Published Dec 8, 2022

#85 Dr. Petar Veličković (Deepmind) - Categories, Graphs, Reasoning [NEURIPS22 UNPLUGGED]

Dr. Petar Veličković delves into the transformative effects of category theory in geometric deep learning at DeepMind, exploring its application in graph neural networks to address challenges like over-squashing and enhance algorithmic reasoning for improved data propagation and model development.

Episode Highlights

Topics covered

Episode Highlights

Related Episodes

Dr. Paul Lessard - Categorical/Structured Deep Learning
Answers 383 questions
#65 Prof. PEDRO DOMINGOS [Unplugged]
Answers 383 questions
Jürgen Schmidhuber - Neural and Non-Neural AI, Reasoning, Transformers, and LSTMs
Answers 383 questions
#82 - Dr. JOSCHA BACH - Digital Physics, DL and Consciousness [UNPLUGGED]
Answers 383 questions
#71 - ZAK JOST (Graph Neural Networks + Geometric DL) [UNPLUGGED]
Answers 383 questions
#50 Christian Szegedy - Formal Reasoning, Program Synthesis
Answers 383 questions
#91 - HATTIE ZHOU - Teaching Algorithmic Reasoning via In-context Learning #NeurIPS
Answers 383 questions
Explainability, Reasoning, Priors and GPT-3
Answers 383 questions
#83 Dr. ANDREW LAMPINEN (Deepmind) - Natural Language, Symbols and Grounding [NEURIPS2022 UNPLUGGED]
Answers 383 questions
#107 - Dr. RAPHAËL MILLIÈRE - Linguistics, Theory of Mind, Grounding
Answers 383 questions
#98 - Prof. LUCIANO FLORIDI - ChatGPT, Superintelligence, Ethics, Philosophy of Information
Answers 383 questions
#66 ALEXANDER MATTICK - [Unplugged / Community Edition]
Answers 383 questions
#106 - Prof. KARL FRISTON 3.0 - Collective Intelligence [Special Edition]
Answers 383 questions
#043 Prof J. Mark Bishop - Artificial Intelligence Is Stupid and Causal Reasoning won't fix it.
Answers 383 questions
Understanding Deep Learning - Prof. SIMON PRINCE [STAFF FAVOURITE]
Answers 383 questions

Dexa/Machine Learning Street Talk (MLST)

#85 Dr. Petar Veličković (Deepmind) - Categories, Graphs, Reasoning [NEURIPS22 UNPLUGGED]

Topics covered

Popular Clips

Graph Neural Networks

Graph Representation Learning

Graph Propagation Efficiency

Algorithmic Reasoning Insights

Categories in Deep Learning

Global Propagation Efficiency

Graph Neural Networks

Graph Representation Challenges

Algorithmic Traffic Routing

Exploring Category Theory

Graph Neural Networks

Neural Network Extrapolation

Neural Network Limitations

Future of Algorithmic Reasoning

Episode Highlights

Category Theory Impact

Graph Neural Networks

Algorithmic Reasoning

Related Episodes

Dr. Paul Lessard - Categorical/Structured Deep Learning

#65 Prof. PEDRO DOMINGOS [Unplugged]

Jürgen Schmidhuber - Neural and Non-Neural AI, Reasoning, Transformers, and LSTMs

#82 - Dr. JOSCHA BACH - Digital Physics, DL and Consciousness [UNPLUGGED]

#71 - ZAK JOST (Graph Neural Networks + Geometric DL) [UNPLUGGED]

#50 Christian Szegedy - Formal Reasoning, Program Synthesis

#91 - HATTIE ZHOU - Teaching Algorithmic Reasoning via In-context Learning #NeurIPS

Explainability, Reasoning, Priors and GPT-3

#83 Dr. ANDREW LAMPINEN (Deepmind) - Natural Language, Symbols and Grounding [NEURIPS2022 UNPLUGGED]

#107 - Dr. RAPHAËL MILLIÈRE - Linguistics, Theory of Mind, Grounding

#98 - Prof. LUCIANO FLORIDI - ChatGPT, Superintelligence, Ethics, Philosophy of Information

#66 ALEXANDER MATTICK - [Unplugged / Community Edition]

#106 - Prof. KARL FRISTON 3.0 - Collective Intelligence [Special Edition]

#043 Prof J. Mark Bishop - Artificial Intelligence Is Stupid and Causal Reasoning won't fix it.

Understanding Deep Learning - Prof. SIMON PRINCE [STAFF FAVOURITE]

#85 Dr. Petar Veličković (Deepmind) - Categories, Graphs, Reasoning [NEURIPS22 UNPLUGGED]

Topics covered

Popular Clips

Episode Highlights

Category Theory ImpactDr. Petar Veličković from DeepMind explores the transformative role of category theory in advancing geometric deep learning models. He discusses how this approach leads to practical innovations in graph neural networks, offering a new perspective on model development.

Category Theory Impact

Graph Neural Networks

Algorithmic ReasoningPetar Veličković explores the intricacies of neural network extrapolation and the future of algorithmic reasoning. He discusses the challenges and potential advancements in refining these networks to better handle real-world applications.

Algorithmic Reasoning

Related Episodes