Data Mining
John Samuel
CPE Lyon
Year
: 2019-2020
Email
: john(dot)samuel(at)cpe(dot)fr
Data Mining
Goals
Artifical Neural Networks
Deep Learning
Reinforcement Learning
Data Licences, Ethics and Privacy
1. Artificial Neural Networks
Inspired by biological neural networks
Collection of connected nodes called artificial neurons.
Artificial neurons can transmit signal from one to another (like in a synapse).
Signal between artificial neurons is a real number
The output of a neuron is the sum of weighted inputs.
Artificial neural networks
1. Artificial Neural Networks
Perceptron
Algorithm for supervised learning of binary classifiers
Binary classifier
Artificial neural networks
1. Artificial Neural Networks
Perceptron: Formal definition
Let
y = f(z)
be output of perceptron for an input vector
z
Let
N
be the number of training examples
Let
X
be the input feature space
Let {
(x
1
, d
1
),...,(x
N
, d
N
)
} be the
N
training examples, where
x
i
is the feature vector of
i
th
training example.
d
i
is the desired output value.
x
j,i
be the
i
th
feature of
j
th
training example.
x
j,0
= 1
Artificial neural networks
1. Artificial Neural Networks
Perceptron: Formal definition
Weights are represented in the following manner:
w
i
is the
i
th
value of weight vector.
w
i
(t)
is the
i
th
value of weight vector at a given time t.
Artificial neural networks
1. Artificial Neural Networks
Perceptron: Steps
Initialize weights and threshold
For each example
(x
j
, d
j
)
in training set
Calculate the weight:
y
j
(t)=f[w(t).x
j
]
Update the weights:
w
i
(t + 1) = w
i
(t) + (d
j
-y
j
(t))x
j,i
Repeat step 2 until the iteration error
1/s (Σ |d
j
- y
j
(t)|)
is less than user-specified threshold.
1. Artificial Neural Networks
Backpropagation
Backward propagation of errors
Adjust the weight of neurons by calculating the gradient of the loss function
Error is calculated and propagated back to the network layers
2. Deep Learning
Deep neural networks
Multiple hidden layers between the input and output layers
2. Deep Learning
Applications
Computer vision
Speech recognition
Drug design
Natural language processing
Machine translation
2. Deep Learning
Convolutional deep neural networks
Analysis of images
Inspired by neurons in the virtual cortex
Network learns the filters
3. Reinforcement Learning
Inspired by behaviourist psychology
Actions to be taken in order to maximize the cumulative award.
4. Data Licences, Ethics and Privacy
Data usage licences
Confidentiality and Privacy
Ethics
4. Data Licences, Ethics and Privacy
Big Data
Volume
Variety
Velocity
Veracity
Value
4. Data Licences, Ethics and Privacy
Privacy
4. Data Licences, Ethics and Privacy
Open Data
4. Data Licences, Ethics and Privacy
Linked Open data cloud
4. Data Licences, Ethics and Privacy
Archived data
References
Online resources
Artificial Neural Network
Perceptron
References
Colors
Color Tool - Material Design
Images
Wikimedia Commons