Index
Symbols
- ℓ² normalization, ℓ2 Normalization, Scaling Bag-of-Words with Tf-Idf Transformation, Scaling Bag-of-Words with Tf-Idf Transformation, Whitening and ZCA, k-Means Clustering
- cross validation classifier accuracy, Tuning Logistic Regression with Regularization
A
- academic paper recommender (see recommender for academic papers)
- activation functions, Rectified Linear Unit (ReLU) Transformation
- AlexNet, Learning Image Features with Deep Neural Networks
- convolutional layers, Convolutional Layers
- fully connected layers, Fully Connected Layers
- pooling layers, Pooling Layers
- ReLU transformation, Rectified Linear Unit (ReLU) Transformation
- response normalization layers, Response Normalization Layers
- structure of, Structure of AlexNet-Structure of AlexNet
- anomaly detection of time series, use of PCA, Use Cases
- approximately leakage-proof statistics, Guarding against data leakage
- ASCII, Parsing and Tokenization
- audio data, Automating the Featurizer: Image Feature Extraction and Deep Learning
B
- back-off bin, What about rare categories?
- bag-of-n-grams, Bag-of-n-Grams
- bag-of-words (BoW) featurization, Bag-of-Words
- scaling with tf-idf transformation, Scaling Bag-of-Words with Tf-Idf Transformation
- basis vectors, From Vectors to Subspaces, From Vectors to Subspaces
- bias, Overview of Linear Classification
- “Big Learning Made Easy—With Counts!” blog post, Bin Counting
- bigrams, Bag-of-n-Grams
- Bilenko, Misha, Bin Counting
- bin counting, Categorical Variables: Counting Eggs in the Age of Robotic Chickens, Bin Counting-Counts without bounds
- counts without bounds, Counts without bounds
- example of, Bin Counting
- example, using data from Avazu Kaggle competition, Bin Counting
- guarding against data leakage, Guarding against data leakage
- odds ratio and log odds ratio for, Bin Counting
- one-hot encoding vs., Bin Counting
- rare categories, What about rare categories?
- trade-offs, Summary
- binarization (of counts), Binarization
- binning
- fixed-width, Fixed-width binning
- quantile, Quantile binning
- binomial distribution, Hypothesis testing for collocation extraction
- Box-Cox transforms, Power Transforms: Generalization of the Log Transform
C
- C-HOG blocks, How are neighborhoods defined? How should they cover the image?
- categorical variables, Categorical Variables: Counting Eggs in the Age of Robotic Chickens-Bibliography
- encodings, Encoding Categorical Variables-Pros and Cons of Categorical Variable Encodings
- dummy coding, Dummy Coding-Effect Coding
- effect coding, Effect Coding
- one-hot encoding, One-Hot Encoding
- pros and cons of, Pros and Cons of Categorical Variable Encodings
- large, dealing with, Dealing with Large Categorical Variables-Counts without bounds
- bin counting, Bin Counting-Counts without bounds
- feature hashing, Feature Hashing-Feature Hashing
- encodings, Encoding Categorical Variables-Pros and Cons of Categorical Variable Encodings
- chunking, Chunking and part-of-speech tagging
- class-imbalanced dataset, Creating a Classification Dataset
- classification
- using k-means featurization, k-Means Featurization for Classification-Alternative Dense Featurization
- with logistic regression, Classification with Logistic Regression-Classification with Logistic Regression
- classification dataset, creating, Creating a Classification Dataset
- clustering algorithms, Nonlinear Featurization via K-Means Model Stacking
- (see also k-means)
- collisions, Feature Hashing
- collocations, Collocation Extraction for Phrase Detection-Summary
- extracting using chunking and part-of-speech tagging, Chunking and part-of-speech tagging
- extracting using frequency-based methods, Frequency-based methods
- extracting using hypothesis testing, Hypothesis testing for collocation extraction
- n-grams vs., Collocation Extraction for Phrase Detection
- column space, Column space
- complex features, Fancy Tricks with Simple Numbers
- handcrafted, Interaction Features
- convolutional layers in neural networks, Convolutional Layers-Convolutional Layers
- convolutional filter, Convolutional Layers
- convolutional filter example, Convolutional Layers-Convolutional Layers
- intuition behind convolutions, Convolutional Layers
- convolutions, Image Gradients
- count-min sketch, What about rare categories?
