Day-9_SUMMER_TRAINING_AI/ML 🤑

day_9_dhruvdhayal_ai_ml (1).py

@@ -0,0 +1,224 @@
+# -*- coding: utf-8 -*-
+"""Day_9_DHRUVDHAYAL_AI_ML.ipynb
+
+Automatically generated by Colab.
+
+Original file is located at
+    https://colab.research.google.com/drive/1LH7a-QFZzgtKiDJ4bpBERaOdFycQWH2R
+
+#Neural Networks!
+"""
+
+# Step-1: Now, we have to import the Libraries Inbuilt or Pre-defined.
+from tensorflow import keras;
+import numpy as np;
+import matplotlib.pyplot as plt;
+
+#Load the MNIST Datasets from the keras Library.
+(Xtrain,ytrain),(Xtest,ytest)=keras.datasets.mnist.load_data();
+print(Xtrain.shape,ytrain.shape);
+print(Xtest.shape,ytest.shape);
+
+#Data Scaling.
+Xtrain=Xtrain/Xtrain.max();
+Xtest=Xtest/Xtest.max();
+
+print("\n --> Number of the Labels: ",len(np.unique(ytrain)));
+
+#Visualise the Values of the Data/Informations.
+import numpy as np;
+plt.figure(1,figsize=(6,8));
+for i in range(1,36+1,1):
+  temp=np.random.randint(0,60000);
+  im=Xtrain[temp,:,:];
+  lab=ytrain[temp];
+  plt.subplot(6,6,i);
+  plt.imshow(im,cmap='gray');
+  plt.title(lab);
+  plt.axis('off')
+
+#Creating the Neural Networks.
+nn_model=keras.Sequential(); #Create the empty 'nn' feed forward.
+#Framework.
+nn_model.add(keras.layers.Flatten(input_shape=(Xtrain.shape[1],Xtrain.shape[2])));  #An , Input Layer.
+
+#Hidden Layers.
+nn_model.add(keras.layers.Dense(128,activation='relu')); # Hidden Layer - 1.
+nn_model.add(keras.layers.Dense(256,activation='relu')); # Hidden Layer - 2.
+nn_model.add(keras.layers.Dense(256,activation='relu')); # Hidden Layer - 3.
+
+#Output Layer.
+nn_model.add(keras.layers.Dense(len(np.unique(ytrain))));
+
+print(nn_model.summary());
+
+#Optimizer Adding.
+nn_model.compile(optimizer='SGD',loss=keras.losses.SparseCategoricalCrossentropy(from_logits=True),metrics=['accuracy']) # Changed 'Loss' to 'loss'
+
+#TRrain the model for the Xtrain and Ytrain.
+history=nn_model.fit(Xtrain,ytrain,epochs=20);
+
+print(Xtrain.min," - ",Xtrain.max());
+
+# visualize the training loss and acc
+plt.figure(1,figsize=(4,2));
+plt.plot(history.epoch,history.history['loss'],'--ko');
+plt.grid('on');
+plt.title('Training loss graph ');
+plt.xlabel('Epochs');
+plt.ylabel('Loss');
+
+plt.figure(2,figsize=(4,2));
+plt.plot(history.epoch,history.history['accuracy'],'--ko');
+plt.grid('on');
+plt.title('Training accuracy graph ');
+plt.xlabel('Epochs');
+plt.ylabel('Accuracy');
+
+#Now, Evaluate the Test Data.
+[loss,acc]=nn_model.evaluate(Xtest,ytest);
+print("\n 1. Loss Values: ",loss);
+print("\n 2. Test Accuracy: ",acc);
+
+"""#Importing the Fashion Datsets."""
+
+# Step-1: Now, we have to import the Libraries Inbuilt or Pre-defined.
+from tensorflow import keras;
+import numpy as np;
+import matplotlib.pyplot as plt;
+
+#Load the MNIST Datasets from the keras Library.
+(Xtrain,ytrain),(Xtest,ytest)=keras.datasets.fashion_mnist.load_data();
+print(Xtrain.shape,ytrain.shape);
+print(Xtest.shape,ytest.shape);
+
+#Data Scaling.
+Xtrain=Xtrain/Xtrain.max();
+Xtest=Xtest/Xtest.max();
+
+print("\n --> Number of the Labels: ",len(np.unique(ytrain)));
+
+#Visualise the Values of the Data/Informations.
+import numpy as np;
+plt.figure(1,figsize=(6,8));
+for i in range(1,36+1,1):
+  temp=np.random.randint(0,60000);
+  im=Xtrain[temp,:,:];
+  lab=ytrain[temp];
+  plt.subplot(6,6,i);
+  plt.imshow(im,cmap='gray');
+  plt.title(lab);
+  plt.axis('off')
+
+#Creating the Neural Networks.
+nn_model=keras.Sequential(); #Create the empty 'nn' feed forward.
+#Framework.
