SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2310.18443 · NeurIPS · 2023

Towards a fuller understanding of neurons with Clustered Compositional Explanations

La Biagio, Rosa, Leilani Gilpin, Roberto Capobianco

arXiv · PDF · Open in the Atlas

Code that ran

We lifted 29 functions out of this paper's own repositories and ran 23 of them in a sandbox. "Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results.

FunctionStatusWhere it lives
And Ran krlgroup/clustered-compositional-explanations/src/heuristic_search.py
code served (permissive licence) · get_code("6e38a1a9bf49e62e")
F Ran krlgroup/clustered-compositional-explanations/src/heuristic_search.py
code served (permissive licence) · get_code("77d55b4b5690d4e1")
Leaf Ran krlgroup/clustered-compositional-explanations/src/heuristic_search.py
code served (permissive licence) · get_code("d19ee8c373497030")
Node Ran krlgroup/clustered-compositional-explanations/src/heuristic_search.py
code served (permissive licence) · get_code("307306f38b84eb6e")
Not Ran krlgroup/clustered-compositional-explanations/src/heuristic_search.py
code served (permissive licence) · get_code("b078feb4ba8b2545")
OrderedFormula Ran krlgroup/clustered-compositional-explanations/src/heuristic_search.py
code served (permissive licence) · get_code("2eb76680fde29ee0")
UnaryNode Ran krlgroup/clustered-compositional-explanations/src/heuristic_search.py
code served (permissive licence) · get_code("5e67839d7dc4c8c7")
build_dense_label_array Ran KRLGroup/Clustered-Compositional-Explanations/src/segmentations.py
code served (permissive licence) · get_code("1ca87d6074add59d")
build_numpy_category_map Ran KRLGroup/Clustered-Compositional-Explanations/src/segmentations.py
code served (permissive licence) · get_code("c841173740b51fe3")
build_ranges_from_clusters Ran KRLGroup/Clustered-Compositional-Explanations/src/activation_utils.py
code served (permissive licence) · get_code("62e245e9fea48980")
compute_hits Ran krlgroup/clustered-compositional-explanations/src/heuristic_search.py
code served (permissive licence) · get_code("8904ea91efc13ec0")
compute_hits Ran KRLGroup/Clustered-Compositional-Explanations/src/metrics.py
code served (permissive licence) · get_code("51278ffc0e50fce0")
extract_mask Ran KRLGroup/Clustered-Compositional-Explanations/src/mask_utils.py
code served (permissive licence) · get_code("d9cbb5def766dac8")
get_mask_type Ran KRLGroup/Clustered-Compositional-Explanations/src/mask_utils.py
code served (permissive licence) · get_code("29d583b830de2e63")
get_rectangles_overlap Ran KRLGroup/Clustered-Compositional-Explanations/src/heuristics.py
code served (permissive licence) · get_code("93c52c7617a3d277")
hook Ran KRLGroup/Clustered-Compositional-Explanations/src/model_utils.py
code served (permissive licence) · get_code("51e482d6b1330aa9")
index_has_any_data Ran KRLGroup/Clustered-Compositional-Explanations/src/segmentations.py
code served (permissive licence) · get_code("130670bcfa93f2d8")
iou Ran krlgroup/clustered-compositional-explanations/src/heuristic_search.py
code served (permissive licence) · get_code("6aee6866c4385318")
load_model_from_settings Ran KRLGroup/Clustered-Compositional-Explanations/src/model_utils.py
code served (permissive licence) · get_code("084259c7a0ea2d61")
load_sparse_masks Ran KRLGroup/Clustered-Compositional-Explanations/src/mask_utils.py
code served (permissive licence) · get_code("6f526af36b5ca85d")
set_seed Ran KRLGroup/Clustered-Compositional-Explanations/src/utils.py
code served (permissive licence) · get_code("b2a19cb38c877ae0")
sparse_to_torch Ran krlgroup/clustered-compositional-explanations/src/heuristic_search.py
code served (permissive licence) · get_code("e602316fb47c122d")
torch_to_sparse Ran KRLGroup/Clustered-Compositional-Explanations/src/utils.py
code served (permissive licence) · get_code("e98d8f1b76401340")
Or Not yet run krlgroup/clustered-compositional-explanations/src/heuristic_search.py
code served (permissive licence) · get_code("e2b83fbe48e387a5")
beam_search Not yet run krlgroup/clustered-compositional-explanations/src/heuristic_search.py
code served (permissive licence) · get_code("4cbadf168b410b2a")
compute_bitmaps Not yet run KRLGroup/Clustered-Compositional-Explanations/src/activation_utils.py
code served (permissive licence) · get_code("f68205752587488c")
compute_hash_value Not yet run krlgroup/clustered-compositional-explanations/src/heuristic_search.py
code served (permissive licence) · get_code("90164d2b4896308d")
get_formula_mask Not yet run krlgroup/clustered-compositional-explanations/src/heuristic_search.py
code served (permissive licence) · get_code("e83325ca92154c23")
get_formula_str Not yet run KRLGroup/Clustered-Compositional-Explanations/src/formula.py
code served (permissive licence) · get_code("95abb69df3bd1d4e")

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

Compositional Explanations [30] is a method for identifying logical formulas of concepts that approximate the neurons' behavior. However, these explanations are linked to the small spectrum of neuron activations (i.e., the highest ones) used to check the alignment, thus lacking completeness. In this paper, we propose a generalization, called Clustered Compositional Explanations, that combines Compositional Explanations with clustering and a novel search heuristic to approximate a broader spectrum of the neuron behavior. We define and address the problems connected to the application of these methods to multiple ranges of activations, analyze the insights retrievable by using our algorithm, and propose desiderata qualities that can be used to study the explanations returned by different algorithms. Preprint. Under review.

For agents

The same record, over MCP at https://syntology.ai/mcp:

get_harvested_code_for_paper("2310.18443")
get_code_for_paper("2310.18443")
have("2310.18443")

Connect an agent — have() is free.