What is Data Science?

What is unsupervised learning?

Given guidelines, and here are the animals, figure it out

What else does data analytics do?

Mathematical techniques that EXPLAIN the data, determine if its relevant, predict trends, extract hidden patterns, associate patterns to operational models, recommend action items

What is regression?

Predicting patterns in terms of input and output

What is data science?

Using model and tools to do AI, ML, DL, big data, etc0

White box vs black box

We want white boxes. Can be understood and repeated. Neural networks like Chat GPT can be a black box

When is supervised learning used?

When the correct value of the target is known

When is error more important?

When we have trained with never before observation. SEPARATE test data

What is AI?

algorithms that try to do what humans do (fixed like calculator)

What is the error signal for reinforcement learning?

can be pass/fail or reward/penalty

What is data analytics?

convert raw data to meaningful and actionable information

What are the 4 types of data analytics?

descriptive (what happened), diagnostic (why did it happen), predictive (what will happen), prescriptive (what to do now)

Is supervised learning implemented online or offline?

either

How do you find the error signal?

error is the correct target value - estimated target value

A good analytical model should be:

inclusive (does not ignore important data), general (applicable to not seen before cases), unbiased (not over or under estimate), interpretable (easy to understand), reasonable (does not contradict common sense), low cost (capturing, processing, analyzing data). IGU, IRL im going under in real life

What are the 3 ways learning can be performed?

offline (using historical data), online (new as they become available), hybrid (combo of both)

How is unsupervised learning usually implemented?

offline, criteria is used

What is learning?

process through which a model adapts itself to the training patterns and eventually produces right output

What are the challenges of big data?

raw data, volume, velocity (rate), arriving from heterogeneous devices, different formats, different quality, bad data, missing data

What are the 4 types of analytics model?

regression, classification, clustering, outlier analysis

Draw Data Analytics process

see picture

What else makes a good model?

short term memory, long term memory, detest uselss and irrelevant io, use previous info to draw conclusions

what is big data?

turning raw data into information, giving confidence to data

Good training data is based on what kind of patterns?

Consistent

What is reinforcement?

Critic (teahcer) tells what is right or wrong

What is supervised learning?

Teacher tells the output, it produces the output, and the difference is the error

What is model development?

Training the model

What is Maching Learning?

Trying to do what humans do AND learning from past mistakes

What is Deep Learning?

Trying to do what humans do AND mimic human brain activity or neural networks

What is our job as data scientists?

Understand the why, and choose the best model based on the data and patterns

What do we call the test performance set?

test, validation, or hold out set

In supervised learning, training continues until what?

the error goes towards zero

In unsupervised learning, training continues until?

there are no obvious changes observed in model parameters

What do we train the model with?

training set