ExplainableAI.jl
ExplainableAI.jl copied to clipboard
Julia-XAI
Explainable AI in Julia.
[`Flux.Parallel` docs](https://fluxml.ai/Flux.jl/stable/models/layers/#Flux.Parallel).
This will be the first step towards #10 by allowing nested model structures. This requires the following changes: - [ ] new internal representation of rules, e.g. via `LRPRulesChain` /...
A visualisation method such as `plot_text_heatmap` from [this iNNvestigate notebook](https://github.com/albermax/innvestigate/blob/master/examples/sentiment_analysis.ipynb) would be useful for language tasks such as sentiment analysis.
Add a `Shapley` analyzer when ShapML is loaded. [LazyModules.jl](https://github.com/johnnychen94/LazyModules.jl) could be used for this purpose.
for more performant use of Tullio.jl.
Add the `Threads.@threads` macro to the for-loop in `gradients_wrt_batch`: https://github.com/adrhill/ExplainableAI.jl/blob/f1b89ab9c784ff2a86de9997efa904580b2af6dd/src/gradient.jl#L5-L16
GPU support is currently untested. In theory, GPU tests could be run on CI using the [JuliaGPU Buildkite CI](https://github.com/JuliaGPU/buildkite). Locally, a first test of GPU support can be run by...
This issue keeps track of which [Flux layers in the model reference](https://fluxml.ai/Flux.jl/stable/models/layers/) got LRP implementations. ___ #### Basic layers - [x] `Dense` - [x] `flatten` #### Convolution - [x] `Conv`...