Aleksander Molak: Practical graph neural networks in Python with TensorFlow and Spektral

Показать описание

Speaker:: Aleksander Molak

Track: PyData: Deep Learning
Graph neural networks (GNNs) have become one of the hottest research topics in recent years. Their popularity is reinforced by hugely successful industry applications in social networks, biology, chemistry, neuroscience and many other areas. One of the main challenges faced by data scientists and researchers who want to apply graph networks in their work is that they require different data structures and a slightly different training approach than traditional deep learning models. During the workshop we’ll demonstrate how to implement graph neural networks, how to prepare your data and – finally – how to train a GNN model for node-level and graph-level tasks using Spektral and TensorFlow.

Recorded at the PyConDE & PyData Berlin 2022 conference, April 11-13 2022.

Рекомендации по теме

Комментарии

Thanks for the tutorial. Very helpful. I only wished the tutorial can be expanded to preparing custom dataset on Spektral.

kingdomman

Thanks for the most useful video, I need some sort of help, If I want to provide my own dataset, what are steps need to follow before loading data into GCN model for label prediction.

sonyannem

Hi, I have gone through this session and nicely explained but in this 1st part of code for Node classification, looks like dataset for training and validation is same: Code :
loader_tr = SingleLoader(dataset)
loader_va = SingleLoader(dataset)

loader_tr.load()
Output: <RepeatDataset element_spec=((TensorSpec(shape=(2708, 1433), dtype=tf.float32, name=None), SparseTensorSpec(TensorShape([2708, 2708]), tf.float32)), TensorSpec(shape=(2708, 7), dtype=tf.float32, name=None))>

loader_va.load()
Output: <RepeatDataset element_spec=((TensorSpec(shape=(2708, 1433), dtype=tf.float32, name=None), SparseTensorSpec(TensorShape([2708, 2708]), tf.float32)), TensorSpec(shape=(2708, 7), dtype=tf.float32, name=None))>

Since both gives same output, but ideally nodes should be splitted accordingly. Any suggestion regarding this will be helpful.

InnovateDotAIDS

Aleksander Molak: Practical graph neural networks in Python with TensorFlow and Spektral

Aleksander Molak: Practical graph neural networks in Python with TensorFlow and Spektral

IJCNN 2021 Tutorial: Deep Learning For Graphs

Aleksander Molak: The Battle of Giants - Causality vs NLP - From Theory to Practice

[DSC Europe 21] Graph Neural Networks in Action - Avi Aminov

Graph+AI Breakout: Introduction to Graph Neural Networks using TF-GNN

Recast's Marketing Measurement Coffee-Break Session #12 - Aleksander Molak

Graph Neural Networks in Tensorflow: A Practical Guide -- Welcome

Causal discovery in Python - Aleksander Molak, Lingaro | GHOST Day: AMLC 2022

027 TensorFlow Graphs

GNN Project #3.1 - Graph-level predictions

Graph Neural Networks on Molecules with Spektral and Keras

Tracking with Graph Neural Networks

🀄Understanding Graph Convolutional Networks

Train a Graph Neural Network for Note Classification Using DGL

Democratizing Causality - Aleksander Molak

Representation Learning of Histopathology Images using Graph Neural Networks | AISC

Data Learning: Graph Neural Networks for Charged Particle Reconstruction at Large Hadron Collider

An Attempt at Demystifying Graph Deep Learning - Eric Ma | PyData Global 2021

Tutorial: Graph Neural Networks in TensorFlow: A Practical Guide

Causality Upside Down, Aleksander Molak; Causal AI Conference 2024

Dmitry Korobchenko - PyTorch Geometric for Graph Neural Nets | PyData Yerevan 2022

A. Himich - Graph Neural Networks : Code Pytorch

On Recoverability of Graph Neural Network Representation | Maxim Fishman

On the Bottleneck of Graph Neural Networks and its Practical Implications | Authors explain ML Paper