Index

Search in book...
Toggle Font Controls
Create new playlist

Name your new playlist

Playlist description (optional)
Sign In

Email address

Password

Forgot Password?

or

Continue with Facebook

Continue with Google
Sign Up

Full Name

Email address

Confirm Email Address

Password

or

Continue with Facebook

Continue with Google

Previous Chapter

Appendix. Capturing data on the web

List of Figures

Index

[SYMBOL][A][B][C][D][E][F][G][H][I][J][K][L][M][N][O][P][Q][R][S][T][U][V][W][X][Z]

SYMBOL

2-D array 2-D combinations 2-D linear model 4-D Gaussian clustering 10-fold cross-validation

A

A/B testing code suitability of theory vs. Bayesian bandits accuracy, 2^nd activating neural networks activation function, 2^nd, 3^rd activation profiles activity trends ad creative optimization adversarial bandits advertising. See click prediction for online advertising. adview type AI (artificial intelligence) algorithms. See intelligent algorithms. all value alpha term AND function Apache Hadoop. See Hadoop framework. Apache Kafka. See Kafka log-processing platform. artificial intelligence. See AI. attributes AUC (area under a curve) automated method awk tool

B

backpropagation balanced data Bandit objects Bayes theorem, 2^nd Bayesian bandits A/B vs. factors impacting Bayesian networks Bengio, Yoshua Bernoulli Restricted Boltzmann Machine. See BRBMs. beta distribution, 2^nd bias term, 2^nd bid win (or loss) notification binary classification binomial distribution bipartite graph bootstrapping, 2^nd Bottou BRBMs (Bernoulli Restricted Boltzmann Machines) build method

C

call number campaigns capturing data on web data collection managing at scale naïve approach Kafka log-processing platform consumer groups design patterns evaluating plus Hadoop plus Storm replication in showing ads online categorical features causation, difference from correlation CF (collaborative filtering) choices A/B testing code suitability of theory vs. Bayesian bandits Bayesian bandits A/B vs. factors impacting multi-armed bandits epsilon decreasing strategy epsilon first strategy epsilon greedy strategy extensions to Cinematch algorithm classes of intelligent algorithms artificial intelligence machine learning predictive analytics classification classifiers lifecycle of statistical classification algorithms structural classification algorithms credibility of results fraud detection with logistic regression from linear to logistic regression implementing fraud detection linear regression primer need for structural algorithms with very large datasets cleverer algorithm click prediction for online advertising bidders decisioning engine building, complexities of contextual information data preparation decisioning engine model feature engineering information about placement information about user mapping predicted click-through rate to bid price model training exchange and ad monitoring ad placement bid win (or loss) notification cookie matching future of real-time prediction history and background of with Vowpal Wabbit machine-learning library model calibration preparing dataset testing model VW data format click type click-streams topic, 2^nd, 3^rd click-through rate. See CTR. cluster centroids, 2^nd clustering, 2^nd codomain collaborative filtering. See CF. colon character cols value concepts confidence interval confusion matrix consumer group, 2^nd, 3^rd, 4^th contextual bandits contextual information continuous online optimization contrastive divergence converged algorithm convergence convert type cookie ID cookie matching correct classification correlation, difference from causation cost function cost of one impression. See CPI. cost per 1,000 impressions. See CPM. cost per click. See CPC. covariance_type parameter CPC (cost per click), 2^nd CPI (cost of one impression) CPM (cost per mille), 2^nd Criteo CTR (click-through rate), 2^nd Cun, Yann Le curse of dimensionality, 2^nd Cutter number

D

dac.tar.gz file DAGs (directed acyclic graphs) data array, 2^nd data axis, transforming eigenvectors and eigenvalues principal component analysis data capture. See capturing data on web. Data class data collection managing at scale naïve approach data preparation data reliability data size data types data_categoricals array data_non_categoricals array datasets, large decision tree algorithms. See DT. decisioning engine click prediction for online advertising contextual information data preparation decisioning engine model feature engineering information about placement information about user mapping predicted click-through rate to bid price model training decisions deduplication bolts data window deep learning. See also neural networks. demand-side platform. See DSP. DESCR attribute dimensional spaces dimensionality directed acyclic graphs. See DAG. distance distance-based algorithms, 2^nd down-sampled dataset, 2^nd, 3^rd drift DSP (demand-side platform), 2^nd, 3^rd DT (decision tree) algorithms duplicate partitions durability

