# matheustrindade2002/multiobjetglad architecture

> Architecture diagram of the GitHub repository matheustrindade2002/multiobjetglad, generated by GitDiagram from its file tree, README and a sample of source files (last updated 2026-10-03). It is an AI-made overview: check details against the source.

- Interactive diagram: https://gitdiagram.com/matheustrindade2002/multiobjetglad
- Repository: https://github.com/matheustrindade2002/multiobjetglad

## Overview

InstanceSelectionMO is an experimental SST-2 text-classification workflow for comparing confidence-only pseudo-labeling with robust multiobjective selection. The main path builds labeled, unlabeled, and hold-out splits; trains a BERT baseline; scores the unlabeled pool; selects examples; retrains and evaluates both strategies; then writes diagnostics, metrics, and plots. Selection internals are shown where source supports them. The graph follows the sampled modules and README; unsampled implementation details are not inferred.

Main entry points: [`data_utils.py`](https://github.com/matheustrindade2002/multiobjetglad/blob/main/data_utils.py) (Dataset splits), [`modeling.py`](https://github.com/matheustrindade2002/multiobjetglad/blob/main/modeling.py) (Pool predictions and embeddings), [`selection.py`](https://github.com/matheustrindade2002/multiobjetglad/blob/main/selection.py) (Selection diagnostics), [`plotting.py`](https://github.com/matheustrindade2002/multiobjetglad/blob/main/plotting.py) (Analysis plots)

