so you want to learn machine learning

and you somehow ended up here well I've

got good news and bad news the bad news

I'm not here to sell you some become an

ml engineer in 3 months fantasy could it

happen sure people also win the lottery

the good news I've been exactly where

you are about 8 years ago I thought I

knew enough Basics to just apply for

jobs and learn the rest while working

spoiler alert I failed miserably but

those failures taught me exactly what

works and what doesn't and that's what

I'm sharing today I've since taught all

of this to hundreds of students in

several countries so I think I have a

good idea on what works and what doesn't

hopefully I can save you months of

frustration by showing you the smart way

to learn machine learning learning data

science was one of the best decisions I

ever made and most of you can do it and

you will learn some cool stuff on the

way that even if you don't become a data

scientist or machine learning engineer

you will have learned programming how to

build apps how to analyze and visualize

data you will have strong statistics and

research skills and be able to

communicate data clearly

many amazing job options will be open to

you even if you don't become a data

scientist but you will have to work hard

but what should you work hard on and how

do we even start that's what I'm here to

tell you learning how to learn before we

even touch machine learning let's talk

about something crucial learning how to

learn why because machine learning and

AI like most things in Tech are

constantly evolving and what matters

isn't just what you know but how quickly

you can adapt and learn new things a

little secret I actually suck at at

programming and algorithms but I am

really good at learning new stuff here's

why this matters specifically for

machine learning technology changes fast

new platforms and Frameworks drop

constantly and new papers come out daily

what's hot today might be obsolete

tomorrow problem solving is everything

machine learning isn't about memorizing

algorithms it's about understanding data

and patterns about breaking down complex

problems and finding Creative Solutions

confidence AKA don't get overwhelmed or

scared of big problems people who know

how to learn and problem solve don't get

paralyzed when faced with a new big

problem they develop a strategy for how

to look at a new problem and break it

down into manageable problems they have

faced before they know how to look up

Solutions and find tools necessary to

solve new problems they adapt more

quickly when Tech changes efficiency if

you know how to learn you won't waste

time on unnecessary things time is money

learn what you actually need to get

where you want to be there's no one size

fits all solution for learning something

it depends on your style of learning but

also on your goals not everyone needs to

learn everything so how do you learn how

to learn this one you kind of have to

figure out for yourself because what

works for one person doesn't necessarily

work for the next some people learn well

with graphs and diagrams others with

text others maybe with voice notes some

people need to understand the theory

before applying it others need to jump

right in and use an algorithm before

asking what it actually does in this

video I will try to show you what worked

for me while giving you resources that I

believe will get you there as quickly as

possible I will just mention a principle

that has helped me a lot throughout my

career the the Paro principle sometimes

called 8020 principle it says that 80%

of the results come from 20% of the

effort constantly ask yourself why am I

doing this is this actually getting me

where I want to be or can I do something

more useful with my time well the answer

to this question isn't always the same

for everyone I will try to now give you

the 20% of the work that would have

gotten me 80% of the way to becoming a

data scientist adapt as needed but where

do I start now let's build your machine

Learning Foundation the right way

here's your road map python while the

next skill is at least as important I

would start with learning python the

main reason is that you will get a

feeling of achievement fairly quickly

and python is super simple why python

python is the main language of data

analytics data science and machine

learning while also being a full-fledged

programming language allowing you to

write scripts build apps and websites

and much more python will allow you to

actually start writing real code within

days without having to learn super

complicated computer science Concepts

like pointers memory allocation and

garbage collection also with python you

will be able to get a job as a

programmer or data analyst or web

developer even if you don't learn all

the hard machine learning stuff I

suggest you first install jupyter

notebooks as they make learning much

easier and jupyter notebooks are also a

core tool for data analysts and data

scientists all over the world then learn

about these Core Concepts programming

fundamentals basic syntax indentation

rules comments and so on variables math

if else Loops printing data types like

strings ins floats booleans lists

dictionaries functions classes and

objects modules packages and importing

do a pandas tutorial pandas is Python's

primary data manipulation Library built

for handling tabular data through data

frame objects imagine it as Excel

spreadsheets on steroids it will be your

main tool for data analysis cleaning and

transformation with powerful functions

for merging reshaping and analyzing data

the library strength lies in combining

the power of numpy arrays with

spreadsheet like functionality and SQL

database like joints it also comes with

built-in plotting functionality built on

