Unsupervised Learning Algorithms Examples
Formal Metadata
Title 
Unsupervised Learning Algorithms Examples

Title of Series  
Part Number 
2

Number of Parts 
2

Author 

License 
No Open Access License:
German copyright law applies. This film may be used for your own use but it may not be distributed via the internet or passed on to external parties. 
Identifiers 

Publisher 

Release Date 
2019

Language 
English

Production Year 
2019

Production Place 
Dubrovnik, Croatia

Content Metadata
Subject Area  
Keywords  machine learning 
00:00
Laptop
Boss Corporation
Time zone
Unsupervised learning
Link (knot theory)
Algorithm
Virtual machine
Port scanner
Form (programming)
00:26
CAN bus
Data model
Type theory
Unsupervised learning
Performance appraisal
Slide rule
Algorithm
Texture mapping
Port scanner
Videoconferencing
Form (programming)
Data type
00:46
Cluster sampling
System call
Outlier
Likelihoodratio test
Variance
Connected space
Mathematics
Type theory
Different (Kate Ryan album)
Object (grammar)
Core dump
Circle
Noise
Social class
Area
Algorithm
View (database)
Building
Point (geometry)
Parameter (computer programming)
Instance (computer science)
Port scanner
Connected space
Message passing
Process (computing)
Uniform resource name
Order (biology)
Hill differential equation
Right angle
Video game console
Resultant
Spacetime
Row (database)
Laptop
Point (geometry)
Metre
Dataflow
Neighbourhood (graph theory)
Algorithm
Calculation
Gene cluster
Maxima and minima
Moisture
Graph coloring
Hypercube
Number
Sound effect
Chain
Goodness of fit
Unsupervised learning
Population density
Radius
Ideal (ethics)
Form (programming)
Data type
Noise (electronics)
Distribution (mathematics)
Dialect
Axiom of choice
Slide rule
Forcing (mathematics)
Neighbourhood (graph theory)
Core dump
Shape (magazine)
Inequality (mathematics)
CAN bus
Number
Population density
Film editing
Personal digital assistant
Game theory
Object (grammar)
Form (programming)
09:46
Point (geometry)
Metre
Distribution (mathematics)
Algorithm
Distribution (mathematics)
Sampling (statistics)
Parameter (computer programming)
Basis <Mathematik>
Port scanner
Rule of inference
Hypercube
Power (physics)
Type theory
Arithmetic mean
Benz plane
Population density
Circle
Spacetime
11:28
Rule of inference
Modal logic
Slide rule
Algorithm
View (database)
Multiplication sign
Range (statistics)
ACID
Port scanner
Mereology
Open set
Unsupervised learning
Performance appraisal
Videoconferencing
Local ring
Form (programming)
Data type
11:54
Pattern recognition
Building
Group action
Linear regression
Decision theory
Outlier
Source code
Execution unit
1 (number)
Set (mathematics)
Mereology
Variable (mathematics)
Data model
Labour Party (Malta)
Semiconductor memory
Core dump
Endliche Modelltheorie
Determinant
Linear regression
Attribute grammar
Variable (mathematics)
Flow separation
Product (business)
Cognition
Type theory
Data stream
Mechatronics
Prediction
Coefficient of determination
Selforganization
Right angle
Volume
Resultant
Point (geometry)
Metre
Optical character recognition
Functional (mathematics)
Identifiability
Support vector machine
Computergenerated imagery
Virtual machine
Decision tree learning
Rule of inference
Twitter
Wave packet
Number
Latent heat
Causality
Musical ensemble
Linear map
Distribution (mathematics)
Computer network
Line (geometry)
Continuous function
Cartesian coordinate system
System call
Software
Network topology
Social class
Musical ensemble
Object (grammar)
17:52
Point (geometry)
Slide rule
Random number
Algorithm
INTEGRAL
Multiplication sign
Decision theory
Tournament (medieval)
Source code
Motion capture
Division (mathematics)
Set (mathematics)
Decision tree learning
Mereology
Hypercube
Wave packet
Attribute grammar
Heegaard splitting
Medical imaging
Vector space
Spacetime
Pairwise comparison
Social class
Area
Decision theory
Parameter (computer programming)
Computer network
Line (geometry)
Flow separation
Forest
Right angle
Quicksort
Resultant
Spacetime
20:47
Sample (statistics)
Decision theory
Decision theory
Network topology
Touch typing
Decision tree learning
19 (number)
21:35
Point (geometry)
Pattern recognition
Functional (mathematics)
Support vector machine
Line (geometry)
Decision theory
View (database)
Archaeological field survey
Virtual machine
Maxima and minima
Decision tree learning
Rule of inference
Dimensional analysis
Revision control
Performance appraisal
Profil (magazine)
Kernel (computing)
Vector graphics
Vector space
Negative number
Endliche Modelltheorie
Hyperplane
Linear map
Social class
Form (programming)
God
Area
Distribution (mathematics)
Dependent and independent variables
Polynomial
Decision theory
Planning
Division (mathematics)
Line (geometry)
Flow separation
Demoscene
Type theory
Word
Radial basis function
Function (mathematics)
Social class
Whiteboard
Object (grammar)
Boundary value problem
Marginal distribution
Resultant
Domain name
Spacetime
25:45
Point (geometry)
Machine learning
Sheaf (mathematics)
00:00
i've heard that you know most of you install the. the condo in your machines and the you have the link do they have a link to the notebooks. i go by the end zone is the is the boss or exactly just put it here and. a.
00:29
in the in the previous presentation. our k. thinks they are. and the can can maybe you just type eat.
00:44
one doesn't work so smoothly.
00:50
it. and. the. it is. and ok this is the nine year old of consolation your hand on his the euro for the notebooks.
01:25
