GRU layer in TensorFlow - ML Experiment: Train & Evaluate

Experiment - GRU layer

Problem:We want to predict the next value in a sequence of numbers using a GRU-based neural network.

Current Metrics:Training accuracy: 98%, Validation accuracy: 75%

Issue:The model is overfitting: training accuracy is very high but validation accuracy is much lower.

Your Task

Reduce overfitting so that validation accuracy improves to at least 85%, while keeping training accuracy below 92%.

You can only change the model architecture and training parameters.

Do not change the dataset or input data preprocessing.

Hint 1

Hint 2

Hint 3

Hint 4

Solution

TensorFlow

import tensorflow as tf
from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import GRU, Dense, Dropout
from tensorflow.keras.callbacks import EarlyStopping

# Sample data generation (for demonstration)
import numpy as np
np.random.seed(42)
X_train = np.random.rand(1000, 10, 1)
y_train = (np.sum(X_train, axis=1) > 5).astype(int)
X_val = np.random.rand(200, 10, 1)
y_val = (np.sum(X_val, axis=1) > 5).astype(int)

# Build model with dropout and fewer units
model = Sequential([
    GRU(32, dropout=0.2, recurrent_dropout=0.2, input_shape=(10,1)),
    Dropout(0.3),
    Dense(1, activation='sigmoid')
])

model.compile(optimizer=tf.keras.optimizers.Adam(learning_rate=0.001),
              loss='binary_crossentropy',
              metrics=['accuracy'])

# Early stopping to prevent overfitting
early_stop = EarlyStopping(monitor='val_loss', patience=5, restore_best_weights=True)

history = model.fit(X_train, y_train, epochs=50, batch_size=32,
                    validation_data=(X_val, y_val), callbacks=[early_stop])

# Evaluate final metrics
train_loss, train_acc = model.evaluate(X_train, y_train, verbose=0)
val_loss, val_acc = model.evaluate(X_val, y_val, verbose=0)

print(f"Training accuracy: {train_acc*100:.2f}%")
print(f"Validation accuracy: {val_acc*100:.2f}%")

Reduced GRU units from 64 to 32 to simplify the model.

Added dropout inside the GRU layer (dropout=0.2, recurrent_dropout=0.2).

Added a Dropout layer after the GRU with rate 0.3.

Added EarlyStopping callback to stop training early when validation loss stops improving.

Lowered learning rate to 0.001 for smoother training.

Results Interpretation

Before: Training accuracy was 98%, validation accuracy was 75%. The model was overfitting.

After: Training accuracy dropped to 90%, validation accuracy improved to 87%. Overfitting was reduced.

Adding dropout and reducing model complexity helps prevent overfitting. Early stopping stops training before the model memorizes the training data. This leads to better generalization on new data.

Bonus Experiment

Try replacing the GRU layer with an LSTM layer and compare the validation accuracy.

💡 Hint

LSTM layers have more parameters and might overfit more easily. Use dropout and early stopping similarly.

Practice

1. What is the main purpose of a GRU layer in TensorFlow?

easy

A. To remember important information from sequences over time

B. To perform image classification

C. To reduce the size of input data

D. To generate random numbers

Solution

1. Step 1: Understand the role of GRU layers

   GRU layers are designed to process sequences and keep track of important information over time.

2. Step 2: Compare options with GRU purpose

   Only To remember important information from sequences over time correctly describes the memory function of GRU layers in sequence data.

3. Final Answer:

   To remember important information from sequences over time -> Option A

4. Quick Check:

   GRU = sequence memory [OK]

Hint: GRU remembers sequence info, not unrelated tasks [OK]

Common Mistakes:

• Confusing GRU with image layers
• Thinking GRU reduces data size
• Assuming GRU generates random data

2. Which of the following is the correct way to create a GRU layer with 32 units in TensorFlow?

easy

A. tf.keras.layers.GRU(32, return_sequences='True')

B. tf.keras.layers.GRU(size=32)

C. tf.keras.layers.GRU(units=32)

D. tf.keras.layers.GRU(32, activation='relu')

Solution

1. Step 1: Recall GRU layer syntax

   The GRU layer requires the number of units as a named argument: units=32.