top of Python's powerful met plot lib

libraries pandas is a true data analysis

Powerhouse if you truly Master pandas

you will excel at most data analysis

positions in the world also because

exploratory data analysis and data

preparation are about 60 to 80% of a

data scientist job it will also lay the

foundation for that 8020 principle

remember your First Data analysis

project so before you get into any

machine learning I would take the time

here to work on an actual project to

deepen your python pandas and data

analysis knowledge as I mentioned in my

previous videos real projects beat

tutorials at developing a good data

scientist find some data you want to

analyze maybe from one of your old jobs

or school maybe you can export some data

from your favorite health tracker or ask

some friends if they have some data they

want analyzed or maybe download public

data from the government the World Bank

or a nonprofit any topic you're

interested in it could be economics

Sports politics video games the board

games this last one was a passion of one

of my former students work on importing

the data into pandas clean up the data

make the units uniform decide what to do

about missing data and outliers plot the

different variables look at correlations

between variables and come up with some

hypothesis about the data and test them

by making more plots turn your results

into a slideshow with nice graphs that

tell a story that you can present to

friends and family Pro tip Jupiter

notebooks with data and plots can be

turned directly into a slideshow this

will also be the first project for your

port portfolio which you can show when

applying for jobs as a data analyst

essential math for machine learning this

might be the part that most of you fear

the most but I think it is the most

important part for anyone wanting to

learn machine learning you should take

this seriously you don't need to be a

math genius or know about all of math to

become good at machine learning but you

need to really understand the Core

Concepts from the areas I'm about to

mention for more details on math for

machine learning check out my video on

the topic basic statistics and

probability this for me is the most

important Branch as a data analyst and

data scientist now there are many online

resources for statistics but I highly

suggest taking the Con Academy

statistics and probability course this

course is completely free and is the one

I took when I prepared for my first job

as a data scientist the full course is

probably around 50 hours of content so

if you have prior math knowledge you

probably won't spend more than 100 hours

on this but it might take you longer

that's around 2 3 weeks of full-time

self-study more if much of this is

completely new to you but please take

the time to do this it will make

everything that follows so much easier

and save you much more than 100 hours of

headaches later on ideally while you

learn new Concepts here you go to your

data set from the previous data analysis

project phase and try to apply them

there to deepen your intuition linear

algebra

fundamentals while also important linear

algebra for machine learning is much

more about learning some tools and rules

this should be much quicker than

learning probability and statistics

Concepts the main thing you want to

learn is how to operate with vectors and

matrices and learn what the different

operations mean this is more about

mathem iCal tools and notations than

Concepts I think learning this will take

about a quarter to a third of the time

it took you to learn the statistics

Concepts so one or two weeks of studying

should be enough for people with prior

math

knowledge I will also leave the link to

the Khan Academy linear algebra course

in the description calculus here again

it's about learning some tools but also

understanding what derivatives are

conceptually and how they help in

optimization problems you should really

understand how functions and their

derivatives work and know the basic

rules of differentiation like the chain

rule I would again calculate with one or

two weeks if you have prior math

knowledge I will also leave a con

Academy Link in the description Pro tip

you need working knowledge not a math

PhD focus on intuition over proofs spend

most your time on statistical concepts

for linear algebra and calculus focus on

learning the tools like Matrix

operations and how to take the

derivative of a function the core

machine learning Concepts and algorithms

now here's where many people mess up

they jump straight to Deep learning but

that's a mistake in my opinion you

should spend most of your time on simple

algorithms the reasons for that are

manifold and discussed in my previous

videos but basically many problems don't

require complicated solution simple

algorithms like linear regression are

quicker to run they're more

generalizable more interpretable and

easier to learn from and communicate and

more importantly these algorithms form

the basis for the more complicated

algorithms like neural networks so truly

understanding them will help you

understand the more complicated

algorithms better too check out my video

on machine learning algorithms but

basically before getting into neural

networks or even svm make sure you

understand how linear regression and

logistic regression work then look at

decision trees and Ensemble algorithms

like random forests and gradient

boosting I learned most of these topics

from the book an introduction to

statistical learning the majority of you

will prefer learning from videos so just

watch the Youtube video series about

this book by the authors themselves

completely for free on YouTube I will

leave a link in the description

all the videos together are about 20

