From f3f2f2f70748b3cf10377113e24d6d1c3b65a2d8 Mon Sep 17 00:00:00 2001 From: Peng Tan Date: Mon, 13 Nov 2023 22:26:51 +0800 Subject: [PATCH] [DOC] modify details for hetero readme --- learnware/market/heterogeneous/organizer/hetero_map/README.md | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/learnware/market/heterogeneous/organizer/hetero_map/README.md b/learnware/market/heterogeneous/organizer/hetero_map/README.md index d2ab834..73c8687 100644 --- a/learnware/market/heterogeneous/organizer/hetero_map/README.md +++ b/learnware/market/heterogeneous/organizer/hetero_map/README.md @@ -7,8 +7,8 @@ This package contains code modified from the paper "TransTab: A Flexible Transfe ## Contents - `__init__.py`: The `__init__.py` file defines the `HeteroMap` class, which forms the main network structure of the market engine. It includes methods for handling heterogeneous tabular data, focusing on mapping data from diverse feature spaces into a unified "specification world". -- `trainer.py`: The `trainer.py` file focuses on the unsupervised training process of the market engine. It defines the `TransTabCollatorForCL` class, which builds positive and negative samples from tabular vertical partitions for unsupervised learning. -- `feature_extractor.py`: This file encompasses NLP-related functionalities crucial for processing and understanding table data features. It includes classes like `WordEmbedding`, which encodes tokens from column names into word embeddings, and `FeatureTokenizer`, which prepares tables for processing by the market engine. +- `trainer.py`: The `trainer.py` file focuses on the unsupervised training process of the market engine. The `TransTabCollatorForCL` class is used for generating positive and negative samples from tabular vertical partitions for unsupervised learning. +- `feature_extractor.py`: This file is utilized for the purpose of tokenizing feature descriptions and transforming them into word embeddings. ## Handling heterogeneous learnwares