- counts, Dealing with Counts-Log Transformation
- binarization, Binarization
- quantization or binning, Quantization or Binning-Log Transformation
- visualizing business review counts in Yelp dataset, Quantization or Binning
- without bounds, Counts without bounds
- CountVectorizer transformer, Bag-of-n-Grams, Scaling Bag-of-Words with Tf-Idf Transformation
- covariance between random variables (in PCA), Variance and Empirical Variance
D
- data
- about, Data
- answering questions with, Tasks
- importing in recommender for academic papers (example), Academic Paper Recommender: Naive Approach
- data leakage
- guarding against in bin counting, Guarding against data leakage
- in k-means feature classification, Pros, Cons, and Gotchas
- data matrix, Deep Dive: What Is Happening?-Deep Dive: What Is Happening?, Overview of Linear Classification
- data space, Scalars, Vectors, and Spaces
- feature vectors in, Bag-of-Words
- vs. feature space, ℓ2 Normalization
- data visualization, importance of, Log Transform in Action
- decision surface, Overview of Linear Classification
- decision tree models, Interaction Features, Feature Selection
- input feature scale and, Fancy Tricks with Simple Numbers
- use in model stacking, Pros, Cons, and Gotchas
- deep learning
- learning image features with deep neural networks, Learning Image Features with Deep Neural Networks-Structure of AlexNet
- fully connected layers, Fully Connected Layers
- use of PCA or ZCA in preprocessing, Use Cases
- learning image features with deep neural networks, Learning Image Features with Deep Neural Networks-Structure of AlexNet
- delimiters, Parsing and Tokenization
- dense featurization with k-means, Alternative Dense Featurization
- dimensionality of subspaces, From Vectors to Subspaces
- dimensionality reduction, Dimensionality Reduction: Squashing the Data Pancake with PCA
- (see also PCA)
- nonlinear, Nonlinear Featurization via K-Means Model Stacking
- distance, Intuition
- distribution, Fancy Tricks with Simple Numbers
- document frequency, Scaling Bag-of-Words with Tf-Idf Transformation
- document-term matrix, Deep Dive: What Is Happening?
- dummy coding, Dummy Coding-Effect Coding
- pros and cons of, Pros and Cons of Categorical Variable Encodings
E
- effect coding, Effect Coding
- pros and cons of, Pros and Cons of Categorical Variable Encodings
- eigen decomposition of a matrix, Principal Components: Matrix-Vector Formulation
- eigenvectors, Singular Value Decomposition (SVD)
- embedded methods, Feature Selection
- empirical variance, Variance and Empirical Variance
- encodings
- categorical variables, Encoding Categorical Variables-Pros and Cons of Categorical Variable Encodings
- dummy coding, Dummy Coding-Effect Coding
- effect coding, Effect Coding
- one-hot encoding, One-Hot Encoding
- pros and cons of, Pros and Cons of Categorical Variable Encodings
- string objects, Parsing and Tokenization
- categorical variables, Encoding Categorical Variables-Pros and Cons of Categorical Variable Encodings
- Euclidean distance, k-Means Clustering
- Euclidean norm, ℓ2 Normalization
F
- factor analysis, Use Cases
- feature engineering, Fancy Tricks with Simple Numbers
- (see also numeric data; text data)
- defined, Features
- in machine learning workflow, Model Evaluation
- numeric data, Fancy Tricks with Simple Numbers
- feature extraction, The Simplest Image Features (and Why They Don’t Work)
- (see also image feature extraction)
- automatic, in deep learning, Automating the Featurizer: Image Feature Extraction and Deep Learning
- nonlinear manifold feature extraction by k-means, Nonlinear Featurization via K-Means Model Stacking
- feature hashing, Feature Hashing-Feature Hashing
- for word features, Feature Hashing
- signed, Feature Hashing
- storage and interpretability tradeoffs, Feature Hashing
- trade-offs, Summary
- feature normalization, Feature Scaling or Normalization
- (see also feature scaling)
- feature scaling, Feature Scaling or Normalization-ℓ2 Normalization, The Effects of Feature Scaling: From Bag-of-Words to Tf-Idf-Summary
- min-max scaling, Min-Max Scaling
- standardization (variance scaling), Standardization (Variance Scaling)
- testing scaled vs. unscaled features, Putting It to the Test-Tuning Logistic Regression with Regularization
- classification with logistic regression, Classification with Logistic Regression-Classification with Logistic Regression
- creating classification dataset, Creating a Classification Dataset
- scaling bag-of-words with tf-idf transform, Scaling Bag-of-Words with Tf-Idf Transformation
- tuning logistic regression with regularization, Tuning Logistic Regression with Regularization-Tuning Logistic Regression with Regularization
- understanding the results, Deep Dive: What Is Happening?-Deep Dive: What Is Happening?
