Machine Learning
&
Neural Networks
![](https://s3.amazonaws.com/media-p.slid.es/uploads/82819/images/2448635/pasted-from-clipboard.png)
- Rishabh, Neuron
About me
- Technical Lead and Chief Deep Learning Engineer at Neuron
- Google Summer of Code Intern'14 with Mifos Initiative
- Google Summer of Code Mentor'16 with CLTK.org
![](https://s3.amazonaws.com/media-p.slid.es/uploads/82819/images/2448691/pasted-from-clipboard.png)
![](https://s3.amazonaws.com/media-p.slid.es/uploads/82819/images/2448693/pasted-from-clipboard.png)
What is Machine Learning?
...Machine Learning
Standard algorithms need a hard coded logic for input and output... On the contrarty machine learning models develops the logic using the input and output data
Input
Output
Standard algorithm has the function:
Machine Learning algorithm estimates the function
(y = f(x)) given the data:
0 : 0
1 : 2
n : 2 * n
Hypothesis: f(x) = 2 * x
Supervised and Unsupervised Learning
Supervised Learning vs. Unsupervised Learning
In Supervised Learning, we have a predefined input and output. Example - for training a sentiment analysis system, we have the text and the sentiment of each text
In Unsupervised Learning, we only have the input data and a defined action to take on them. Example - Clustering, Anomaly Detection, Dimensionality Reduction, etc.
Neural Networks
![](https://s3.amazonaws.com/media-p.slid.es/uploads/82819/images/2448655/pasted-from-clipboard.png)
...Neural Networks
a beautiful, highly flexible and generic biologically-inspired programming paradigm which enables a computer to learn from observational data
Neural Network Architecture
![](https://s3.amazonaws.com/media-p.slid.es/uploads/82819/images/2448762/pasted-from-clipboard.png)
Components
- Inputs
- Bias Inputs
- Weights - W
- Hidden layer activation
- Output layer activation
Significance of Weights(Network Parameter)
![](https://s3.amazonaws.com/media-p.slid.es/uploads/82819/images/2448788/pasted-from-clipboard.png)
Activation Functions
![](https://s3.amazonaws.com/media-p.slid.es/uploads/82819/images/2448924/pasted-from-clipboard.png)
![](https://s3.amazonaws.com/media-p.slid.es/uploads/82819/images/2448928/activation.png)
Significance of Biases
![](https://s3.amazonaws.com/media-p.slid.es/uploads/82819/images/2448784/pasted-from-clipboard.png)
Forward Propagation
![](https://s3.amazonaws.com/media-p.slid.es/uploads/82819/images/2448950/pasted-from-clipboard.png)
Back Propagation
![](https://s3.amazonaws.com/media-p.slid.es/uploads/82819/images/2448959/pasted-from-clipboard.png)
Backpropagation - Logic
![](https://s3.amazonaws.com/media-p.slid.es/uploads/82819/images/2448661/pasted-from-clipboard.png)
Loss(Cost) Function
![](https://s3.amazonaws.com/media-p.slid.es/uploads/82819/images/2448965/pasted-from-clipboard.png)
Classification - Cross Entropy Loss:
Regression - Mean Squared Error:
![](https://s3.amazonaws.com/media-p.slid.es/uploads/82819/images/2448968/pasted-from-clipboard.png)
Convex Optimization
![](https://s3.amazonaws.com/media-p.slid.es/uploads/82819/images/2448970/pasted-from-clipboard.png)
Neural Network - Equations
![](https://s3.amazonaws.com/media-p.slid.es/uploads/82819/images/2448976/pasted-from-clipboard.png)
![](https://s3.amazonaws.com/media-p.slid.es/uploads/82819/images/2448978/pasted-from-clipboard.png)
![](https://s3.amazonaws.com/media-p.slid.es/uploads/82819/images/2448979/pasted-from-clipboard.png)
Forward Propagation
Backward Propagation
Loss Function
Suggested Reading:
Neural Networks
Introduction to Theano
Theano is a Python library that allows you to define, optimize, and evaluate mathematical expressions involving multi-dimensional arrays efficiently.
- Built on top of numpy
- Symbolic Expressions
- Automatic Differentiation
- In built integration for GPU Computation
- Python Interface
- Shared Variables
Theano Syntax
import numpy
import theano
import theano.tensor as T
from theano import pp
x = T.dscalar('x')
y = x ** 2
gy = T.grad(y, x)
f = theano.function([x], gy)
f(4)
# array(8.0)
Suggested Reading:
Theano
Deep Neural Networks
![](https://s3.amazonaws.com/media-p.slid.es/uploads/82819/images/2448668/pasted-from-clipboard.png)
Suggested Reading:
Deep Learning Models
This is It.
Email: rishabh@neuronme.com
Github: https://github.com/rishy
Linkedin: https://www.linkedin.com/in/rishabhshukla1
Hellomeets-ml-nn
By Rishabh Shukla
Hellomeets-ml-nn
- 2,324