## Diagram (Mermaid)

```mermaid
flowchart TD

subgraph group_data["Data preparation"]
  node_config["Experiment config<br/>[config.py]"]
  node_splits["Dataset splits<br/>[data_utils.py]"]
  node_torchdata["Tokenized datasets<br/>[data_utils.py]"]
  node_augmented_data["Pseudo-labeled training data<br/>[pipeline.py]"]
end

subgraph group_models["BERT modeling"]
  node_tokenizer["BERT tokenizer<br/>[modeling.py]"]
  node_bert["BERT model<br/>[modeling.py]"]
  node_training["Train and evaluate<br/>[modeling.py]"]
  node_pool_scores["Pool predictions and embeddings<br/>[modeling.py]"]
end

subgraph group_selection["Pseudo-label selection"]
  node_typicality["kNN typicality<br/>[selection.py]"]
  node_robust_confidence["Bootstrap confidence<br/>[selection.py]"]
  node_pareto["MOO Pareto selection<br/>[selection.py]"]
  node_naive["Naive top-confidence<br/>[selection.py]"]
  node_selection_table["Selection diagnostics<br/>[selection.py]"]
end

subgraph group_results["Analysis and results"]
  node_test_metrics["Hold-out metrics<br/>[pipeline.py]"]
  node_dataset_report["Split report"]
  node_results_json["Experiment results<br/>[results.json]"]
  node_results_md["Results summary<br/>[RESULTS.md]"]
  node_plots["Analysis plots<br/>[plotting.py]"]
  node_plot_artifacts["Plot files"]
end

node_researcher(("Experiment runner"))
node_sst2[("SST-2 dataset")]

node_researcher -->|"runs pipeline"| node_splits
node_sst2 -->|"provides examples"| node_splits
node_config -->|"configures splits"| node_splits
node_config -->|"configures model"| node_bert
node_config -->|"sets output path"| node_plots
node_splits -->|"creates datasets"| node_torchdata
node_tokenizer -->|"tokenizes examples"| node_torchdata
node_torchdata -->|"supplies batches"| node_training
node_bert -->|"trains and evaluates"| node_training
node_training -->|"scores pool"| node_pool_scores
node_pool_scores -->|"provides embeddings"| node_typicality
node_pool_scores -->|"provides confidences"| node_robust_confidence
node_typicality -->|"supplies typicality"| node_robust_confidence
node_robust_confidence -->|"supplies robust scores"| node_pareto
node_pool_scores -->|"supplies predictions"| node_pareto
node_pool_scores -->|"supplies confidences"| node_naive
node_pareto -->|"records selections"| node_selection_table
node_naive -->|"records selections"| node_selection_table
node_selection_table -->|"indexes pseudo-labels"| node_augmented_data
node_splits -->|"provides labeled data"| node_augmented_data
node_augmented_data -->|"supplies retraining data"| node_training
node_splits -->|"provides hold-out"| node_training
node_training -->|"returns metrics"| node_test_metrics
node_splits -->|"reports split counts"| node_dataset_report
node_test_metrics -->|"writes metrics"| node_results_json
node_selection_table -->|"writes selection diagnostics"| node_results_json
node_test_metrics -->|"writes summary"| node_results_md
node_selection_table -->|"renders diagnostics"| node_plots
node_test_metrics -->|"renders comparison"| node_plots
node_pool_scores -->|"provides embeddings"| node_plots
node_plots -->|"writes figures"| node_plot_artifacts

click node_config "https://github.com/matheustrindade2002/multiobjetglad/blob/main/config.py"
click node_splits "https://github.com/matheustrindade2002/multiobjetglad/blob/main/data_utils.py"
click node_torchdata "https://github.com/matheustrindade2002/multiobjetglad/blob/main/data_utils.py"
click node_tokenizer "https://github.com/matheustrindade2002/multiobjetglad/blob/main/modeling.py"
click node_bert "https://github.com/matheustrindade2002/multiobjetglad/blob/main/modeling.py"
click node_training "https://github.com/matheustrindade2002/multiobjetglad/blob/main/modeling.py"
click node_pool_scores "https://github.com/matheustrindade2002/multiobjetglad/blob/main/modeling.py"
click node_typicality "https://github.com/matheustrindade2002/multiobjetglad/blob/main/selection.py"
click node_robust_confidence "https://github.com/matheustrindade2002/multiobjetglad/blob/main/selection.py"
click node_pareto "https://github.com/matheustrindade2002/multiobjetglad/blob/main/selection.py"
click node_naive "https://github.com/matheustrindade2002/multiobjetglad/blob/main/selection.py"
click node_selection_table "https://github.com/matheustrindade2002/multiobjetglad/blob/main/selection.py"
click node_augmented_data "https://github.com/matheustrindade2002/multiobjetglad/blob/main/pipeline.py"
click node_test_metrics "https://github.com/matheustrindade2002/multiobjetglad/blob/main/pipeline.py"
click node_dataset_report "https://github.com/matheustrindade2002/multiobjetglad/blob/main/data/dataset_report.json"
click node_results_json "https://github.com/matheustrindade2002/multiobjetglad/blob/main/outputs/results.json"
click node_results_md "https://github.com/matheustrindade2002/multiobjetglad/blob/main/outputs/RESULTS.md"
click node_plots "https://github.com/matheustrindade2002/multiobjetglad/blob/main/plotting.py"
click node_plot_artifacts "https://github.com/matheustrindade2002/multiobjetglad/tree/main/plots"

classDef toneNeutral fill:#f8fafc,stroke:#334155,stroke-width:1.5px,color:#0f172a
classDef toneBlue fill:#dbeafe,stroke:#2563eb,stroke-width:1.5px,color:#172554
classDef toneAmber fill:#fef3c7,stroke:#d97706,stroke-width:1.5px,color:#78350f
classDef toneMint fill:#dcfce7,stroke:#16a34a,stroke-width:1.5px,color:#14532d
classDef toneRose fill:#ffe4e6,stroke:#e11d48,stroke-width:1.5px,color:#881337
classDef toneIndigo fill:#e0e7ff,stroke:#4f46e5,stroke-width:1.5px,color:#312e81
classDef toneTeal fill:#ccfbf1,stroke:#0f766e,stroke-width:1.5px,color:#134e4a
class node_config,node_splits,node_torchdata,node_augmented_data toneBlue
class node_tokenizer,node_bert,node_training,node_pool_scores,node_sst2 toneAmber
class node_typicality,node_robust_confidence,node_pareto,node_naive,node_selection_table toneMint
class node_test_metrics,node_dataset_report,node_results_json,node_results_md,node_plots,node_plot_artifacts toneRose
class node_researcher toneIndigo
```

## Components

### Data preparation

- [Experiment config](https://github.com/matheustrindade2002/multiobjetglad/blob/main/config.py): `config.py`
- [Dataset splits](https://github.com/matheustrindade2002/multiobjetglad/blob/main/data_utils.py): `data_utils.py`
- [Tokenized datasets](https://github.com/matheustrindade2002/multiobjetglad/blob/main/data_utils.py): `data_utils.py`
- [Pseudo-labeled training data](https://github.com/matheustrindade2002/multiobjetglad/blob/main/pipeline.py): `pipeline.py`

### BERT modeling

- [BERT tokenizer](https://github.com/matheustrindade2002/multiobjetglad/blob/main/modeling.py): `modeling.py`
- [BERT model](https://github.com/matheustrindade2002/multiobjetglad/blob/main/modeling.py): `modeling.py`
- [Train and evaluate](https://github.com/matheustrindade2002/multiobjetglad/blob/main/modeling.py): `modeling.py`
- [Pool predictions and embeddings](https://github.com/matheustrindade2002/multiobjetglad/blob/main/modeling.py): `modeling.py`