- tf-idf, Tf-Idf : A Simple Twist on Bag-of-Words-Putting It to the Test
- use case for, ℓ2 Normalization
- ℓ² normalization, ℓ2 Normalization
- feature selection, Feature Selection-Feature Selection
- embedded methods, Feature Selection
- filtering techniques, Feature Selection
- focus of, Solving a Linear System
- with interaction features, Interaction Features
- wrapper methods, Feature Selection
- feature space, Scalars, Vectors, and Spaces
- data space vs., ℓ2 Normalization
- text sentences in 3D feature space, Bag-of-Words
- FeatureHasher, scikit-learn, Feature Hashing
- features
- about, Features
- multiple, composed into more complex features, Fancy Tricks with Simple Numbers
- filtering, Feature Selection, Item-Based Collaborative Filtering
- (see also item-based collaborative filtering)
- for cleaner features, using text data, Filtering for Cleaner Features-Stemming
- frequency-based filtering, Frequency-Based Filtering
- stemming, Stemming
- stopwords, Stopwords
- techniques for collocation extraction, Hypothesis testing for collocation extraction
- filters
- 2D convolutional filter, Convolutional Layers-Convolutional Layers
- applying to an image, Image Gradients
- financial modeling, use of PCA in, Use Cases
- fixed-width binning, Fixed-width binning
- in recommender for academic papers (example), Academic Paper Recommender: Take 2
- part two, Academic Paper Recommender: Take 2
- in recommender for academic papers (example), Academic Paper Recommender: Take 2
- frequency-based filtering (text data), Frequency-Based Filtering
- frequent words, filtering from text, Frequent words
- fully connected layers (in neural networks), Fully Connected Layers
G
- Gaussian distribution, Fancy Tricks with Simple Numbers
- Gaussian filter, applying to an image, Convolutional Layers-Convolutional Layers
- gradient boosting tree (GBT) classifiers, k-Means Featurization for Classification, Pros, Cons, and Gotchas
- gradient orientation histograms, Gradient Orientation Histograms
- bins, considerations for, How many bins should there be? Should they span from 0°–360° (signed gradients) or 0°–180° (unsigned gradients)?
- image neighborhoods, How are neighborhoods defined? How should they cover the image?
- normalization, What kind of normalization should be done?
- weight functions for gradient magnitude, What weight functions should be used?
- grid search, Tuning Logistic Regression with Regularization
- tuning logistic regression hyperparameters with, Tuning Logistic Regression with Regularization
- GridSearchCV function, scikit-learn, Tuning Logistic Regression with Regularization
H
- hard clustering, k-Means Clustering
- hash functions, Feature Hashing
- (see also feature hashing)
- heatmap of paper recommendations, Academic Paper Recommender: Naive Approach
- heavy-tailed distribution, Log Transformation
- in text data, Rare words
- HOG (Histogram of Oriented Gradients), Manual Feature Extraction: SIFT and HOG-Learning Image Features with Deep Neural Networks
- gradient orientation histograms, Gradient Orientation Histograms-What kind of normalization should be done?
- bins, considerations for, How many bins should there be? Should they span from 0°–360° (signed gradients) or 0°–180° (unsigned gradients)?
- image neighborhoods, How are neighborhoods defined? How should they cover the image?
- normalization, What kind of normalization should be done?
- weight functions, What weight functions should be used?
- image gradients, Image Gradients-Image Gradients
- gradient orientation histograms, Gradient Orientation Histograms-What kind of normalization should be done?
