Professional Documents
Culture Documents
Convolutional Nets
Yann LeCun
NYU - Courant Institute & Center for Data Science
Facebook AI Research
“Hypernetwork”
x y
Y. LeCun
MAX
Yann LeCun
NYU - Courant Institute & Center for Data Science
Facebook AI Research
Y. LeCun
Recurrent Networks
Recurrent Networks
RNN tricks
[Pascanu, Mikolov, Bengio, ICML 2013; Bengio, Boulanger & Pascanu, ICASSP 2013]
Differential equation
Time discretization
The network updates the state
Delay
Y. LeCun
Sequence to Sequence
Representation
x x x
Of the input sequence
Y. LeCun
x x x x x
Convolutional
Networks
For sequences, audio, speech, images, volumetric
images, video, and other natural signals.
Y. LeCun
MAX
It's deep if it has more than one stage of non-linear feature transformation
Feature visualization of convolutional net trained on ImageNet from [Zeiler & Fergus 2013]
Y. LeCun
weighted thresholded
detectors sums outputs
Linear
classifier
input
image
"C"
"D"
input input
image image
Y. LeCun
Definition
convolution
In practice
Cross-correlation
In 2D
Y. LeCun
Convolution
(really: cross-correlation)
Backprop to input
Sometimes called
“back-convolution”
Backprop to weights
Y. LeCun
Regular convolution
“dense”
Stride
subsampling convolution
Reduces spatial resolution
Pooling
Filter Bank +non-linearity
Pooling
Multiple Convolutions
ReLU ( x )= max ( x , 0)
Pooling: aggregation over space or feature type
– Max, Lp norm, log prob.
p p 1 bX
MAX : Max i ( X i ) ; L p : √ X i ; PROB : log ∑ e
b ( i )
i
Y. LeCun
LeNet5
Simple ConvNet
for MNIST
[LeCun 1998]
LeNet5
Simple ConvNet
for MNIST
[LeCun 1998]
5x5 5x5
5x5 2x2 2x2
convolution convolution
convolution pooling/ pooling/
Y. LeCun
“Simple cells”
“Complex
cells”
pooling
Multiple subsampling
convolutions
Single layer Two layers FC locally connected Shared weights Shared weights
- Convolutions with stride
(subsampling)
- No separate pooling layers
Y. LeCun
First “Real” ConvNets at Bell Labs [LeCun et al 89]