Data cleaning for supervised learning, case study from SmartPDF AI project at Basware

Описание

Visiting lecture at University of Helsinki, 8.11.2023, Dmitry Lebedev

0:02:24 Introducing company and project
0:04:55 Theoretical part starts here
0:05:20 Problem statement
0:06:27 Exercises: removing outliers from dataset
0:14:45 Can large language models replace the need for prior knowledge, opinion of Noam Chomsky
0:22:17 The concept of model
0:31:44 The idea of universal model and 'No free lunch' theorems
0:35:23 Evolutionary programming as free lunch
0:38:38 Theory of evolution and limits of genetic programming
0:41:10 Synergetic effect and emergence in complex systems
0:46:12 Examples of evolution
0:53:40 Conclusions
0:55:55 Practical part starts here, Mosquito data refinery
1:05:16 Under the hood of Mosquito model, CFG
1:09:54 Alternatives to CFG, ML-based invoice understanding, GLOM, DINO
1:16:22 Q&A

Related:
• Debate: "Does AI Need More Innate Machinery"
• Why Does Noam Chomsky Say AI Failed?
• #51 FRANCOIS CHOLLET - Intelligence and Ge...
• Watch antibiotic resistance evolve | Scien...
• I programmed some creatures. They Evolved.
When and Why Metaheuristics Researchers Can Ignore "No Free Lunch" Theorems (https://arxiv.org/abs/1906.03280)
The Evolution of Bacteria on a “Mega-Plate” Petri Dish (https://vimeo.com/180908160)