and the past war is clear what our hands on this or. it is. the world. but. but. i. but so good. and again. weekends and gore won the very very fast or the algorithms all these two gains and the bees can now read them and they year of the key message already dismissed that we fail. with they'll the algorithm before how many klosters do we want to cash in the data said and the algorithm price tool to close the space accordingly and here is one example of data points the these are some features on the one on the x. and each point is the use of soft object to. and the first thing the first step is the enemy choose the two data point two two points in this space as a cost of centers for example if i want to have to klosters it will enrich was two points and then it would assign the nearest points to to the the to do this. center to the course when exempt are like here so that these this one isn't your to the to these them to move to another one and after the cost of space in this way the next the next step is to find the center also the cost of that will pain here and no direct bought. the to the centers and then you again the assigned the data points to the centers and you can do. you can eat or t.v. a. the repeat these ideals the cluster doesn't change any more so the distribution doesn't change any more significantly right and these ones small in their active the more how how it works and we have here before and. the colors and the the cost of center surfers randomly chosen and then to the. i. to the to the dense areas that say. so good. they said we have to speed up any questions in uk means probably urinals. except for some. so. and. yet there are some the advantages of the game means algorithm it's very efficient and finds the local people out of but the problem is that you need to pay well but in how many customers who want to care for doesn't will not find it by itself also you can of course try was different number of cluster see what comes out and. the just choose the optimal distribution of the of the space. no good. and here are some results which are which war for example working also with for example here we kept incorrect number of gloucester so if we don't have here he says that we have three klosters what think with all that in find to then it will find something radical find some two clusters which makes sense and also if the classes are. but not circles let's say you have a problem to find find correct one. these are the different the different problems of all of the one on the particular datasets. in order to overcome this problem with knowing the number of clusters in advance of bears another already been called in sick you base cost cutting and da da is that the open looks at the distribution of the data into space and the dense reasons of the reasons with a lot so for instance us. and surrounded by regions of low density the awful for this instance is our can see that this clusters and advantages is that the they are independent of the cost of form they just look for the city in the in the vicar space and the outlier you can find the outlier better than with canes because. payments will assign everyday to point to every the couple some cluster center but instead these costs and will not. in and there are some efficiency it into just like the just one scan through the data and known the thief process and high permit the ok a number of pastors the will need to define number of cluster see you in advance but there are other party meters that would have to define. and here are some basic emotions flow for these are very them. romano the. the new mission the strange so we define the the hyperbaric meter eps which is a radius with with the within the three youths are we will measure the density and the. and if we keep this reduces the small like one then every of jet in our data space and will be a separate plaster but if we have to choose these reviews to bake larch then old jets will fall into one clusters of course and there are three. we lives of low points at four point is the. is a point which is surrounded by at least new points rights of these another hyper for a metre the radio was and the number of points the way through to reach the require density and their border points the game the campaign lists then the required number of points in their us. i bought a the they live within the neighborhood of another portable or point. and there are most points neither poured more border with the if he doesn't have enough this city and there were no border the the. they've been on the most dense neighbor food then there can see that this is a moist one. from two points are directed there's a teachable you. if he is in direct the neighborhood a few and few is a is a core porn the m.p. and you are detected in situ reachable. and it in this city is reachable if there is a change of points from people you where every point is the record then city reachable forces of. and the even they can be done city the points can be density connected if the. there is a point from which you can then see to reach both of these points. and all the once actually which are which are density quit the density connected. foreman cluster two years he is one example so these. these forms like the very dense for the cluster and the once outside actually the noise points. in the algorithms a more peaceful also randomly choose the starting point find old institutional points from the starting point and in case the starting was a core point we found the cost or and if it's what the point then we continue with the next next one and repeat until. once were chosen and also year i have some the example although it didn't work so you see it's just pulls the neighborhood and once the he's done it starts the next cluster and the wrist points are can see they're just as it is a voice.
09:49
three. he and the if we come back to our power example. me.
10:03
their son.
10:08
