Please enable JavaScript.
Coggle requires JavaScript to display documents.
FUNDAMENTALS OF MACHINE LEARNING - Coggle Diagram
FUNDAMENTALS OF MACHINE LEARNING
COVERS
MACHINE LEARNING
BEYOND
CLASSIFICATION
REGRESSION
EVALUATION
PROCEDURES
MACHINE LEARNING MODELS
DATA
DEEP LEARNING
FEATURE ENGINEER
TACKLING OVERFITTING
UNIVERSAL WORKFLOW
MACHINE LEARNING PROBLEMS
INTRODUCTION
CONSOLIDATE
DATA PREPROCESSING
FEATURE ENGINEERING
MODEL EVALUATION
TACKLING OVERFITTING
SEVEN STEP WORKFLOW
SUMMARY
DEFINE
PROBLEM
DATA
TRAIN
LABELS
DETERMINE
SUCESS
WHAT METRICS
MONITOR
VALIDATION DATA
EVALUATION
PROTOCOL
K-FOLD
HOLD OUT
PORTION
DATA
VALIDATION
COLLECT
ANNOTATE
QUESTIONS
QUESTION 2
HOW
MEASURE
SUCESS
QUESTION 1
WHAT METRICS
MONITOR
DEVELOP
FIRST MODEL
PERFORM BETTER
BASIC BASELINE
MODEL
OVERFIT
REGULARIZE
MODEL
TUNE
HYPERPARAMETERS
PERFORMANCE
VALIDATION DATA
QUESTIONS
QUESTION 2
HOW
MEASURE
SUCESS
QUESTION 1
WHAT METRICS
MONITOR
MACHINE LEARNING
UNSUPERVISED LEARNING
UNDERSTAND
DATA
SELF-SUPERVISED
SUPERVISION
WITHOUT PEOPLE
LABELS
CREATED BY
INPUT DATA
EXAMPLE
AUTOENCODERS
GENERATED TARGETS
INPUTS
PREDICT
NEXT
FRAME
PREVIOUS
FRAME
WORD
PREVIOUS
VIDEO
SUPERVISED LEARNING
MAPS
INPUT DATA
TARGETS
SEQUENCE GENERATION
PICTURE
CAPTION
DESCRIBING
SERIES
CLASSIFICATION PROBLEMS
SENTENCE
DECOMPOSITION
SYNTAX TREE
OBJECT DETECTION
PICTURE
BOUNDING LINE
CERTAIN OBJECTS
CLASSIFICATION
REGRETION
IMAGE SEGMENTATION
PICTURE
PIXEL LEVEL MASK
SPECIFIC OBJECTS
REINFORCEMENT LEARNING
CLASSIFICATION AND REGRESSION TERM
CLASSIFICATION
COMMON TERMS
SAMPLE OR INPUT
DATA POINT
INSIDE MODEL
PREDICTION OR OUTPUT
OUT OF MODEL
TARGET
EXPECTED OUTPUT
PREDICTION ERROR (LOSS VALUE)
DISTANCE
MODEL'S PREDICTION
TARGET
GROUND TRUTH (ANNOTATIONS)
TARGETS
DATASETS
CLASSES
SET
LABELS
CLASSIFICATION PROBLEM
LABEL
CLASS ANNOTATION
EXAMPLE
IMAGE
DOG
ANNOTATION
DOG
BINARY CLASSIFICATION
INPUT SAMPLE
CATEGORIZED
TWO EXCLUSIVE CATEGORIES
MULTICLASS CLASSIFICATION
INPUT SAMPLE
MULTIPLE CATEGORIES
MULTILABEL CLASSIFICATION
INPUT SAMPLE
MULTIPLE LABELS
EXAMPLE
IMAGE
CAT
LABELED
CAT
DOG
DOG
REGRESSION
SCALAR REGRESSION
TASK
TARGET
CONTINUOUS SCALAR VALUE
VECTOR REGRESSION
TASK
TARGET
SET
CONTINUOUS VALUES
MINI BATCH OR BATCH
SMALL SET
SAMPLES
PROCESSED
SIMULTANEOUSLY
MODEL
MACHINE LEARNING MODEL EVALUATION
TRAINING SET
AVOID
OVERFITTING
VALIDATION SET
GOAL
REDUCE
OVERFITTING
MEASURE
PERFORMANCE
TRAINING, VALIDATION AND TEST SETS
TUNING
CHOOSE CAREFULLY
NUMBER
LAYER
HYPERPARAMETER
SIZE
USE
FEEDBACK SIGNAL
PERFORMANCE
MODEL
VALIDATION DATA
OVERFITTING
DONE TOO MUCH
HYPERPARAMETER
INFORMATION
VALIDATION DATA
LEAKS IN
ONLY ONCE
FINE
PARAMETERS
WEIGHTS
TEST DATASET
SIMPLE HOLD-OUT VALIDATION
DIVIDE
DATA
TRAIN
TEST
PROBLEM
TOO FEW
DATA
K-FOLD VALIDATION
ITERATED K-FOLD VALIDATION
VERY FEW DATA
APPLY
K-FOLD VALIDATION
EACH SHUFFLED DATA
PER PARTITION
IMPORTANT DETAILS
ARROW OF TIME
TEST SET
AFTER
TRAINING SET
NO RESHUFLE
DATA
TRAIN
FUTURE DATA
REDUNDANCY DATA
NUMBER
APPEAR TWICE
REDUNDANCY
SHUFFLING
REDUNDANCY
TRAIN DATA
SAME VALUE
TEST DATA
DATA REPRESENTATIVENESS
TRAINING SET
REPRESENTATIVE
DATA
TEST SET
EXAMPLE
ARRAY
NUMBERS
0-9
RANDOMLY
RESHUFLE
NUMBER ORDER