- horizontal image gradients, Image Gradients
- hyperparameter tuning, Tuning Logistic Regression with Regularization
- for logistic regression with grid search, Tuning Logistic Regression with Regularization
- in comparing models, importance of, Tuning Logistic Regression with Regularization
- in k-means feature classification, k-Means Featurization for Classification
- using grid search, Tuning Logistic Regression with Regularization
- hyperparameters, Tuning Logistic Regression with Regularization
- hyperplanes, Overview of Linear Classification
- hypothesis testing, using for collocation extraction, Hypothesis testing for collocation extraction
I
- image descriptors, Manual Feature Extraction: SIFT and HOG
- image feature extraction, Automating the Featurizer: Image Feature Extraction and Deep Learning-Summary
- learning image features with deep neural networks, Learning Image Features with Deep Neural Networks-Structure of AlexNet
- convolutional layers, Convolutional Layers-Convolutional Layers
- fully connected layers, Fully Connected Layers
- pooling layers, Pooling Layers
- ReLU transformation, Rectified Linear Unit (ReLU) Transformation-Rectified Linear Unit (ReLU) Transformation
- response normalization layers, Response Normalization Layers
- structure of AlexNet, Structure of AlexNet-Structure of AlexNet
- manual feature extraction with SIFT and HOG, Manual Feature Extraction: SIFT and HOG-Learning Image Features with Deep Neural Networks
- simplest image features (why they don't work), The Simplest Image Features (and Why They Don’t Work)
- learning image features with deep neural networks, Learning Image Features with Deep Neural Networks-Structure of AlexNet
- image gradients, Image Gradients-Image Gradients
- image neighborhoods, How are neighborhoods defined? How should they cover the image?
- ImageNet Large Scale Visual Recognition Challenge (ILSVRC), Learning Image Features with Deep Neural Networks
- indices, maintaining assignments during coversions, Academic Paper Recommender: Take 2
- inner product, From Vectors to Subspaces, From Vectors to Subspaces
- interaction features, Fancy Tricks with Simple Numbers, Interaction Features-Interaction Features
- computational expense of, Interaction Features
- example of use in prediction, Interaction Features
- intercept, Dummy Coding
- intercept term, Overview of Linear Classification
- intrinsic dimensionality, Intuition
- inverse document frequency, Tf-Idf : A Simple Twist on Bag-of-Words
- item-based collaborative filtering, Item-Based Collaborative Filtering
- building an item-based recommender, Item-Based Collaborative Filtering
- filter for academic paper recommender, First Pass: Data Import, Cleaning, and Feature Parsing
- for academic paper recommender
- stage 1, building item feature matrix, Academic Paper Recommender: Naive Approach
- stage 2, searching for similar items, Academic Paper Recommender: Naive Approach
- origins of, Item-Based Collaborative Filtering
- recommendations, Academic Paper Recommender: Naive Approach, Academic Paper Recommender: Take 2, Academic Paper Recommender: Take 3
K
- k-means, k-Means Clustering-Summary
- clustering as surface tiling, Clustering as Surface Tiling-Clustering as Surface Tiling
- clustering with, k-Means Clustering
- featurization for classification, k-Means Featurization for Classification-Pros, Cons, and Gotchas
- k-nearest neighbors (kNN), k-Means Featurization for Classification, Pros, Cons, and Gotchas
- kernel, Convolutional Layers
L
- left null space, Left null space
- likelihood ratio test analysis, Hypothesis testing for collocation extraction
- algorithm for detecting common phrases with, Hypothesis testing for collocation extraction
- linear algebra
- anatomy of a matrix, The Anatomy of a Matrix
- from vectors to subspaces, From Vectors to Subspaces
- useful concepts, From Vectors to Subspaces
- fundamental subspaces of the data matrix, The Four Fundamental Subspaces of the Data Matrix-Left null space
- singular value decomposition (SVD), Singular Value Decomposition (SVD)-Singular Value Decomposition (SVD)
- tips for navigating formulas, Intuition
- useful definitions, From Vectors to Subspaces
- linear classification
- overview, Overview of Linear Classification-Overview of Linear Classification
- solving a linear system, Solving a Linear System-Solving a Linear System
- linear combination, From Vectors to Subspaces
- linear correlation, Variance and Empirical Variance
- linear dependent features, One-Hot Encoding
- linear independence, From Vectors to Subspaces
- linear operators, The Anatomy of a Matrix
- linear projection (in PCA), Linear Projection
- using to transform features, Transforming Features
- linear regression
- learned coefficients, Dummy Coding
- on categorical variable, using one-hot and dummy codes, Dummy Coding
- with effect coding, Effect Coding
- log transforms, Fancy Tricks with Simple Numbers, Log Transformation-Feature Scaling or Normalization
- generalization of, in power transforms, Power Transforms: Generalization of the Log Transform-Feature Scaling or Normalization
- using log transformed data to make predictions, Log Transform in Action-Power Transforms: Generalization of the Log Transform
- using with inverse document frequency, Tf-Idf : A Simple Twist on Bag-of-Words
- log-odds ratio for bin counting, Bin Counting
- logical functions, Fancy Tricks with Simple Numbers
- logistic regression
- classification with, Classification with Logistic Regression-Classification with Logistic Regression
- tuning logistic regression with regularization, Tuning Logistic Regression with Regularization-Tuning Logistic Regression with Regularization
- looking at model's use of features, Deep Dive: What Is Happening?-Deep Dive: What Is Happening?
