ann model

Question 1

What is Sequential in TensorFlow?

Answer 1

In TensorFlow/Keras, Sequential is a simple way to build a feedforward neural network (ANN, CNN, etc.). It stacks layers one after another in a linear fashion.

Question 2

How to import sequential model?

Answer 2

from tensorflow.keras.models import Sequential

from tensorflow.keras.layers import Dense

Question 3

How to Create a Sequential Model for regression?

Answer 3

model = Sequential([
Dense(16, activation=’relu’, input_shape=(X_train.shape[1],)), # Input layer

Dense(8, activation='relu'),  # Hidden layer

Dense(1)  # Output layer (Regression problem, no activation function) ])

Question 4

How to create sequential model for binary classification ?

Answer 4

model = Sequential([
Dense(64, activation=’relu’, input_shape=(10,)),

Dense(32, activation='relu'),

Dense(1, activation='sigmoid')  # For binary classification ])

Question 5

How to create sequential model using .add() method?

Answer 5

model = Sequential()
model.add(Dense(64, activation=’relu’, input_shape=(10,)))
model.add(Dense(32, activation=’relu’))
model.add(Dense(1, activation=’sigmoid’))

Question 6

What are the common layers used in sequential?

Answer 6

Dense(units, activation) – Fully connected layer

Conv2D(filters, kernel_size, activation) – CNN layer

Dropout(rate) – Reduces overfitting

Flatten() – Converts multi-dimensional input to 1D

Question 7

Give example of sequential model with 20% dropout

Answer 7

model = Sequential([
Dense(128, activation=’relu’, input_shape=
(784,)), # Input Layer
Dropout(0.2), # Dropout layer (20% neurons
randomly dropped)

Dense(64, activation='relu'),  # Hidden Layer
Dropout(0.2),

Dense(10, activation='softmax')  # Output 
                                     Layer (for classification) ])

Question 8

How to import dropout?

Answer 8

from tensorflow.keras.layers import Dropout

model.add(Dropout(0.2))
# Drops 20% of neurons randomly

Question 9

How to Compile Sequential Model in Keras? and explain parameters

Answer 9

model.compile(optimizer=’adam’,
loss=’categorical_crossentropy’,
metrics=[‘accuracy’])

After defining a neural network model in Keras, you need to compile it before training. Compilation configures the model with:

✅ Optimizer → Decides how the model updates weights
✅ Loss Function → Measures how wrong the predictions are
✅ Metrics → Tracks performance (e.g., accuracy)

Explanation of Parameters
1️⃣ Optimizer → Controls how weights are updated
‘adam’ → Best for most cases (adaptive learning)
‘sgd’ → Slower but works well for simple models
‘rmsprop’ → Good for RNNs

2️⃣ Loss Function → Measures how far predictions are from actual values

For Classification:
‘categorical_crossentropy’ (Multi-class with one-hot labels)
‘sparse_categorical_crossentropy’ (Multi-class with integer labels)
‘binary_crossentropy’ (For binary classification)

For Regression:
‘mean_squared_error’ (MSE)
‘mean_absolute_error’ (MAE)

3️⃣ Metrics → Used for evaluation
‘accuracy’ → Works for classification
‘mae’ → Mean Absolute Error (for regression)

Question 10

How to Train Sequential Model?

Answer 10

history = model.fit(X_train, y_train, epochs=100, batch_size=32, validation_data=(X_test, y_test))

epochs=100 → Number of times model sees data
batch_size=32 → Number of samples processed at once
validation_data=(X_test, y_test) → Checks performance on unseen data

Question 11

How to Predict for new data?

Answer 11

For Regression (Continuous Output)
predictions = model.predict(X_test)
print(predictions[:5]) # Show first 5 predictions
Output: [250000, 180000, 320000, 410000, 150000] (Predicted house prices)

For Binary Classification (Yes/No, 0/1)
predictions = model.predict(X_test)
predicted_labels = (predictions > 0.5).astype(int) # Convert probabilities to 0 or 1
print(predicted_labels[:5])
Output: [0, 1, 0, 0, 1] (Spam or Not Spam)

For Classification (Multi-Class)
predictions = model.predict(X_test)
predicted_classes = predictions.argmax(axis=1) # Get class with highest probability
print(predicted_classes[:5])
Output: [1, 0, 2, 1, 2] (Predicted class indices)