the dollar sample of ministers and the carse and would cost it with gains than the cost or would look like like you're like circles right and the if we if we would do the density based costing than would find more even more. masters and the seeds it's also most of the most right but it requires actually other eye popper meters than the key means. what then would the exercises but at this point rules keep them. i just want to save on the basis that which we continue with i read the diesel cars in the detroit on cars on the same you think the year and the senior type and the here to dip into into space of four spots then by price. and and by by thousands of metres where each point is a one car is obviously there are certain distributions one is the news of another as a pro and i was the question what the other didn't will do with it and it makes exercises you can try it out and tried to cluster fifty and find the clusters of off. or is different cars that if it does have correctly. it's.
11:33
we will do it online because of time reasons.
11:38
question so far more some remarks. but let's let's then you go further to supervise part.
11:57
i. so we have different applications in our earth for four super was building is that aggression suffocation where we will see vote the decision tree and support the machines and your own networks the and learn about ensembles anyone not familiar with the decision tree or support that much. again. ok for them and also the. so. regression problem generally is the nearest the influence of two variables or for the the. all for entrance of the features on the target variable and the for example like how hard to predict the price of a car based on its its features. the can be used for weeks the spanish but the action of this determination and course if occasion there are some examples like much justification identify you've the particular object is on the mitchell not of fraud detection of identify the behavior is the of your specific data stream. these are normal or it's the memorial the face of cognition character recognition that these are all the old all belong to the core suffocation problem. he is one example of offered aggression the beginning to take a big one car and the type and we need a four for this type of car each point here again these is one specific objects and floated over the price and over the thousands of kilometres that the tron.
13:45
and if you can see you can absorb the trend here right that the more acute meters the car was the running the lorries the prize action. and then i can calculate the straight line and can say this is my model if i have a car which i didn't see it but they know how much it from the neck and just read the price of the car based on these omi my straight line. do you have any idea if there would be a problem with these simple model. what can go wrong. here or is it fine young it mean it. you did it is not clean year have been the model is of course the assumption is wrong which it the beginning that it should be aligned yeah ok other problem with the problems. you are. it can have your feet you just can learn the data set instead of ignoring the problem behind just go on that particular cars. the outliers have have the influence on the result the us. maybe i was absolutely right maybe i would i would name some interesting problems the is well the line is actually endless right so that means that you too ridiculous can predict for a knee point but these were the will not make sense. be on the part which was seen by the was a by the other isn't right. it's like a monkey magical function new york you can put the some some number and he will get number out but you always have to keep in mind does this number doesn't make sense or not and it's only makes a unit in their the which the organ is the source. the and also what i can see here is that the actually for the most of the performance this is just from the result result who go on and because prices not the only party meter not only feature so if you can do machine learning you should be careful about the future is that you use the dish would be. the euro. them. they have the right to choose the right futures ok your but advantage is that you can interpret the results very very easy to write you can see i predict a price for discovery causes was running and of one hundred fifty and was of the points the rules of the cars which were running one hundred fifty head. price of that old the feisty write off woods. then the it's going to cost implications and eight he and the data said that there was shown before the red ones are the usual cars here and the new ones of the petroleum cars because it's not feasible if you don't all the labour than the the yet these days. fusion is not the only visible and what the pacific nation of that in try to achieve is to find the border to separate these two distributions and it can be done on the on different ways the executive always four for one car for a new car to predict if it's in his usual petroleum and. the this separation line should work very well on the objects which agency before it's more important than to work well on the training said. so they have a is that there are different possibilities how you would actually sleep the school distributions can be revealing yearn for the optional you have borders that say many many opportunities and indeed the the different organisms. still if these in different ways and here again is more human course occasional the rooms every call him here is another all that in every ruiz of data sets an example from the early season trees by asian plus the first known of course you listening or so very very different the number of organs which makes the front of functions and.
18:11