- with k-means cluster features, k-Means Featurization for Classification, Pros, Cons, and Gotchas
- classification with, Classification with Logistic Regression-Classification with Logistic Regression
M
- machine learning
- progress in text analysis vs. audio and images, Automating the Featurizer: Image Feature Extraction and Deep Learning
- workflow, Model Evaluation
- magnitude of numeric data, Fancy Tricks with Simple Numbers
- manifold (nonlinear subspace), Nonlinear Featurization via K-Means Model Stacking
- manifold learning, Nonlinear Featurization via K-Means Model Stacking
- mask, Convolutional Layers
- mathematical formulas, Models
- mathematical modeling, Models
- matrices
- anatomy of, The Anatomy of a Matrix
- decomposition of, methods for, Solving a Linear System
- matrix-vector formulation, principal components, Principal Components: Matrix-Vector Formulation
- mean, Scaling Bag-of-Words with Tf-Idf Transformation
- metric (k-means), k-Means Clustering
- Microsoft Academic Graph dataset, Item-Based Collaborative Filtering
- min-max scaling, Min-Max Scaling
- caution when performing on sparse features, Standardization (Variance Scaling)
- missing data, Models
- model evaluation, Model Evaluation
- model stacking, Fancy Tricks with Simple Numbers
- about, Pros, Cons, and Gotchas
- k-means featurization, k-Means Featurization for Classification-Alternative Dense Featurization
- key intuition for, Pros, Cons, and Gotchas
- models
- about, Models
- based on space-partitioning trees, Fancy Tricks with Simple Numbers
- comparing, using hyperparameter tuning in, Tuning Logistic Regression with Regularization
- data scheme for academic paper recommender model, Academic Paper Recommender: Naive Approach
- evaluating for use with categorical variable features, Summary
- good linear model that is relatively unique, Solving a Linear System
- model-driven feature engineering, PCA as example of, Summary
N
- n-grams, Bag-of-n-Grams, Parsing and Tokenization
- collocations vs., Collocation Extraction for Phrase Detection
- computing, Bag-of-n-Grams
- natural language processing (NLP), Collocation Extraction for Phrase Detection
- neighborhoods (image), How are neighborhoods defined? How should they cover the image?
- neural networks (deep), learning image features with, Learning Image Features with Deep Neural Networks-Structure of AlexNet
- convolutional layers, Convolutional Layers-Convolutional Layers
- fully connected layers, Fully Connected Layers
- pooling layers, Pooling Layers
- ReLU transformation, Rectified Linear Unit (ReLU) Transformation-Rectified Linear Unit (ReLU) Transformation
- response normalization layers, Response Normalization Layers
- structure of AlexNet, Structure of AlexNet-Structure of AlexNet
- NLP (natural language processing), Collocation Extraction for Phrase Detection
- NLTK Python package, Stemming
- nonlinear dimensionality reduction, Nonlinear Featurization via K-Means Model Stacking
- nonlinear embedding, Nonlinear Featurization via K-Means Model Stacking
- nonlinear featurization, Nonlinear Featurization via K-Means Model Stacking
- nonlinear manifold feature extraction (k-means), Nonlinear Featurization via K-Means Model Stacking
- nonordinal values, Categorical Variables: Counting Eggs in the Age of Robotic Chickens
- normalization, Fancy Tricks with Simple Numbers
- feature, Feature Scaling or Normalization
- (see also feature scaling)
- of gradient orientation histograms, What kind of normalization should be done?
