100 likes | 446 Views
CS228: Deep Learning & Unsupervised Feature Learning. Andrew Ng. Pieter Abbeel Adam Coates Zico Kolter Ian Goodfellow Quoc Le Honglak Lee Rajat Raina Andrew Saxe. TexPoint fonts used in EMF.
E N D
CS228: Deep Learning & Unsupervised Feature Learning Andrew Ng Pieter Abbeel Adam Coates Zico Kolter Ian Goodfellow Quoc Le Honglak Lee Rajat Raina Andrew Saxe TexPoint fonts used in EMF. Read the TexPoint manual before you delete this box.: AAAAAA
How is computer perception done? Object detection Low-level vision features Image Recognition Image Grasp point Low-level features Computer vision is hard!
How is computer perception done? Object detection Recognition Vision features Image Audio classification Audio Audio features Speaker ID Text classification, MT, IR, etc. Image Grasp point Low-level features NLP Text features Text
Sensor representations Learning/AI algorithm Low-level features Input
A plethora of sensors Visible light image 3d range scan (laser scanner) Visible light image Audio Thermal Infrared Thermal Infrared 3d range scans (flash lidar) Camera array A general-purpose algorithm for good sensor representations?
Sensor representation in the brain Seeing with your tongue Human echolocation (sonar) Auditory cortex learns to see. Auditory Cortex [BrainPort; Martinez et al; Roe et al.]
Learning abstract representations object models object parts (combination of edges) edges pixels [Related work: Deep learning, Hinton, Bengio, LeCun, and others.]
Feature learning for audio Algorithm: Learned features Learned features correspond to phonemes and other “basic units” of sound.
Audio Images Galaxy Video Multimodal (audio/video) Other feature learning records: Different phone recognition task (Hinton), PASCAL VOC object classification (Yu)