0% found this document useful (0 votes)
7 views5 pages

Deep Learning Model Code Overview

The document provides a detailed explanation of various deep learning models including Artificial Neural Networks (ANN), Convolutional Neural Networks (CNN), Recurrent Neural Networks (RNN), Long Short-Term Memory (LSTM), and Gated Recurrent Units (GRU). Each model's code is broken down step-by-step, highlighting the purpose of each layer and the training process. Key components such as activation functions, loss functions, and model compilation are also discussed.

Uploaded by

Hadia Ahmad
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
7 views5 pages

Deep Learning Model Code Overview

The document provides a detailed explanation of various deep learning models including Artificial Neural Networks (ANN), Convolutional Neural Networks (CNN), Recurrent Neural Networks (RNN), Long Short-Term Memory (LSTM), and Gated Recurrent Units (GRU). Each model's code is broken down step-by-step, highlighting the purpose of each layer and the training process. Key components such as activation functions, loss functions, and model compilation are also discussed.

Uploaded by

Hadia Ahmad
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd

Code Explanation

Task – Deep Learning Models Code Explanation

ANN – Artificial Neural Network

1. from [Link] import Sequential

 This line imports the Sequential model from Keras.


 Sequential allows you to add layers one by one, like stacking blocks.

2. from [Link] import Dense

 This brings in the Dense layer (also called a fully connected layer).
 It means every neuron in this layer is connected to all neurons in the next.

3. model = Sequential()

 This starts an empty neural network model.


 Layers will now be added to this model.

4. [Link](Dense(64, activation='relu',
input_shape=(input_dim,)))

 Adds the first hidden layer with 64 neurons.


 ReLU helps the model learn better.
 input_shape tells the model how many input values to expect.

5. [Link](Dense(1, activation='sigmoid')

 Adds the final layer with 1 neuron.


 Sigmoid is used for outputs between 0 and 1 (yes/no).

6. [Link](optimizer='adam',
loss='binary_crossentropy', metrics=['accuracy'])

 Prepares the model to learn using Adam optimizer.


 Uses binary crossentropy for binary classification.
 Tracks accuracy.

7. [Link](X_train, y_train, epochs=10, batch_size=32)

 Trains the model with input and answer data.


 Goes through the data 10 times in batches of 32.

CNN – Convolutional Neural Network

1. from [Link] import Conv2D, MaxPooling2D, Flatten,


Dense

 Brings in all necessary CNN layers:


 - Conv2D: detects patterns
 - MaxPooling2D: reduces image size
 - Flatten: reshapes data
 - Dense: normal layers

2. model = Sequential()

 Starts a new CNN model.

3. [Link](Conv2D(32, (3, 3), activation='relu',


input_shape=(28, 28, 1)))

 Adds a Conv layer with 32 filters.


 Each filter looks at 3x3 parts of the image.
 ReLU is the activation function.
 Input is 28x28 grayscale image.

4. [Link](MaxPooling2D(pool_size=(2, 2)))

 Downsamples the image to reduce size while keeping important parts.

5. [Link](Flatten())

 Flattens 2D image data into 1D before sending to Dense layers.

6. [Link](Dense(64, activation='relu'))

 Adds a Dense hidden layer to learn patterns.

7. [Link](Dense(10, activation='softmax'))

 Final layer with 10 outputs for 10 classes.


 Softmax converts output into probabilities.

8. [Link](optimizer='adam',
loss='categorical_crossentropy', metrics=['accuracy'])

 Compiles the model for multiclass classification.


9. [Link](X_train, y_train, epochs=10, batch_size=32)

 Trains the model for 10 epochs.

RNN – Recurrent Neural Network (for text)


1. from [Link] import Embedding, SimpleRNN, Dense

 Loads layers for sequence models.


 Embedding turns words into number vectors.

2. model = Sequential()

 Starts the model.

3. [Link](Embedding(input_dim=5000, output_dim=64))

 Turns each word into a 64-value vector.


 Only the 5000 most common words are used.

4. [Link](SimpleRNN(64))

 Adds an RNN layer that remembers previous steps.

5. [Link](Dense(1, activation='sigmoid'))

 Adds the output layer for binary prediction.

6. [Link](optimizer='adam',
loss='binary_crossentropy', metrics=['accuracy'])

 Compiles the model.

7. [Link](X_train, y_train, epochs=5, batch_size=64)

 Trains the RNN model.

LSTM – Long Short-Term Memory


1. from [Link] import LSTM

 Imports LSTM layer which remembers long-term info.

2. model = Sequential()

 Starts the model.

3. [Link](Embedding(input_dim=5000, output_dim=64))

 Turns words into 64-dimension vectors.


4. [Link](LSTM(128))

 Adds LSTM layer with 128 units to remember important words.

5. [Link](Dense(1, activation='sigmoid'))

 Output layer for binary output.

6. [Link](optimizer='adam',
loss='binary_crossentropy', metrics=['accuracy'])

 Compiles the model.

7. [Link](X_train, y_train, epochs=5, batch_size=64)

 Trains the model.

GRU – Gated Recurrent Unit


1. from [Link] import GRU

 Imports GRU layer, which is simpler and faster than LSTM.

2. model = Sequential()

 Starts the model.

3. [Link](Embedding(input_dim=5000, output_dim=64))

 Embeds the input words.

4. [Link](GRU(128))

 Adds the GRU layer with 128 units.

5. [Link](Dense(1, activation='sigmoid'))

 Output layer with sigmoid activation.

6. [Link](optimizer='adam',
loss='binary_crossentropy', metrics=['accuracy'])

 Prepares model for training.

7. [Link](X_train, y_train, epochs=5, batch_size=64)

 Trains the GRU.

_________________________________________________________________________________________________________

You might also like