and class those things they just face in different way and if you compare to other isms just based on their outcome problem because the fire the market they can distinguish between two classes they might appear very similar but behind it would be very easy for these people for the or for the full of the its.
18:31
look at one of the result is the and chidambaram and decisions we've found on a decision three had the same countries a political suffocation approach. and is based. it's actually the past nice to see this sort of space in fact offered three where you descend from their work down three each time you spend your source these the the in even if they mean by hundred thirty two parts. so you you half you have half of the source these with each livable for three and the integration is how to do it now for example our data said. the were even if i would be for will split the data set where he worked with the line if we can see they're just one the image of for example price the probably somewhere up somewhere like here i would say everything about the slide is like one costs everything they will this line is another plus i will kill some terrorists here but i am. i captured hear them to do most of the data the with these with these a separation and i continue these and needs now based on the this second attribute split it here and say everything to the left from these line is different from this paper says. separation the border is like the tournament he is the diesel right then the neck and continuing split it further and further and make the the try frightened actually capture for all the old the eerie us and they could make it more and more for us and the. of course here it is very dangerous because if i continue its further the i can make every data point in this separate areas was really very very fine and this is the time were able or feed my data sets so than a lot of my data set but the in the eye. can cause the fight each point in my training said correctly but you can you point comes the eye. i probably a movie you kind of friends.
20:48
and this is one example of suffocation the decision three looks like it's in his the fighting or book if you run if you can try yourself and the changeup army the unseen the how this threes constructive and i wasn't here even indeed it did three nadal been decided that.
21:08
at seventeen thousand tickets to somewhere here it was funny. the latest face and then it will have to go further but a buyer the thousands of kilometres like by nineteen will make the second season so on so it's very into a thief to every touch and all these decisions three the other in is that the it's interpreted also you can say. ok good the then it would it will descend the tree.
21:38
for for you point it was it would look the where it belongs and the belief of the three the movie the course of the pacific nations decision for its include you because you can explain exactly what on the public why here is a problem car why he is the reason car. and because it's in the particular area where is the speed speed by the three.
22:07
but the and the seventh largest disadvantages are that it's the result of the most sensitive and you can or feet if you if you decide to have a lot of notes in the street to split the them. in the space than you can more easily if another our approach is support that the first or second machine and the idea here is due to find this separating proper place in separating line in into the engines or separating couple isn't much of the mission of space to do. divide the food distributions and the there are lots of problems that you can speak in this way but in the whole thing you know in your separable problems but they're also known nino several problems like your shoulder where you will not find a single line which the which will split them. of. we are currently in you to integrate and and distributions you. if we have two decisions again there and there are many possible solutions for four separate in carrefour plane and the support machine makes them popular assumption in truth this the war of the board of support a separate up a plane which has the largest margin. the supporters for actually and sportacus did data points which other immune response to this the type of plane itself so if you if you hear them out and four he won the margin is large and for me to the margin is more than he would prefer to be one or be too although that is the rule for the high profile. they and these these type of plane would be the model which helps them to decide for unseen objects which costs of the wings to the green or to divide them. i think the scenes are left and it is a also for form only acceptable problem the support of machine has a particular solution if it's just it's another dimension. it's simple conflict and he's another them the division you can make the problem again we need a separate of rules also moved to go to a deeply here but if you are interested i would recommend you to read the boat. god is the political tricks. is one example for four different or no function so you can you can choose the airline to separate the distribution the or you can use a political will or no function the been you can you can use the word enormous year the. did it and that is a decent adjusts of wealth. the there is also will have. the every if your know which tries to find actually circles. it assumes that each each the closet useless fifteen within this or call and and again depending on the problem if your problem is like one class of problem of where the uk for couples looking for example just the ok to have one place and the negative examples of their. and the survey of view quite fine because he would find the area was example in tissue everything else is in a get them. and if. i will speak in their own the rest.
25:47
other questions at this point.