- response normalization layers in neural networks, Response Normalization Layers
- feature, Feature Scaling or Normalization
- normalization constant, ℓ2 Normalization
- null space, Null space, Solving a Linear System
- numeric data, Fancy Tricks with Simple Numbers-Summary
- counts, Dealing with Counts-Log Transformation
- binarization, Binarization
- quantization or binning, Quantization or Binning-Log Transformation
- feature scaling or normalization, Feature Scaling or Normalization-ℓ2 Normalization
- min-max scaling, Min-Max Scaling
- standardization (variance scaling), Standardization (Variance Scaling)
- ℓ² normalization, ℓ2 Normalization
- feature selection, Feature Selection-Feature Selection
- interaction features, Interaction Features-Interaction Features
- log transformation, Log Transformation-Feature Scaling or Normalization
- generalization of, in power transforms, Power Transforms: Generalization of the Log Transform-Feature Scaling or Normalization
- using log transformed data to make predictions, Log Transform in Action-Power Transforms: Generalization of the Log Transform
- scalars, vectors, and spaces, Scalars, Vectors, and Spaces-Dealing with Counts
- counts, Dealing with Counts-Log Transformation
- NumPy sparse array, converting Pandas DataFrame to, Academic Paper Recommender: Take 2
O
- odds ratio for bin counting, Bin Counting
- one-hot encoding, One-Hot Encoding, Bin Counting
- of cluster membership categorical variable, k-Means Featurization for Classification
- dense featurization vs., Alternative Dense Featurization
- pros and cons of, Pros and Cons of Categorical Variable Encodings
- trade-offs, Summary
- using scikit-learn DictVectorizer, Academic Paper Recommender: Take 3
- vs. bin counting, Bin Counting
- of cluster membership categorical variable, k-Means Featurization for Classification
- orthogonal basis, From Vectors to Subspaces
- orthogonal vectors, From Vectors to Subspaces
- orthogonality, From Vectors to Subspaces
- orthonormal basis, From Vectors to Subspaces
P
- Pandas
- computing quantiles and mapping data into quantile bins, Quantile binning
- converting Pandas DataFrame to NumPy sparse array, Academic Paper Recommender: Take 2
- maintaining index assignments, Academic Paper Recommender: Take 2
- dummy coding and one-hot encoding implementation, Dummy Coding
- using to compute n-grams, Bag-of-n-Grams
- parsing, Parsing and Tokenization
- part-of-speech (PoS) tagging, Chunking and part-of-speech tagging
- PCA (principal component analysis), Dimensionality Reduction: Squashing the Data Pancake with PCA-Bibliography
- considerations and limitations, Considerations and Limitations of PCA-Considerations and Limitations of PCA
- derivation, Derivation-PCA in Action
- implementing PCA, Implementing PCA
- linear projection, Linear Projection
- principal components, first formulation, Principal Components: First Formulation
- principal components, general solution of, General Solution of the Principal Components
- principal components, matrix-vector formulation, Principal Components: Matrix-Vector Formulation
- transforming features using linear projection, Transforming Features
- variance and empirical variance, Variance and Empirical Variance
- use cases, Use Cases
- using on scikit-learn digits dataset, PCA in Action-PCA in Action
- whitening and ZCA, Whitening and ZCA
- phrase detection, Filtering for Cleaner Features
- collocation extraction for, Collocation Extraction for Phrase Detection-Summary
- frequency-based methods, Frequency-based methods
- using chunking and part-of-speech tagging, Chunking and part-of-speech tagging
- using hypothesis testing, Hypothesis testing for collocation extraction
- collocation extraction for, Collocation Extraction for Phrase Detection-Summary
- Poisson distribution, Power Transforms: Generalization of the Log Transform
- pooling layers (in neural networks), Pooling Layers
- Porter stemmer, Stemming
- power transforms, Fancy Tricks with Simple Numbers, Power Transforms: Generalization of the Log Transform-Feature Scaling or Normalization
- principal component analysis (see PCA)
- probability plots (probplots), Power Transforms: Generalization of the Log Transform
- Pythagorean theorem, ℓ2 Normalization
- Python
- converting Pandas DataFrame to NumPy sparse array, Academic Paper Recommender: Take 2
- maintaining index assignments, Academic Paper Recommender: Take 2
- libraries for chunking and part-of-speech (PoS) tagging, Chunking and part-of-speech tagging
- using to calculate image gradients, Image Gradients
- converting Pandas DataFrame to NumPy sparse array, Academic Paper Recommender: Take 2
Q
- quantiles, Quantile binning
- quantization (or binning), Quantization or Binning-Log Transformation
- fixed-width binning, Fixed-width binning
- quantile binning, Quantile binning
- quantizing a count, Quantization or Binning
R
- R-HOG blocks, How are neighborhoods defined? How should they cover the image?
