MIT 6.S191 (2020): Machine Learning for Scent

Показать описание

MIT Introduction to Deep Learning 6.S191: Lecture 10
Machine Learning for Scent
Lecturer: Alex Wiltschko (Google Brain)
January 2020

Lecture Outline
0:00 - Introduction
2:55 - Digitizing smell
4:28 - The sense of smell
10:11 - Problem setup
12:58 - Molecule fragrance dataset
16:00 - Baseline algorithms
18:14 - Graph neural networks
21:25 - Molecules to graphs
23:03 - Predicting odor descriptors
25:19 - The odor embedding space
27:58 - Molecular neighbors
30:04 - Generalization
32:34 - Explaining/interpreting predictions
36:49 - Summary and future work

Subscribe to stay up to date with new deep learning lectures at MIT, or follow us @MITDeepLearning on Twitter and Instagram to stay fully-connected!!

Рекомендации по теме

Комментарии

Not going to lie, I'm pretty excited about this lecture!

voxfox

This is gonna be like icing on the cake.

rakeshhotker

What an engaging lecture, really good.

Samuel-wlfw

Alexander, I love your uploads! <3

MachineLearningJack

I like being able to smell and all, but the ability to detect chemical composition is so very different from other perceptive abilities like detecting electromagnetic radiation & vibration.

We could potentially take advantage of the neural networks responsible for scent. Think figuring out some very binary system to enhance scent with an extra sensory ability. Perhaps one day we will be "smelling" wavelengths of light outside of our current perceptual ability. As far as my knowledge extends, the olfactory bulb is one of the two areas in the adult human brain that allows for neurogenesis- the other being the hippocampus. We need to examine scent perception through an fMRI and figure out what kind of activity humans and other animals have in response. If we have some fixed actions we may even be able to increase memory association through smell stimulus by administrating chemicals on demand to our brains.

On top of that, it may also help to look at animals with very strong abilities to smell. Perhaps aquatic mammals that more heavily depend on the ability to perceive scents that lead to danger, reproduction, and different foods

I've got plenty of ideas and no money hmu

danielsayre

It may help to look at smaller creatures in how they detect chemical composition to "smell". Like ants!

danielsayre

Anyone found the otter thing he mentioned?

ebonilha

for images, machines are very good at taking in and detecting images/pixels(RGB) en masse
how can machines detect smell in anywhere close to the same scale or accuracy? (for training purposes)

josephwong

Google took that Google Nose april fool joke too far.

muzammilaziz

MIT 6.S191 (2020): Machine Learning for Scent

MIT 6.S191 (2020): Machine Learning for Scent

MIT 6.S191 (2020): Introduction to Deep Learning

MIT 6.S191 (2020): Neurosymbolic AI

MIT 6.S191: Recurrent Neural Networks, Transformers, and Attention

Barack Obama: Intro to Deep Learning | MIT 6.S191

MIT 6.S191 (2023): Deep Generative Modeling

MIT Introduction to Deep Learning (2022) | 6.S191

MIT 6.S191 (2021): Introduction to Deep Learning

MIT 6.S191 (2020): Deep Generative Modeling

MIT Deep Learning 6.S191 Teaser

MIT 6.S191 (2020): Neural Rendering

MIT 6.S191 (2023): Reinforcement Learning

MIT 6.S191 (2023): Convolutional Neural Networks

MIT 6.S191 (2020): Deep Learning New Frontiers

MIT 6.S191 (2020): Recurrent Neural Networks

MIT 6.S191: AI Bias and Fairness

MIT 6.S191 (2020): Reinforcement Learning

MIT 6.S191 (2020): Convolutional Neural Networks

MIT 6.S191: (Google) Generative AI for Media

MIT 6.S191 (2022): Reinforcement Learning

MIT 6.S191 (2023): Deep Learning New Frontiers

MIT 6.S191: AI for Science

MIT 6.S191 (2019): Visualization for Machine Learning (Google Brain)

MIT 6.S191: Evidential Deep Learning and Uncertainty