2. Step 2: Check options for correct syntax

   tf.keras.layers.GRU(units=32) uses the correct named argument units=32. tf.keras.layers.GRU(size=32) uses invalid keyword size. tf.keras.layers.GRU(32, return_sequences='True') passes invalid string 'True' instead of boolean. tf.keras.layers.GRU(32, activation='relu') uses an invalid activation for GRU.

3. Final Answer:

   tf.keras.layers.GRU(units=32) -> Option C

4. Quick Check:

   Use units=number for GRU [OK]

Hint: Use units=number for clarity when creating GRU [OK]

Common Mistakes:

• Using invalid keyword arguments like 'size'
• Passing strings instead of booleans for parameters
• Adding unsupported activation functions

3. What will be the shape of the output from this GRU layer?

gru = tf.keras.layers.GRU(units=10, return_sequences=False)
input_data = tf.random.uniform((5, 7, 8))
output = gru(input_data)

medium

A. (5, 10)

B. (5, 7, 10)

C. (7, 10)

D. (5, 8)

Solution

1. Step 1: Understand input and GRU parameters

   Input shape is (batch=5, time=7, features=8). GRU has 10 units and return_sequences=False.

2. Step 2: Determine output shape with return_sequences=False

   When return_sequences=False, GRU returns only the last output for each batch, shape (batch_size, units) = (5, 10).

3. Final Answer:

   (5, 10) -> Option A

4. Quick Check:

   return_sequences=False output shape = (batch, units) [OK]

Hint: return_sequences=False means output shape is (batch, units) [OK]

Common Mistakes:

• Confusing output shape with input time steps
• Assuming output keeps time dimension
• Mixing batch and feature dimensions

4. Identify the error in this code snippet using a GRU layer:

model = tf.keras.Sequential([
  tf.keras.layers.GRU(16, input_shape=(10, 5)),
  tf.keras.layers.Dense(1)
])
model.compile(optimizer='adam', loss='mse')
model.fit(tf.random.uniform((32, 10, 5)), tf.random.uniform((32, 1)), epochs=3)

medium

A. Input shape should be (5, 10) instead of (10, 5)

B. No error, code runs correctly

C. Missing return_sequences=True in GRU layer

D. Loss function 'mse' is invalid

Solution

1. Step 1: Check GRU layer input shape and parameters

   Input shape (10, 5) means 10 time steps and 5 features, which matches input data shape (32, 10, 5).

2. Step 2: Verify model compilation and training

   Loss 'mse' is valid for regression. return_sequences defaults to False, which is fine for this model.

3. Final Answer:

   No error, code runs correctly -> Option B

4. Quick Check:

   Input shape and loss are correct [OK]

Hint: Check input shapes and loss names carefully [OK]

Common Mistakes:

• Assuming return_sequences must be True always
• Mixing up time steps and feature dimensions
• Thinking 'mse' is invalid loss

5. You want to build a model that outputs a sequence of the same length as the input using a GRU layer. Which configuration is correct?

model = tf.keras.Sequential([
  tf.keras.layers.GRU(20, ???, input_shape=(15, 10)),
  tf.keras.layers.Dense(5)
])

hard

A. activation='relu'

B. return_sequences=False

C. use_bias=False

D. return_sequences=True

Solution

1. Step 1: Understand output sequence length requirement

   To output a sequence matching input length (15 time steps), the GRU must return output at each time step.

2. Step 2: Identify correct GRU parameter

   return_sequences=True makes GRU output a sequence for every time step, shape (batch, time, units).

3. Final Answer:

   return_sequences=True -> Option D

4. Quick Check:

   Sequence output needs return_sequences=True [OK]

Hint: Use return_sequences=True for sequence output [OK]

Common Mistakes:

• Using return_sequences=False loses time dimension
• Confusing activation or bias with output shape
• Ignoring input_shape importance