# aitrics/delta-xai architecture

> Architecture diagram of the GitHub repository aitrics/delta-xai, generated by GitDiagram from its file tree, README and a sample of source files (last updated 2026-11-16). It is an AI-made overview: check details against the source.

- Interactive diagram: https://gitdiagram.com/aitrics/delta-xai
- Repository: https://github.com/aitrics/delta-xai

## Overview

Delta-XAI is a time-series explainability research framework. Its documented workflow loads a MIMIC-III dataset, initializes and loads a prediction model, selects an explainer, and computes feature attributions. The graph also preserves model training, generator-based explainers, TimeX variants, and visualization. Edges follow the README or shown source references; unsampled implementation details are not treated as verified calls.

Main entry points: [`run.py`](https://github.com/aitrics/delta-xai/blob/main/run.py) (Explanation runner), [`modeltrainer.py`](https://github.com/aitrics/delta-xai/blob/main/modeltrainer.py) (Model trainer), [`explainer/base.py`](https://github.com/aitrics/delta-xai/blob/main/explainer/base.py) (Explainer base), [`explainer/fit.py`](https://github.com/aitrics/delta-xai/blob/main/explainer/fit.py) (FIT explainer)

## Diagram (Mermaid)

```mermaid
flowchart TD

subgraph group_workflow["Experiment workflow"]
  node_runner["Explanation runner<br/>[run.py]"]
  node_config["Experiment configuration<br/>[config.py]"]
end

subgraph group_data["Data handling"]
  node_dataset["Dataset loaders<br/>[dataloader.py]"]
  node_preprocess["Data preprocessing<br/>[data_preprocess.py]"]
  node_icu["ICU mortality data<br/>[icu_mortality.py]"]
end

subgraph group_models["Predictive models"]
  node_trainer["Model trainer<br/>[modeltrainer.py]"]
  node_modelbase["Model wrappers<br/>[base.py]"]
  node_sequence["Sequence models<br/>[lstm.py]"]
  node_cnn["CNN models<br/>[cnn.py]"]
  node_transformer["Transformer models<br/>[transformer.py]"]
  node_layers["Interpretability layers<br/>[layers.py]"]
  node_generator["Feature generators<br/>[generator.py]"]
end

subgraph group_explain["Attribution methods"]
  node_explainerbase["Explainer base<br/>[base.py]"]
  node_fit["FIT explainer<br/>[fit.py]"]
  node_winit["WinIT explainer<br/>[winit.py]"]
  node_swing["SWING explainer<br/>[swing.py]"]
  node_timex["TimeX explainer<br/>[timex.py]"]
  node_timexpp["TimeX++ explainer<br/>[timexpp.py]"]
  node_methods["Other explainers"]
end

subgraph group_analysis["Results and visualization"]
  node_plot["Attribution visualization<br/>[plot.py]"]
  node_masker["Masking utilities<br/>[masker.py]"]
end

node_user(("Researcher"))

node_user -->|"configures run"| node_runner
node_runner -->|"uses settings"| node_config
node_runner -->|"loads data"| node_dataset
node_runner -->|"trains model"| node_trainer
node_trainer -->|"consumes loaders"| node_dataset
node_trainer -->|"wraps models"| node_modelbase
node_trainer -->|"selects model"| node_sequence
node_trainer -->|"selects model"| node_cnn
node_trainer -->|"selects model"| node_transformer
node_transformer -->|"uses layers"| node_layers
node_runner -->|"sets up explainer"| node_explainerbase
node_runner -->|"configures FIT"| node_fit
node_runner -->|"configures WinIT"| node_winit
node_runner -->|"configures TimeX"| node_timex
node_runner -->|"configures TimeX++"| node_timexpp
node_runner -->|"runs SWING"| node_swing
node_explainerbase -->|"explains model"| node_modelbase
node_explainerbase -->|"trains generators"| node_generator
node_fit -->|"extends base"| node_explainerbase
node_fit -->|"builds generator"| node_generator
node_runner -->|"visualizes results"| node_plot
node_runner -->|"uses masking"| node_masker

click node_runner "https://github.com/aitrics/delta-xai/blob/main/run.py"
click node_config "https://github.com/aitrics/delta-xai/blob/main/utils/config.py"
click node_dataset "https://github.com/aitrics/delta-xai/blob/main/data/dataloader.py"
click node_preprocess "https://github.com/aitrics/delta-xai/blob/main/data/data_preprocess.py"
click node_icu "https://github.com/aitrics/delta-xai/blob/main/data/icu_mortality.py"
click node_trainer "https://github.com/aitrics/delta-xai/blob/main/modeltrainer.py"
click node_modelbase "https://github.com/aitrics/delta-xai/blob/main/models/base.py"
click node_sequence "https://github.com/aitrics/delta-xai/blob/main/models/lstm.py"
click node_cnn "https://github.com/aitrics/delta-xai/blob/main/models/cnn.py"
click node_transformer "https://github.com/aitrics/delta-xai/blob/main/models/transformer.py"
click node_layers "https://github.com/aitrics/delta-xai/blob/main/models/layers.py"
click node_generator "https://github.com/aitrics/delta-xai/blob/main/models/generator.py"
click node_explainerbase "https://github.com/aitrics/delta-xai/blob/main/explainer/base.py"
click node_fit "https://github.com/aitrics/delta-xai/blob/main/explainer/fit.py"
click node_winit "https://github.com/aitrics/delta-xai/blob/main/explainer/winit.py"
click node_swing "https://github.com/aitrics/delta-xai/blob/main/explainer/swing.py"
click node_timex "https://github.com/aitrics/delta-xai/blob/main/explainer/timex.py"
click node_timexpp "https://github.com/aitrics/delta-xai/blob/main/explainer/timexpp.py"
click node_methods "https://github.com/aitrics/delta-xai/tree/main/explainer"
click node_plot "https://github.com/aitrics/delta-xai/blob/main/utils/plot.py"
click node_masker "https://github.com/aitrics/delta-xai/blob/main/utils/masker.py"

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_runner,node_config toneBlue
class node_dataset,node_preprocess,node_icu toneAmber
class node_trainer,node_modelbase,node_sequence,node_cnn,node_transformer,node_layers,node_generator toneMint
class node_explainerbase,node_fit,node_winit,node_swing,node_timex,node_timexpp,node_methods toneRose
class node_plot,node_masker,node_user toneIndigo
```

