Beruflich Dokumente
Kultur Dokumente
Confidential, unpublished property of aiLabs. Do not duplicate or distribute. Use and distribution limited solely to authorized personnel. (c) Copyright 2018
A Journey into Deep Learning
Facets of AI Research
▪knowledge representation
▪Reasoning
▪natural language understanding
▪natural scene understanding
Facets of AI Research
▪natural speech understanding
▪problem solving
▪Perception
▪Learning
▪planning
Connotations of Learning
▪process of generalization
Connotations of Learning:
low-level parts
mid-level parts Cognitive Elements
high-level parts
additional descriptors
Sparse
Sparse Coder
Representation
Existing
Existing
training setSet
Training
Sparse
Representation Samples
Random
Sampler
Reviewer
Augmented
training set
Adaptive Prediction /
Training
Recognition
Set Learner Model
Unlabelled Predicted /
Residual Corpus Recognized Corpus
Prediction /
Recognition
Model
Human
Predicted
Reviewed
Corpus Stratified sub- Reviewer Stratified sub-
samples samples
Stratified
Sampler
Precision &
Recall
Calculator
▪ NO learning
x1
w2j n
x2 s j wij xi b j y j f (s j )
i 0
yj
wnj
xn
y 1 as x
y 0 as x
w1j
x1
w2j n
x2 s j wij xi b j y j f (s j )
i 0
yj
wnj
1
s j
1 e
xn
E yˆ n En
▪ Now differentiate to get
wi
1
2 n w yˆ
error derivatives for i n
weights
xi ,n ( yn yˆ n )
n
▪ The batch delta rule
changes the weights in
proportion to their error E
derivatives summed wi
over all training cases wi
AILABS (c) Copyright 2018 21
Evolution of Connectionist Models
1
1
x2
X1 X2 Output
0 0 0
0 1 1
1 0 1
1 1 0
(0) (0) (1)
(XOR operation) (0) x2(1)
▪ Two-pass algorithm
▪ forward propagation of activation signals from input to output
▪ backward propagation of error derivatives from output to input
…
…
…
…
…
…
xN yM
Input
Hidden Output
Layer
Layer Layer
▪ Two-pass algorithm
▪ forward propagation of activation signals from input to output
▪ backward propagation of error derivatives from output to input
Machine
“2”
Input Output
x1 y1 y1
0.1 is 1
x2 y2 y2
0.7 is 2
The image
is “2”
…
…
…
…
y10 0.2
y1 is 0
x256
16 x 16 = 256
Color → 1 Each output represents the
No color → 0 confidence of a digit.
x1 y1
x2 y2
Machine “2
…
…
…
…
”
x256 y10
…
…
…
…
…
…
…
…
…
…
xN … yM
Input
… Output
Layer Hidden Layers Layer
Patc
h
Convolution Layer
Channels Feature Maps
Tanh(x) ReLU
…
…
…
…
…
…
…
…
…
…
xN … yM
Input
… Output
Layer Hidden Layers Layer
hand-crafted “Simple”
feature Trainable output
extractor Classifier
Feature visualization of convolutional net trained on ImageNet (Zeiler and Fergus, 2013)
CiFAR10
▪ Manifold learning
▪ Reinforcement learning