- radial basis function support vector machine (RBF SVM), k-Means Featurization for Classification, Pros, Cons, and Gotchas
- random forest classifiers, k-Means Featurization for Classification
- rank or dimensionality (subspaces), From Vectors to Subspaces, From Vectors to Subspaces
- rare categories, What about rare categories?
- rare words, filtering from text, Rare words
- raw counts, PCA and, Considerations and Limitations of PCA
- receiver operating characteristic (ROC) curves, k-Means Featurization for Classification
- recommender for academic papers (example), Back to the Feature: Building an Academic Paper Recommender-Bibliography
- first pass, data import, cleaning, and feature parsing, First Pass: Data Import, Cleaning, and Feature Parsing-Academic Paper Recommender: Naive Approach
- second pass, more engineering and smarter model, Second Pass: More Engineering and a Smarter Model-Third Pass: More Features = More Information
- fixed-width binning and dummy coding (part 2), Academic Paper Recommender: Take 2
- third pass, more features and more information, Third Pass: More Features = More Information-Academic Paper Recommender: Take 3
- rectified linear unit, Rectified Linear Unit (ReLU) Transformation
- rectified linear unit (ReLU) transformation, Rectified Linear Unit (ReLU) Transformation-Rectified Linear Unit (ReLU) Transformation
- redundant data, Models
- reference category, Dummy Coding
- effect, calculating, Effect Coding
- regularization constraints, adding to a model, Solving a Linear System
- regularization, tuning logistic regression with, Tuning Logistic Regression with Regularization-Tuning Logistic Regression with Regularization
- resampling, Tuning Logistic Regression with Regularization
- response normalization layers (in neural networks), Response Normalization Layers
- local response normalization, Response Normalization Layers
- robustness, Binarization
- row space, Row space
S
- scalars, Scalars, Vectors, and Spaces, From Vectors to Subspaces
- scale, Fancy Tricks with Simple Numbers
- scikit-learn
- CountVectorizer transformer, Bag-of-n-Grams, Scaling Bag-of-Words with Tf-Idf Transformation
- digits dataset, PCA in Action
- FeatureHasher, Feature Hashing
- GridSearchCV function, Tuning Logistic Regression with Regularization
- k-means clustering on Swiss roll, Clustering as Surface Tiling
- one-hot encoding, One-Hot Encoding
- one-hot encoding using DictVectorizer, Academic Paper Recommender: Take 3
- PCA package, Considerations and Limitations of PCA
- SciPy, stats package, Power Transforms: Generalization of the Log Transform
- sentences, analysis of, Parsing and Tokenization
- separators, Parsing and Tokenization
- SIFT (Scale Invariant Feature Transform), Manual Feature Extraction: SIFT and HOG-Learning Image Features with Deep Neural Networks
- architecture, SIFT Architecture
- gradient orientation histograms, Gradient Orientation Histograms-What kind of normalization should be done?
- bins, considerations for, How many bins should there be? Should they span from 0°–360° (signed gradients) or 0°–180° (unsigned gradients)?
- image neighborhoods, How are neighborhoods defined? How should they cover the image?
- normalization, What kind of normalization should be done?
- weight functions, What weight functions should be used?
- image gradients, Image Gradients-Image Gradients
- sigmoid function, Classification with Logistic Regression, Rectified Linear Unit (ReLU) Transformation
- signed feature hashing, Feature Hashing
- singular value decomposition (SVD) of a matrix, Derivation, Principal Components: Matrix-Vector Formulation, Singular Value Decomposition (SVD)-Singular Value Decomposition (SVD), Solving a Linear System
- computational expense of, Considerations and Limitations of PCA
- singular vectors, Singular Value Decomposition (SVD)
- space characters, Parsing and Tokenization
- spaces, Scalars, Vectors, and Spaces
- data space vs. feature space, ℓ2 Normalization
- spaCy library, Chunking and part-of-speech tagging
- span of a set of vectors, From Vectors to Subspaces
- sparse data
- caution when performing min-max scaling or standardization on, Standardization (Variance Scaling)
- example document-term matrix, Deep Dive: What Is Happening?
