CS5720 - Week 6
Slide 102 of 120

Types of Sequence Problems

🔵
One-to-One
Traditional neural networks. Each input produces exactly one output, no sequence dependency.
Input
→
Network
→
Output
  • • Image classification
  • • Credit scoring
  • • Medical diagnosis
🔵➡️🟡🟡🟡
One-to-Many
Single input generates a sequence of outputs. Creating sequential content from fixed input.
Image
→
RNN
→
Word₁
Word₂
Word₃
  • • Image captioning
  • • Music generation
  • • Video description
🟡🟡🟡➡️🔵
Many-to-One
Sequence of inputs produces a single output. Analyzing entire sequence to make one prediction.
Word₁
Word₂
Word₃
→
RNN
→
Sentiment
  • • Sentiment analysis
  • • Document classification
  • • Activity recognition
🟡🟡🟡➡️🔴🔴🔴
Many-to-Many
Sequence input produces sequence output. Complex transformations between different sequences.
En₁
En₂
En₃
→
RNN
→
Fr₁
Fr₂
Fr₃
  • • Language translation
  • • Text summarization
  • • Video captioning

Problem Type Comparison

Problem Type Input Output Complexity Key Challenge
One-to-One Fixed size Fixed size Simple Feature extraction
One-to-Many Fixed size Variable sequence Medium When to stop generating
Many-to-One Variable sequence Fixed size Medium Sequence aggregation
Many-to-Many Variable sequence Variable sequence High Alignment and length
💡 Choosing the Right Type:
The problem type determines your network architecture. Understanding these patterns helps you choose the right approach for your specific use case!
← Previous Next →
Prepared by Dr. Gorkem Kar