[Solved] You want to classify movie reviews using a Bidirectional LSTM on sequences of length 50 with... Which model architecture is best suited to produce a fixed-length vector for classification? | NLP

NLP - Sequence Models for NLP

You want to classify movie reviews using a Bidirectional LSTM on sequences of length 50 with 100-dimensional embeddings. Which model architecture is best suited to produce a fixed-length vector for classification?

AEmbedding(input_dim=10000, output_dim=100, input_length=50) -> Bidirectional(LSTM(64)) -> Dense(1, activation='sigmoid')

BEmbedding(input_dim=10000, output_dim=100, input_length=50) -> Bidirectional(LSTM(64, return_sequences=true)) -> Dense(1, activation='sigmoid')

CEmbedding(input_dim=10000, output_dim=100, input_length=50) -> Bidirectional(LSTM(64, return_sequences=true)) -> GlobalMaxPooling1D() -> Dense(1, activation='sigmoid')

DEmbedding(input_dim=10000, output_dim=100, input_length=50) -> LSTM(64) -> Dense(1, activation='sigmoid')

Step-by-Step Solution

Solution:

Step 1: Understand output shape
Bidirectional LSTM without return_sequences outputs a fixed-length vector.
Step 2: Check model options
Embedding(input_dim=10000, output_dim=100, input_length=50) -> Bidirectional(LSTM(64)) -> Dense(1, activation='sigmoid') uses Bidirectional LSTM without return_sequences, producing fixed-size vector for Dense.
Step 3: Analyze other options
Options B and C output sequences, requiring pooling or flattening; D is unidirectional.
Final Answer:
Embedding -> Bidirectional(LSTM(64)) -> Dense(1, activation='sigmoid') -> Option A
Quick Check:
Use return_sequences=false for fixed vector output [OK]

Quick Trick: return_sequences=false outputs fixed vector [OK]

Common Mistakes:

MISTAKES

Using return_sequences=true without pooling
Choosing unidirectional LSTM when bidirectional is needed

Master "Sequence Models for NLP" in NLP

9 interactive learning modes - each teaches the same concept differently

Learn Why Deep Model Try Challenge Experiment Recall Metrics

More NLP Quizzes

Which model architecture is best suited to produce a fixed-length vector for classification?

Step 1: Understand output shape

Step 2: Check model options

Step 3: Analyze other options

Final Answer:

Quick Check:

Want More Practice?