E

eigenvalues eigenvectors EM (expectation maximization) algorithm, 2^nd, 3^rd encoded data energy-based learning epsilon decreasing strategy epsilon first strategy epsilon greedy strategy error attribute Euclidean distance evaluation events exchange, click prediction for online advertising and ad monitoring ad placement bid win (or loss) notification cookie matching expectation maximization explore-exploit conundrum extensibility

F

f1 score feature engineering FeedForwardNetwork object field grouping filtering, user-based collaborative fit method flat geometries flat reference structures flexibility FlumeJava FNR (false negative rate) forecasting forward chaining FP rate FPR (false positive rate), 2^nd F-score, 2^nd FullConnection object future applications of intelligent web home healthcare internet of things personalized physical advertising self-driving vehicle semantic web

G

Gaussian mixture model. See GMM. Gaussian processes GDBTs (gradient-boosted decision trees) generalization generalized linear models Gerbrands Gibbs measure GLM (generalized linear model) GMM (Gaussian mixture model), 2^nd, 3^rd example of learning using Gaussian distribution general discussion k-means and gmm object Google Glass Google Now example grayscale values GUID (globally unique identifier)

H

Hadoop framework, Kafka plus hand-built neuron hashing trick HDFS (Hadoop Distributed File System) head tool healthcare hidden layer, 2^nd hidden nodes, 2^nd, 3^rd hidden units, 2^nd hierarchical reference structures higher-dimensional spaces high-frequency trading Hinton, Geoffrey home healthcare horizontal scaling Huang, Jeubin human-readable tags HW (high watermark) hyperbolic profile hyperbolic tangent function hyperplane

I

if-then clauses incorrect classification inference inhibitory input input layer, 2^nd input variables instances in-sync replicas intelligence, evaluating intelligent algorithms, 2^nd, 3^rd classes of artificial intelligence machine learning predictive analytics data reliability and data size and evaluating performance of evaluating intelligence evaluating predictions examples of generalization and Google Now example inference and lifecycle of scaling characteristics of what they are not intelligent web future applications of home healthcare internet of things personalized physical advertising self-driving vehicle semantic web social implications of internet of things invert_hash option Iris dataset IsFraud label ISRs (in-sync replicas)

J

Jaccard similarity

K

Kafka log-processing platform consumer groups design patterns evaluating plus Hadoop plus Storm replication in message acknowledgement replication, leaders, and in-sync replicas KafkaClient object keyword searches KISS (keep it simple, stupid!) k-means algorithm, 2^nd, 3^rd, 4^th kNN (k-nearest neighbors)

L

landing-page optimization large datasets, classification with latency latent space learning rate learning vector quantization. See LVQ. leave-one-out technique left-most neuron lifecycle of classifiers of intelligent algorithms linear approximation linear models, 2^nd linear regression, 2^nd, 3^rd linear relationship linear supervised learning link function load method, 2^nd log shipping log.dirs parameter logging logic inputs logistic profile logistic regression, 2^nd, 3^rd, 4^th, 5^th, 6^th, 7^th, 8^th log-odds, 2^nd, 3^rd, 4^th LVQ (learning vector quantization)

M

m1-1m dataset MABs (multi-armed bandits) epsilon decreasing strategy epsilon first strategy epsilon greedy strategy extensions to adversarial bandits contextual bandits MAC addresses machine learning. See ML. make_ellipses method mapping, predicted click-through rate to bid price MapReduce McCulloch, Warren MCP model mean_center metaclassifier scheme Millwheel min_values Minsky, Marvin mirroring missing values ML (machine learning) MLP (multilayer perceptron) model training, 2^nd, 3^rd modulo operator MovieLens dataset multi-armed bandits. See MABs. multiclass classification, 2^nd, 3^rd multicluster operation multilayer perceptrons activation functions backpropagation and in scikit-learn learned multiple partitions

N

n dimensional space n_components parameter n_iter parameter n-dimensional Gaussian nearest-neighbor algorithms negative classes negative classification negative exponential profile negative sign Netflix Prize example neurons new_data array Ng, Andrew NNs (neural networks) multilayered activation functions backpropagation and in scikit-learn learned multilayered perceptron overview perceptrons geometric interpretation of for two inputs training RBMs (Restricted Boltzmann Machines) and BRBMs (Bernoulli Restricted Boltzmann Machines) illustrative example overview noise normal interval nudge_dataset function NumPy arrays, 2^nd