## Components

### Experiment workflow

- [Explanation runner](https://github.com/aitrics/delta-xai/blob/main/run.py): `run.py`
- [Experiment configuration](https://github.com/aitrics/delta-xai/blob/main/utils/config.py): `utils/config.py`

### Data handling

- [Dataset loaders](https://github.com/aitrics/delta-xai/blob/main/data/dataloader.py): `data/dataloader.py`
- [Data preprocessing](https://github.com/aitrics/delta-xai/blob/main/data/data_preprocess.py): `data/data_preprocess.py`
- [ICU mortality data](https://github.com/aitrics/delta-xai/blob/main/data/icu_mortality.py): `data/icu_mortality.py`

### Predictive models

- [Model trainer](https://github.com/aitrics/delta-xai/blob/main/modeltrainer.py): `modeltrainer.py`
- [Model wrappers](https://github.com/aitrics/delta-xai/blob/main/models/base.py): `models/base.py`
- [Sequence models](https://github.com/aitrics/delta-xai/blob/main/models/lstm.py): `models/lstm.py`
- [CNN models](https://github.com/aitrics/delta-xai/blob/main/models/cnn.py): `models/cnn.py`
- [Transformer models](https://github.com/aitrics/delta-xai/blob/main/models/transformer.py): `models/transformer.py`
- [Interpretability layers](https://github.com/aitrics/delta-xai/blob/main/models/layers.py): `models/layers.py`
- [Feature generators](https://github.com/aitrics/delta-xai/blob/main/models/generator.py): `models/generator.py`

### Attribution methods

- [Explainer base](https://github.com/aitrics/delta-xai/blob/main/explainer/base.py): `explainer/base.py`
- [FIT explainer](https://github.com/aitrics/delta-xai/blob/main/explainer/fit.py): `explainer/fit.py`
- [WinIT explainer](https://github.com/aitrics/delta-xai/blob/main/explainer/winit.py): `explainer/winit.py`
- [SWING explainer](https://github.com/aitrics/delta-xai/blob/main/explainer/swing.py): `explainer/swing.py`
- [TimeX explainer](https://github.com/aitrics/delta-xai/blob/main/explainer/timex.py): `explainer/timex.py`
- [TimeX++ explainer](https://github.com/aitrics/delta-xai/blob/main/explainer/timexpp.py): `explainer/timexpp.py`
- [Other explainers](https://github.com/aitrics/delta-xai/tree/main/explainer): `explainer`

### Results and visualization

- [Attribution visualization](https://github.com/aitrics/delta-xai/blob/main/utils/plot.py): `utils/plot.py`
- [Masking utilities](https://github.com/aitrics/delta-xai/blob/main/utils/masker.py): `utils/masker.py`

### Other components

- **Researcher**

## Connections

- Researcher → Explanation runner: configures run
- Explanation runner → Experiment configuration: uses settings (evidence: `run.py`)
- Explanation runner → Dataset loaders: loads data (evidence: `run.py`)
- Explanation runner → Model trainer: trains model (evidence: `run.py`)
- Model trainer → Dataset loaders: consumes loaders (evidence: `modeltrainer.py`)
- Model trainer → Model wrappers: wraps models (evidence: `modeltrainer.py`)
- Model trainer → Sequence models: selects model (evidence: `modeltrainer.py`)
- Model trainer → CNN models: selects model (evidence: `modeltrainer.py`)
- Model trainer → Transformer models: selects model (evidence: `modeltrainer.py`)
- Transformer models → Interpretability layers: uses layers (evidence: `models/transformer.py`)
- Explanation runner → Explainer base: sets up explainer (evidence: `run.py`)
- Explanation runner → FIT explainer: configures FIT (evidence: `run.py`)
- Explanation runner → WinIT explainer: configures WinIT (evidence: `run.py`)
- Explanation runner → TimeX explainer: configures TimeX (evidence: `run.py`)
- Explanation runner → TimeX++ explainer: configures TimeX++ (evidence: `run.py`)
- Explanation runner → SWING explainer: runs SWING
- Explainer base → Model wrappers: explains model (evidence: `explainer/base.py`)
- Explainer base → Feature generators: trains generators (evidence: `explainer/base.py`)
- FIT explainer → Explainer base: extends base (evidence: `explainer/fit.py`)
- FIT explainer → Feature generators: builds generator (evidence: `explainer/fit.py`)
- Explanation runner → Attribution visualization: visualizes results (evidence: `run.py`)
- Explanation runner → Masking utilities: uses masking (evidence: `run.py`)
