SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2403.09053 · 2024

Towards a theory of model distillation

arXiv · PDF · Open in the Atlas

Code that ran

We lifted 8 functions out of this paper's own repositories and ran 8 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.

RepositoryRoleRan
eboix/theory-of-model-distillation canonical 8 of 8
FunctionStatusWhere it lives
batch_linear_probe_adam Ran eboix/theory-of-model-distillation/probing_utils.py
pointer only (licence: NONE) · get_code("b62beacb52e18332")
batch_logistic_probe_adam Ran eboix/theory-of-model-distillation/probing_utils.py
pointer only (licence: NONE) · get_code("20fe419b3cd8dfbf")
comp_and Ran eboix/theory-of-model-distillation/decision_tree_utils.py
pointer only (licence: NONE) · get_code("01704e6e4419e521")
comp_and_list Ran eboix/theory-of-model-distillation/decision_tree_utils.py
pointer only (licence: NONE) · get_code("55004cdca28a206f")
get_random_data_unif_binary Ran eboix/theory-of-model-distillation/train_utils.py
pointer only (licence: NONE) · get_code("fcae99f48ee56066")
literals_to_tup Ran eboix/theory-of-model-distillation/decision_tree_utils.py
pointer only (licence: NONE) · get_code("39d9c79e201d3259")
test_class Ran eboix/theory-of-model-distillation/train_utils.py
pointer only (licence: NONE) · get_code("59983924736c5a8a")
validation_split Ran eboix/theory-of-model-distillation/train_utils.py
pointer only (licence: NONE) · get_code("51b5eb19649eb81d")

Repositories linked to this paper

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

Abstract

Distillation is the task of replacing a complicated machine learning model with a simpler model that approximates the original [BCNM06,HVD15]. Despite many practical applications, basic questions about the extent to which models can be distilled, and the runtime and amount of data needed to distill, remain largely open. To study these questions, we initiate a general theory of distillation, defining PAC-distillation in an analogous way to PAC-learning [Val84]. As applications of this theory: (1) we propose new algorithms to extract the knowledge stored in the trained weights of neural networks -- we show how to efficiently distill neural networks into succinct, explicit decision tree representations when possible by using the ``linear representation hypothesis''; and (2) we prove that distillation can be much cheaper than learning from scratch, and make progress on characterizing its complexity.

For agents

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

get_harvested_code_for_paper("2403.09053")
get_code_for_paper("2403.09053")
have("2403.09053")

Connect an agent — have() is free.