O

odds of event offset attribute OMG (Object Management Group) one-hot-encoding online advertising. See click prediction for online advertising. online learning, 2^nd online optimization only_unknowns=True parameter ontologies optimal performance Orkut orthogonal vectors output error output layer, 2^nd output node output profiles

P

PA (predictive analytics), 2^nd Papert, Seymour parametric model PARC (Palo Alto Research Center) partial derivative partition attribute PCA (principal component analysis), 2^nd PDFs (probability density functions), 2^nd perceptron learning algorithm perceptrons geometric interpretation of for two inputs multilayer activation functions backpropagation and in scikit-learn learned training performance evaluation evaluating intelligence evaluating predictions personalized physical advertising petal length and width physical advertising, personalized pipe character Pitts, Walter placement, information about plot_gmm_classifier plotting output positive classes positive classification positive values post_normalize pre_normalize precision predict_rating method predictive analytics. See PA. PredictWallStreet pre-prepared ads principal component analysis. See PCA. prior knowledge probabilities.out file probability distribution functions. See PDFs. probability, obtaining with Gibbs measure ProduceResponse objects production rules production stage programmatic buying publish-subscribe system pull_handle method PyBrain, 2^nd python-recsys package

Q

quantified self

R

–r switch RatingCountMatrix class RBM/LR pipeline RBMs (Restricted Boltzmann Machines) BRBMs (Bermoulli Restricted Boltzmann Machines) illustrative example overview training readable.models recall, 2^nd receiver operating characteristic curve. See ROC. recommend method recommender algorithm recommending relevant content distance and similarity evaluating recommender how recommender engines work model-based recommendation using singular value decomposition choosing movies for given user choosing users for given movie overview Netflix Prize example user-based collaborative filtering regression algorithms regret accumulated maximum relevant content. See recommending relevant content. reliability of data replication in Kafka log-processing platform message acknowledgement replication, leaders, and in-sync replicas Restricted Boltzmann Machines. See RBMs. retargeting Rete algorithm right-most neuron RMSE (root-mean-square error) ROC (receiver operating characteristic) curve, 2^nd, 3^rd Rosenblatt, Frank rows value rudimentary ontology rule-based algorithms runtime performance

S

scalability scale of probability scaling characteristics, of intelligent algorithms scikit-learn, 2^nd, 3^rd SE (standard error) second price auction self-driving vehicle semantic ontology semantic web semi-empirical approach SGD (stochastic gradient descent) similarity similarity of probability SimilarityMatrix class SimpleConsumer SimpleProducer, 2^nd, 3^rd simplicity single hidden units single hyperplane single perceptron single test variable single weight value single-shot optimization Single-Vector Lanczos Method singular value decomposition. See SVD. size of data sklearn.cluster.KMeans sklearn.datasets.base.Bunch sklearn.mixture.GMM social implications of intelligent web sort tool sortModules method specificity, 2^nd speed square root profile standard error. See SE. statistical algorithms, 2^nd, 3^rd, 4^th stochastic gradient descent. See SGD. structural algorithms, 2^nd structure sum_probs_chosen method SVD (singular value decom-position), 2^nd, 3^rd choosing movies for given user choosing users for given movie overview svd.compute method svd.load_data method svd.recommend method, 2^nd SVDLIBC library SVMs (support vector machines)

T

target array, 2^nd target_names attribute test topic test_vw_file testing stage test-replicated-topic tfidf value threshold bias TNR (true negative rate) topic attribute TPR (true positive rate), 2^nd train method train_vw_file trained model trained neural networks, 2^nd training data training pipeline training stage two-layer hidden network two-layer perceptron type I and II errors

U

ubiquitous computing user-based collaborative filtering userdict object users, information about utility problem

V

validation stage, 2^nd vehicles, self-driving Vickrey auction visible units, 2^nd volume Vowpal Wabbit (VW), click prediction with model calibration preparing dataset testing model VW data format

W

Wald interval web. See intelligent web. weighted sum, 2^nd weighted_ratings weighting parameter

X

XOR function, 2^nd, 3^rd, 4^th

Z

z value zip method Zookeeper, 2^nd z-test

..................Content has been hidden....................

You can't read the all page of ebook, please click here login for view all page.