrans('', '', string.punctuation))
main_content_string = main_content_string.replace("‘", '').replace("’", '').replace("'", '')
main_content_string[:300]

Niranjankumar-c

#import required packages
from nltk.tokenize import word_tokenize
from wordcloud import WordCloud, STOPWORDS, ImageColorGenerator
import matplotlib.pyplot as plt 
from collections import Counter
import seaborn as sns
from nltk.corpus import stopwords
import re
import string

Niranjankumar-c

class CNN_BN(nn.Module):
    def __init__(self): 
        super(MyNetBN, self).__init__()
        self.features = nn.Sequential(
            nn.Conv2d(1, 3, 5),         # (N, 1, 28, 28) -> (N,  3, 24, 24)
            nn.ReLU(),
            nn.AvgPool2d(2, stride=2),  # (N, 3, 24, 24) -> (N,  3, 12, 12)
            nn.Conv2d(3, 6, 3),
            nn.BatchNorm2d(6)           # (N, 3, 12, 12) -> (N,  6, 10, 10) 
        )

Niranjankumar-c

class MyNet(nn.Module):
    def __init__(self): 
        super(MyNet, self).__init__()
        self.classifier = nn.Sequential(
            nn.Linear(784, 48),  # 28 x 28 = 784 flatten the input image
            nn.ReLU(),
            nn.Linear(48, 24),
            nn.ReLU(),
            nn.Linear(24, 10)
        )

Niranjankumar-c

#create a neural network with out dropout
N_h = 100 #hidden nodes

model = torch.nn.Sequential(
    nn.Linear(1, N_h),
    nn.ReLU(),
    nn.Linear(N_h, N_h),
    nn.ReLU(),
    nn.Linear(N_h, 1)
)

Niranjankumar-c

N = 50 #number of data points
noise = 0.3

#generate the train data
X_train = torch.unsqueeze(torch.linspace(-1, 1, N),1)
Y_train = X_train + noise * torch.normal(torch.zeros(N,1), torch.ones(N,1))

#generate the test data
X_test = torch.unsqueeze(torch.linspace(-1,1,N),1)
Y_test = X_test + noise * torch.normal(torch.zeros(N,1), torch.ones(N,1))

Niranjankumar-c

N = 50 #number of data points
noise = 0.3

#generate the train data
X_train = torch.unsqueeze(torch.linspace(-1, 1, N),1)
Y_train = X_train + noise * torch.normal(torch.zeros(N,1), torch.ones(N,1))

#generate the test data
X_test = torch.unsqueeze(torch.linspace(-1,1,N),1)
Y_test = X_test + noise * torch.normal(torch.zeros(N,1), torch.ones(N,1))

Niranjankumar-c

#compute occlusion heatmap
heatmap = occlusion(model, images, pred[0].item(), 32, 14)

#displaying the image using seaborn heatmap and also setting the maximum value of gradient to probability
imgplot = sns.heatmap(heatmap, xticklabels=False, yticklabels=False, vmax=prob_no_occ)
figure = imgplot.get_figure()

Niranjankumar-c

#custom function to conduct occlusion experiments

def occlusion(model, image, label, occ_size = 50, occ_stride = 50, occ_pixel = 0.5):
  
    #get the width and height of the image
    width, height = image.shape[-2], image.shape[-1]
  
    #setting the output image width and height
    output_height = int(np.ceil((height-occ_size)/occ_stride))
    output_width = int(np.ceil((width-occ_size)/occ_stride))

Niranjankumar-c

def plot_weights(model, layer_num, single_channel = True, collated = False):
  
  #extracting the model features at the particular layer number
  layer = model.features[layer_num]
  
  #checking whether the layer is convolution layer or not 
  if isinstance(layer, nn.Conv2d):
    #getting the weight tensor data
    weight_tensor = model.features[layer_num].weight.data