### Pseudo-label selection

- [kNN typicality](https://github.com/matheustrindade2002/multiobjetglad/blob/main/selection.py): `selection.py`
- [Bootstrap confidence](https://github.com/matheustrindade2002/multiobjetglad/blob/main/selection.py): `selection.py`
- [MOO Pareto selection](https://github.com/matheustrindade2002/multiobjetglad/blob/main/selection.py): `selection.py`
- [Naive top-confidence](https://github.com/matheustrindade2002/multiobjetglad/blob/main/selection.py): `selection.py`
- [Selection diagnostics](https://github.com/matheustrindade2002/multiobjetglad/blob/main/selection.py): `selection.py`

### Analysis and results

- [Hold-out metrics](https://github.com/matheustrindade2002/multiobjetglad/blob/main/pipeline.py): `pipeline.py`
- [Split report](https://github.com/matheustrindade2002/multiobjetglad/blob/main/data/dataset_report.json): `data/dataset_report.json`
- [Experiment results](https://github.com/matheustrindade2002/multiobjetglad/blob/main/outputs/results.json): `outputs/results.json`
- [Results summary](https://github.com/matheustrindade2002/multiobjetglad/blob/main/outputs/RESULTS.md): `outputs/RESULTS.md`
- [Analysis plots](https://github.com/matheustrindade2002/multiobjetglad/blob/main/plotting.py): `plotting.py`
- [Plot files](https://github.com/matheustrindade2002/multiobjetglad/tree/main/plots): `plots`

### Other components

- **Experiment runner**
- **SST-2 dataset**

## Connections

- Experiment runner → Dataset splits: runs pipeline (evidence: `README.md`)
- SST-2 dataset → Dataset splits: provides examples (evidence: `data_utils.py`)
- Experiment config → Dataset splits: configures splits (evidence: `data_utils.py`)
- Experiment config → BERT model: configures model (evidence: `modeling.py`)
- Experiment config → Analysis plots: sets output path (evidence: `plotting.py`)
- Dataset splits → Tokenized datasets: creates datasets (evidence: `data_utils.py`)
- BERT tokenizer → Tokenized datasets: tokenizes examples (evidence: `data_utils.py`)
- Tokenized datasets → Train and evaluate: supplies batches (evidence: `pipeline.py`)
- BERT model → Train and evaluate: trains and evaluates (evidence: `pipeline.py`)
- Train and evaluate → Pool predictions and embeddings: scores pool (evidence: `pipeline.py`)
- Pool predictions and embeddings → kNN typicality: provides embeddings (evidence: `pipeline.py`)
- Pool predictions and embeddings → Bootstrap confidence: provides confidences (evidence: `pipeline.py`)
- kNN typicality → Bootstrap confidence: supplies typicality (evidence: `selection.py`)
- Bootstrap confidence → MOO Pareto selection: supplies robust scores (evidence: `selection.py`)
- Pool predictions and embeddings → MOO Pareto selection: supplies predictions (evidence: `pipeline.py`)
- Pool predictions and embeddings → Naive top-confidence: supplies confidences (evidence: `pipeline.py`)
- MOO Pareto selection → Selection diagnostics: records selections (evidence: `selection.py`)
- Naive top-confidence → Selection diagnostics: records selections (evidence: `selection.py`)
- Selection diagnostics → Pseudo-labeled training data: indexes pseudo-labels (evidence: `pipeline.py`)
- Dataset splits → Pseudo-labeled training data: provides labeled data (evidence: `pipeline.py`)
- Pseudo-labeled training data → Train and evaluate: supplies retraining data (evidence: `pipeline.py`)
- Dataset splits → Train and evaluate: provides hold-out (evidence: `pipeline.py`)
- Train and evaluate → Hold-out metrics: returns metrics (evidence: `pipeline.py`)
- Dataset splits → Split report: reports split counts
- Hold-out metrics → Experiment results: writes metrics (evidence: `pipeline.py`)
- Selection diagnostics → Experiment results: writes selection diagnostics
- Hold-out metrics → Results summary: writes summary (evidence: `pipeline.py`)
- Selection diagnostics → Analysis plots: renders diagnostics (evidence: `pipeline.py`)
- Hold-out metrics → Analysis plots: renders comparison (evidence: `pipeline.py`)
- Pool predictions and embeddings → Analysis plots: provides embeddings (evidence: `pipeline.py`)
- Analysis plots → Plot files: writes figures (evidence: `plotting.py`)
