Fichas
machine learning chap 1
Fichas del conjunto (38)
- 1what is machine learninga computer program that is said to learn from experience with respect to some task and a performance measure if its performance continues to improve
- 2training instance vs training setinstance: one piece of data
set: group of data - 3performance measure and accuracypm: measures whether the algorithm did a good job
accuracy: performance measure thats either good data or bad data - 4data mininglooks at data and finds anomalies that are not easily detectable
- 5supervised learningtraining data has info about whether the data is good or bad
- 6classgood or bad
- 7attributedata type
- 8featureattribute + class
- 9logistic regressiontakes large sets of data predicts whether theyre right or wrong through trial and error and outputs a probability of it being right or wrong
- 10unsupervisedtraining set doesnt have information of whether its right or wrong
- 11clustering algorithmgroupss data based on similarities
- 12hierarchal clustering algorithmsubdivides each group of data into smaller groups
- 13visualiation algorithmplots similar data on a 2d or 3d graph
- 14dimentionality reductionsimplfies data by seeing which data corresponds to each other by producing similar values
- 15feature extractionmerging multiple features into one
- 16anomily detectionknows what normal data looks like, spots different data
- 17novelty detectionconsiders different data new rather than wrong
- 18association rule learninggrouping items if they go together even if they arent similar
- 19semisupervised learningsome of the training data is labeled as right or wrong, rest of it isnt
- 20reinforcement learninggives algorithm rewards and penalties based on whether it does something right or wrong(used for bots in video games)
- 21batch learningeverything is done offline, if data is updated you have to retrain the entire algorithm
- 22online learningfetch data incrementally using mini-batches
- 23out of core learningis done offline, incrementally
- 24learning ratehow fast your algorithm adapts
- 25high learning rateadapts quickly to new data but then forgets old data
- 26low learning ratelearns slower but is less sensitive to noise of new data
- 27what to do if someone spams bad dataswitch learning off when you detect a drop in performance and use a anomily detection algorithm
- 28instance based learningremembers instances by heart and detects similarities between new and old instANCes
- 29model based learningused to make predictions. choosing a specific learning algorithm based on which is the most accurate
- 30utility vs cost functionutility is how good a model is cost is bad
- 31sampling noisenonrepresentative data as a result of chance
- 32sampling biaswhen the sampling method is flawed
- 33what to do with poor quality datadiscard, fix the errors manually if they are clearly outliers
vs
if missing a few features, ignore, fill in with average, train a model with the feature and one witout it - 34feature selectionchoose which features are relevant
- 35feature extractionusing existing features to produce a new one
- 36overfitting and how to fixdata is too complex relative to the amount of noise
vs
simplify model with fewer parameters, more teraining data, fix errors manually - 37underfitting and how to fixmodel is too sipmle and doesnt have enough features
use a more powerful model
feed better features
reduce constrains of the model - 38test set vs training settraining set is the set you train the model on, after, you test the model on the test set. Never train the model with the test set
Información
- Descripción
- Fichas sobre machine learning chap 1, creado por hasdfhfsd gsdhhgsdh el 01/08/2025.
- Creado por
- hasdfhfsd gsdhhgsdh
- Visitas
- 0
- Valoración