Algorithm Research Roadmap Template (Deep Learning)

flowchart TB
    subgraph P1[Problem definition]
        direction LR
        A[Task and evaluation metrics] --> B[Survey existing methods and gaps]
    end
    subgraph P2[Data]
        direction LR
        C[Data collection] --> D[Cleaning and labelling] --> E[Train / validation / test split]
    end
    subgraph P3[Method]
        direction LR
        F[Reproduce baselines] --> G[Proposed model] --> H[Loss function and training strategy]
    end
    subgraph P4[Experiments]
        direction LR
        I[Comparison with baselines] --> J[Ablation study] --> K[Sensitivity analysis]
    end
    subgraph P5[Conclusion]
        direction LR
        L[Analysis of results] --> M[Conclusions and future work]
    end
    P1 --> P2 --> P3 --> P4 --> P5

Draw a deep learning research roadmap in Mermaid flowchart TB syntax with five subgraph phases: Problem definition (task and evaluation metrics → survey existing methods and gaps), Data (data collection → cleaning and labelling → train/validation/test split), Method (reproduce baselines → proposed model → loss function and training strategy), Experiments (comparison with baselines → ablation study → sensitivity analysis), Conclusion (analysis of results → conclusions and future work). Add direction LR inside each subgraph so the steps run horizontally, and connect the phases by their subgraph ids (for example S1 --> S2). Output only the mermaid code block.