+nn_model.add(keras.layers.Flatten(input_shape=(Xtrain.shape[1],Xtrain.shape[2])));  #An , Input Layer.
+
+#Hidden Layers.
+nn_model.add(keras.layers.Dense(128,activation='relu')); # Hidden Layer - 1.
+nn_model.add(keras.layers.Dense(256,activation='relu')); # Hidden Layer - 2.
+nn_model.add(keras.layers.Dense(256,activation='relu')); # Hidden Layer - 3.
+
+#Output Layer.
+nn_model.add(keras.layers.Dense(len(np.unique(ytrain))));
+
+print(nn_model.summary());
+
+#Optimizer Adding.
+nn_model.compile(optimizer='SGD',loss=keras.losses.SparseCategoricalCrossentropy(from_logits=True),metrics=['accuracy']) # Changed 'Loss' to 'loss'
+
+#TRrain the model for the Xtrain and Ytrain.
+history=nn_model.fit(Xtrain,ytrain,epochs=20);
+
+# visualize the training loss and acc
+plt.figure(1,figsize=(4,2));
+plt.plot(history.epoch,history.history['loss'],'--ko');
+plt.grid('on');
+plt.title('Training loss graph ');
+plt.xlabel('Epochs');
+plt.ylabel('Loss');
+
+plt.figure(2,figsize=(4,2));
+plt.plot(history.epoch,history.history['accuracy'],'--ko');
+plt.grid('on');
+plt.title('Training accuracy graph ');
+plt.xlabel('Epochs');
+plt.ylabel('Accuracy');
+
+#Now, Evaluate the Test Data.
+[loss,acc]=nn_model.evaluate(Xtest,ytest);
+print("\n 1. Loss Values: ",loss);
+print("\n 2. Test Accuracy: ",acc);
+
+"""#CIFAR DATSETS."""
+
+# Step-1: Now, we have to import the Libraries Inbuilt or Pre-defined.
+from tensorflow import keras;
+import numpy as np;
+import matplotlib.pyplot as plt;
+
+#Load the MNIST Datasets from the keras Library.
+(Xtrain,ytrain),(Xtest,ytest)=keras.datasets.cifar10.load_data();
+print(Xtrain.shape,ytrain.shape);
+print(Xtest.shape,ytest.shape);
+
+#Data Scaling.
+Xtrain=Xtrain/Xtrain.max();
+Xtest=Xtest/Xtest.max();
+
+print("\n --> Number of the Labels: ",len(np.unique(ytrain)));
+
+#Visualise the Values of the Data/Informations.
+import numpy as np;
+plt.figure(1,figsize=(6,8));
+for i in range(1,36+1,1):
+  temp=np.random.randint(0,50000); # Generate random integers within the valid range
+  im=Xtrain[temp,:,:];
+  lab=ytrain[temp];
+  plt.subplot(6,6,i);
+  plt.imshow(im,cmap='gray');
+  plt.title(lab);
+  plt.axis('off')
+
+#Creating the Neural Networks.
+nn_model=keras.Sequential(); #Create the empty 'nn' feed forward.
+#Framework.
+nn_model.add(keras.layers.Flatten(input_shape=(Xtrain.shape[1],Xtrain.shape[2], Xtrain.shape[3])));  #An , Input Layer. # Added Xtrain.shape[3] to account for color channels
+
+#Hidden Layers.
+nn_model.add(keras.layers.Dense(128,activation='relu')); # Hidden Layer - 1.
+nn_model.add(keras.layers.Dense(256,activation='relu')); # Hidden Layer - 2.
+nn_model.add(keras.layers.Dense(256,activation='relu')); # Hidden Layer - 3.
+
+#Output Layer.
+nn_model.add(keras.layers.Dense(len(np.unique(ytrain)))); # Removed the dense layer that was causing the shape mismatch
+
+print(nn_model.summary());
+
+#Optimizer Adding.
+nn_model.compile(optimizer='SGD',loss=keras.losses.SparseCategoricalCrossentropy(from_logits=True),metrics=['accuracy']) # Changed 'Loss' to 'loss'
+
+#TRrain the model for the Xtrain and Ytrain.
+history=nn_model.fit(Xtrain,ytrain,epochs=50);
+
+# visualize the training loss and acc
+plt.figure(1,figsize=(4,2));
+plt.plot(history.epoch,history.history['loss'],'--ko');
+plt.grid('on');
+plt.title('Training loss graph ');
+plt.xlabel('Epochs');
+plt.ylabel('Loss');
+
+plt.figure(2,figsize=(4,2));
+plt.plot(history.epoch,history.history['accuracy'],'--ko');
+plt.grid('on');
+plt.title('Training accuracy graph ');
+plt.xlabel('Epochs');
+plt.ylabel('Accuracy');
+
+#Now, Evaluate the Test Data.
+[loss,acc]=nn_model.evaluate(Xtest,ytest);
+print("\n 1. Loss Values: ",loss);
++print("\n 2. Test Accuracy: ",acc);
+