hours but since you will want to pause

and take notes and read up on certain

Concepts I think this will be another

100 hours or so of study time so that

should be another two or more weeks of

full-time self-study all these numbers

are estimates as everyone learns at

different speeds do a scikit learn

tutorial scit learn is the number one

machine learning library in the world

for basic machine learning algorithms

you can do a basic sklearn tutorial in a

day or two and the good thing is that

the simple and consistent syntax makes

it such that once you know how to use

the library for one algorithm you know

how to use it for any algorithm as long

as you know what algorithm is meant for

what which you now know because you just

learned it pyit learn also comes with

great documentation and toy data sets to

play around with I suggest you start

using psyit learn while you are learning

about the algorithms in the statistical

learning course the genius move while

you learn about the theory behind new

algorithms for example going through the

statistical learning course and starting

with linear regression Implement and use

the algorithm in the following three

three ways implemented from scratch

using basic python implemented using

scikit learn using a toy data

set then use both your own

implementation and the syit learn

toolkit to try out the algorithm on a

real data set that you have prepared

yourself now there's a common Pitfall

that many beginners get stuck in

tutorial hell it's where you essentially

just keep following tutorials without

striking out on your own and actually

building something most learning comes

from the trial and error of building an

application so if you always follow a

tutorial you get Stu at a basic level

how do you not get stuck there do just

one or two tutorials per area Max and

then work on a real project your first

machine learning project here you can

either continue with your data analysis

project from before or find a new more

interesting data set for a machine

learning project but don't forget to

still use pandas to do an exploratory

data analysis to prepare your data for

modeling and form hypothesis about your

data more often than not the goal of a

machine learning project will be to

predict some variable from other

variables research the industry of your

project a bit look at the data and make

some hypothesis about what might

influence your target variable either

based on intuition about the industry or

from looking at correlations and Scatter

Plots of different

variables design new features based on

your knowledge of the problem then start

modeling but start with simple

algorithms like linear regression

logistic regression and decision trees

then move on to more complex algorithms

like svm random forests or gradient

boosting note how as complexity

increases accuracy usually increases but

interpretability decreases your goal is

usually to find a sweet spot also don't

forget over fitting as you increase in

model complexity keep validation and

test sets aside before starting to model

and compare your models using the test

set at the very end many times the more

complex algorithms don't look as good

anymore once you use the final test set

it might be a good idea to work with

data sets that have been published on

sites like kaggle to then compare your

Solutions and accuracies to other

people's Solutions and get an idea of

how well your models are in comparison

to others but don't get frustrated a lot

of people on kaggle are professionals

with years of experience if you get

anywhere close in accuracy you should be

happy don't know what to work on you can

start with a tutorial but instead of

following it directly after building the

core features add some features change

some features swap out the data set and

try to break your code and then fix it

this is one of the best ways to learn

while not getting stuck in tutorial hell

collaborate and share your projects with

others learning ML and isolation is the

slowest way to learn instead find coding

buddies to work on a project with

present your work to friends and family

or post it publicly on GitHub or in

machine learning communities have

someone more advanced than you give you

feedback this will speed up your

learning 10 times don't know anyone to

work on a project with participate in a

hackathon or write to people with

similar interests on kaggle GitHub

Discord Reddit LinkedIn Etc the

connections you form this way will not

only help you learn better but boost

your career in unexpected ways check out

my most recent video to learn more about

the importance of networking and data

science Advanced topics only now should

you look at more advanced topics deep

learning architectures cnns for computer

vision RNN for sequential data or

Transformers for NLP Advanced

optimization techniques model deployment

strategies and the latest research

papers remember learn these by need not

by fomo you don't need to know

everything just learn these techniques

if they are important to your project

here some dos and don'ts don't don't get

stuck in tutorial hell don't try to

memorize everything don't learn in

isolation don't chase every new trend

don't copypaste code without

understanding don't try to learn every

new fancy tool or research paper instead

build real projects focus on

understanding share your progress join

communities Master fundamentals first

Implement from scratch learn by doing if

you found this video helpful share it

with someone who you think might also

like it and get started on one of the

tutorials in the description or on this

very Channel also consider liking the

video and subscri subcribing to be

notified about similar content in the

future thanks for watching