Seven Steps of Machine Learning

Question 1

To recognize what is wine and beer

Answer 1

Features:
alcohol content
color

spectrometer - to measure the colour ( wavelength of light)
hydrometer - to measure the alcohol content

Question 2

1st Step

Answer 2

Data Collection

The quantity and quality of data would determine the result

Question 3

2nd Step

Answer 3

Data preparation

we load our data to a suitable place, randomise the order
if we load too many pictures of beer and less of wine, our model will be too much biased
so while preparing data, we do measure the quality of the data we have collected

The data for evaluation should not be same as the data given in training since then it would be more like answering questions from the homework given

Sometimes, our training data needs manipulation like correction , normalisation etc

Question 4

Choosing a model

Answer 4

Some are suited for images,
some for text
some for numerics

y=mx+b

In ML, there must me many slopes since there are various inputs

These m’s are called weights and bs are called biases

we form the matrices for wiehgts and biases

Training process involves initialising some random values of w and b
and attempting to predict the output
and then test w & b again
and update them by testing with evaulation data.

This entire cycle is called one training step

Training 80%
Evaluation 20%

Question 5

Parameter Tuning

Answer 5

How many times we showed the same training data while training
Learning rate - how fast we shift the line between each training steps
Shifting the line means -
Suppose there is a line which separates wine from beer
initially this line would be not distinctively separating wine from beer but after training, this separation becomes prominent

These parameters are called Hyper parameters
The adjustments and tuning is an experimental process
it depends on your training dataset , model and the process

Question 6

Prediction

Answer 6

Now given a alcohol percentage and color, we can determine if its beer or wine

so, rather than using human judgements, model can decide on its own

Question 7

All seven steps

Answer 7

Gathering data
Preparing that data 
Choosing a model 
Training 
Evaluation 
Hyperparameter training
Prediction

Question 8

Tensorflow

Answer 8

Browser based sandbox to try out Machine Learning