- spectrum (of a matrix), Considerations and Limitations of PCA, Singular Value Decomposition (SVD)
- unevenness in, Solving a Linear System
- standardization, Standardization (Variance Scaling)
- caution when performing on sparse features, Standardization (Variance Scaling)
- statistical factor model, Use Cases
- statistical modeling, Models
- stats package, Power Transforms: Generalization of the Log Transform
- stemming, Stemming
- stocks, correlation patterns in, Use Cases
- stopwords, Stopwords
- string objects, Parsing and Tokenization
- subspaces, From Vectors to Subspaces
- fundamental, of the data matrix, The Four Fundamental Subspaces of the Data Matrix-Left null space
- organizing vectors into, From Vectors to Subspaces
- Swiss roll, Nonlinear Featurization via K-Means Model Stacking
- k-means clustering on, Clustering as Surface Tiling-Clustering as Surface Tiling
T
- tanh function, Rectified Linear Unit (ReLU) Transformation
- target engineering, Fancy Tricks with Simple Numbers, Binarization
- target hints, k-means featurization with/without, k-Means Featurization for Classification
- text data, Text Data: Flattening, Filtering, and Chunking-Summary
- atoms of meaning, Atoms of Meaning: From Words to n-Grams to Phrases-Summary
- parsing and tokenization, Parsing and Tokenization
- bag-of-x, turning text into flat vectors, Bag-of-X: Turning Natural Text into Flat Vectors-Filtering for Cleaner Features
- bag-of-n-grams, Bag-of-n-Grams
- bag-of-words (BoW) featurization, Bag-of-Words
- filtering for cleaner features, Filtering for Cleaner Features-Stemming
- frequency-based filtering, Frequency-Based Filtering
- stemming, Stemming
- text analysis in machine learning, Automating the Featurizer: Image Feature Extraction and Deep Learning
- atoms of meaning, Atoms of Meaning: From Words to n-Grams to Phrases-Summary
- TextBlob library, Chunking and part-of-speech tagging
- tf-idf (term frequency-inverse document frequency), Tf-Idf : A Simple Twist on Bag-of-Words-Putting It to the Test
- scaling bag-of-words with tf-idf transform, Scaling Bag-of-Words with Tf-Idf Transformation
- using in recommender for academic papers (example), Academic Paper Recommender: Take 3
- tokenization, Bag-of-n-Grams, Parsing and Tokenization
- scikit-learn's tokenizing pattern, Scaling Bag-of-Words with Tf-Idf Transformation
- training
- fitting feature transformer on training data, Scaling Bag-of-Words with Tf-Idf Transformation
- of logistic regression classifiers with default parameters, Classification with Logistic Regression
- tree-based models, Fancy Tricks with Simple Numbers
- (see also decision tree models)
- input feature scale and, Feature Scaling or Normalization
U
- Unicode, Parsing and Tokenization
- uniform hash functions, Feature Hashing
- unigrams, Bag-of-n-Grams
V
- variance, Scaling Bag-of-Words with Tf-Idf Transformation
- and empirical variance in PCA, Variance and Empirical Variance
- estimating via resampling, Tuning Logistic Regression with Regularization
- in a blob of data points, Intuition
- variance scaling, Standardization (Variance Scaling)
- variance-stabilizing transformations, Power Transforms: Generalization of the Log Transform
- (see also log transforms; power transforms)
- vector quantization, Clustering as Surface Tiling
- vector spaces, Scalars, Vectors, and Spaces
- vectors, Scalars, Vectors, and Spaces, From Vectors to Subspaces
- Euclidean distance between, k-Means Clustering
- from vectors to subspaces, From Vectors to Subspaces
- turning text into flat vectors with bag-of-words (BoW), Bag-of-Words
- vertical image gradients, Image Gradients
- volume, Intuition
W
- weighting schemes, gradient magnitude, What weight functions should be used?
- whitening, Whitening and ZCA
- word features, feature hashing for, Feature Hashing
- wrapper methods, Feature Selection
- wrong data, Models
Z
- ZCA (zero-phase component analysis), Whitening and ZCA
- use cases, Use Cases
..................Content has been hidden....................
You can't read the all page of ebook, please click here login for view all page.