Change Logs
2024-10-13
Initialized documentation framework: Established the overall structure for the Read the Docs documentation.
Created main sections under three categories:
Tutorials: Comprehensive guides and tutorials for users.
API Documents: Placeholder for detailed API documentation.
Notes: Additional information and project notes.
Created initial
.rstfiles for each section:Tutorials:
install.rst: Installation instructions.getting_started.rst: Quick start guide.tasks.rst: Overview of supported tasks.datasets.rst: Information on available datasets.adding_new_datasets.rst: Guide for adding new datasets.attacks.rst: Description of available attack methods.adding_new_attacks.rst: Guide for adding new attack methods.poison_dataset_wrapper.rst: Documentation for the poison dataset wrapper.models.rst: Overview of supported models.adding_new_models.rst: Guide for adding custom models.defenses.rst: Description of defense mechanisms.adding_new_defenses.rst: Guide for adding new defenses.configs.rst: Configuration file guidelines.
API Documents:
api_reference.rst: Placeholder for API reference documentation.
Notes:
scalability.rst: Information on scalability strategies.contributing.rst: Guidelines for contributing to BackdoorMBTI.changelog.rst: Changelog file to track documentation updates.
Updated index.rst: Linked all newly created
.rstfiles inindex.rstto provide a structured navigation within the documentation.
2024-10-15
Updated document framework: Corrected minor errors in the overall documentation structure.
Added content to
contributing.rst: Included a guide to help new contributors get started with BackdoorMBTI.Added content to
models.rst: Documented instructions on defining and registering custom models, along with data preprocessing transformations.
2024-10-16
Added content to
getting_started.rst: Created a quick start guide for BackdoorMBTI, including steps for downloading data, running backdoor attacks, and executing defense experiments.Added content to
tasks.rst: Documented the various tasks supported by BackdoorMBTI, including new tasks like video, audiovisual, VQA, and contrastive learning.Modified
models.rst: Updated thedef load_model(args):function to improve functionality.Updated titles: Adjusted and standardized section titles across the documentation.
Revised
getting_started.rst: Made additional updates to the content for clarity and completeness.Fixed error in
install.rst: Corrected minor errors in the installation guide.
2024-10-17
Updated
getting_started.rst: Continued making improvements to the quick start guide.Modified
tasks.rst: Added descriptions for new tasks: Contrastive Learning (CL) and Visual Question Answering (VQA).Filled content in
add_new_attacks.rst: Added a guide on how to implement new backdoor attacks, using an example similar to the image-basedBadNetattack.
2024-10-18
Updated titles in
tasks.rst: Revised section titles for improved readability.Modified
add_new_models.rstandcontributing.rst: Made formatting adjustments to ensure consistency in.rststructure.
2024-10-19
Filled content in
add_new_datasets.rst: Provided instructions on adding new datasets to BackdoorMBTI, with CIFAR-100 as an example.
2024-10-20
Filled content in
add_new_defenses.rst: Added a guide on implementing new defense methods, usingSTRIPas an example for detecting poisoned data.
2024-10-22
Updated
add_new_defenses.rst: Corrected.rstformatting issues for consistency.
2024-10-23
Filled content in
models.rst: Provided an overview of models used in the project, categorized by task type.Modified
attacks.rst: Adjusted content for improved organization and readability.
2024-10-24
Modified
attacks.rst: Made additional content adjustments for clarity.
2024-10-26
Filled content in
defenses.rst: Added a table listing the supported defenses in BackdoorMBTI.
2024-11-07
Filled content in
poison_dataset_wrapper.rst: This wrapper allows users to inject specific labels into a dataset at a controlled rate, making it especially useful for creating backdoor attacks in machine learning datasets.Filled content in
configs.rst: TheConfigsdirectory contains various configuration files and settings used for attacks, defenses, and training in the backdoor learning system.