Deep Learning and Robotics

n State-of-the-art object detection until 2012: n Deep Supervised Learning (Krizhevsky, Sutskever, Hinton 2012; also LeCun, Beng ...

Many Layer Neural Network cat car dog nothing different weights àdifferent computation Neural Net Training: Find the weights tha ...

n Training / Learning: n Cars: n Cats: n Dogs: n Test time: à àLabel = Cat? Dog? Car? Neural Net Learning Image Recognition labe ...

Performance graph credit Matt Zeiler, Clarifai ...

Performance graph credit Matt Zeiler, Clarifai ...

Performance graph credit Matt Zeiler, Clarifai AlexNet ...

Performance graph credit Matt Zeiler, Clarifai AlexNet ...

Performance graph credit Matt Zeiler, Clarifai AlexNet ...

MS COCO Image Captioning Challenge Karpathy& Fei-Fei, 2015; Donahue et al., 2015; Xuet al, 2015; many more Pieter Abbeel --U ...

Visual QA Challenge Stanislaw Antol, AishwaryaAgrawal, JiasenLu, Margaret Mitchell, DhruvBatra, C. Lawrence Zitnick, Devi Parikh ...

Traditional Programming: program by writing lines of code Change in Programming Paradigm! Pieter Abbeel --UC Berkeley | Gradesco ...

Speech Recognition graph credit Matt Zeiler, Clarifai ...

Machine Translation Google Neural Machine Translation (in production) Pieter Abbeel --UC Berkeley | Gradescope| Covariant.AI ...

History Is deep learning 3 , 3 0, or 6 0 years old? 2000sSparse, Probabilistic, and Energy models (Hinton, Bengio, LeCun, Ng) Ro ...

n Data n Compute n AI innovation Why Now? Pieter Abbeel --UC Berkeley | Gradescope| Covariant.AI ...

n Deep learning successes n Supervised learning = pattern recognition, if enough data (input -> output pairs), then neural ne ...

n fast.ai n deeplearning.ai(Andrew Ng) n cs231n.stanford.edu (Andrej Karpathy et al) n fullstackdeeplearning.com(Pieter Abbeel [ ...

n Deep learning successes n Supervised learning = pattern recognition, if enough data (input -> output pairs), then neural ne ...

n Deep learning successes n Supervised learning = pattern recognition, if enough data (input -> output pairs), then neural ne ...

[Wyrobek, Berger, van der Loos, Salisbury, ICRA 2008] PR- 1 Pieter Abbeel --UC Berkeley | Gradescope| Covariant.AI ...