- A
- activations, neural networks, 113, 114
- AGI (Artificial General Intelligence), 11
- agile development, 201
- AI (artificial intelligence)
- Amazon, 10, 290
- applications, 8–11
- art generation, 6, 77–88, 169–180, 290
- engines, 10
- Google, 10
- image generation
- GAN, 180
- generative models, 180
- neural networks, 181–185
- probability distribution, 180
- knowledge representation, 8–9
- Netflix, 10
- overview, 1–2, 7–8
- photograph generation, 290
- planning, 10
- research pages
- self‐driving cars, 9
- smart task examples, 5–6
- software development, 202–203
- strategy, 10
- video surveillance and, 10
- Ajax, 204–205
- algorithms
- decision trees, 54
- Random Forest, 56
- Amazon. See also AWS (Amazon Web Services)
- AI (Artificial Intelligence) and, 290
- Alexa, 104
- SageMaker, 266
- analytics, 5–7
- applications
- descriptive, 13–14
- diagnostic, 14–15
- predictive, 15
- prescriptive, 16–17
- building, 12–13
- decision logic
- models, data‐driven, 18
- physics‐based analytics, 17–18
- rules‐based analytics, 17–18
- development, system building, 18–21
- user interface, 271–275
- Analytics Vidhya, 291, 293
- Apache Spark, 269
- applications
- Cloud‐native, 211
- containerized, 209
- containers, packaging as, 233–237
- growth in size, 205
- Kubernetes, deployment as microservice, 238–240
- web applications, 203
- Ajax, 204–205
- HTML (HyperText Markup Language), 203
- HTML 2.0, 205
- HTTP (HyperText Transfer Protocol), 204
- JavaScript, 204–205
- server‐side scripting, 204–205
- architectures
- CNN (Convolutional Neural Networks), 143–145
- feed‐forward architecture, 160
- microservices, containers and, 212–214
- models, 144
- SOA (Services Oriented Architecture), 212
- arrays, words, 97
- art generation, 169–180
- calculating loss, 175–176
- content, 174–175
- Prizma, 170, 179
- The Starry Night (van Gogh), 171
- style, 174–175
- artificial neural networks, 112
- ASCII (American Standard Code for Information Interchange), 73
- audio data, 104
- FFT (Fast Fourier Transform), 105
- pressure waves, 105
- sequence‐to‐sequence models, 104
- autoencoders, 188–189
- AutoML, 251–252
- AWS (Amazon Web Services), 200, 207
- Elastic Beanstalk, 208
- SageMaker, 245, 270
- virtual machines, 200
- Azure ML Studio, 245
- B
- back‐propagation, 43
- Batch Gradient Descent, 119–120
- batch normalization network layer, 135–136
- Bengio, Yoshua, 293
- Bethge, Matthias, 170, 293
- bias, 57–61
- Big Data, Hadoop, 208
- binary classification, 43–44
- blackboxes, neural networks, 115
- boilerplate code, 225
- Bontempi, Gianluca, 188
- C
- CaaS (Container‐as‐a‐Service), 199, 209, 211
- containers, 209
- DockerHub, 211
- Kubernetes, 214–220
- microservices and, 211
- Caelen, Olivier, 188
- calculating loss, 175–176
- CD (continuous delivery), 245
- CGI‐Scripts, HTML content, 204
- cGroups (Linux), 210
- chatbots, 91
- Chollet, François, 292
- CI (continuous integration), 245
- CI/CD (Continuous Integration and Continuous Delivery), 201–202, 202–203
- classification, 43–44
- binary, 43–44
- data augmentation and, 150–160
- datasets and, 48–52
- Ensemble method, 56
- ground truth, 53
- logistic regression, 53
- Random Forest algorithm, 56
- transfer learning and, 150–160
- clickstreams, 73
- cloud, terminology, 206
- cloud computing, 205
- application size growth, 205
- AWS (Amazon Web Services), 207
- CaaS (Container‐as‐a‐Service), 209
- globalization and, 205
- IaaS (Infrastructure‐as‐a‐Service), 207, 208
- PaaS (Platform‐as‐a‐Service), 207, 208
- SaaS (Software‐as‐a‐Service), 207, 209
- virtualization, 207
- Cloud‐native applications, 211
- CNN (Convolutional Neural Networks), 73, 128, 131, 250
- codecs (video), 89
- columnar data, 72
- computer vision, 77–88
- computer vision libraries, 77
- containerized applications, 209
- containers
- CaaS (Container‐as‐a‐Service), 209
- Cloud‐native applications and, 200
- DevOps and, 211
- Docker, 210
- environment, 210
- Linux, 210
- packaging applications as, 233–237
- video, 89
- convolution layers, 69
- convolution network layer, 133–135
- corpus, 91
- Coursera, 5
- Courville, Aaron, 293
- credit card fraud detection, 188
- autoencoders, 188–189
- creditcard.cvs file, 188
- cross‐validation, 127
- K‐fold cross‐validation, 127
- Crystal Reports, 13
- D
- Dal Pozzolo, Andrea, 188
- DAQ (data acquisition system), 72
- data
- acquisition, 267–270
- CSV files, 267–268
- gold datasets, 270
- Kafka, 268
- analytics, building, 12–13
- augmentation, 149–150
- classification and, 150–160
- cleansing, 32, 246–248
- collection, 246–248
- descriptive statistics, 246–247
- frames, 33
- overview, 1–2
- patterns (See patterns)
- preparation, 246–248
- data centers, 206
- data‐driven models, 18
- DataRobot, 271
- datasets
- autoencoder, 188–189
- classification and, 48–52
- creditcard.csv file, 188
- decoding, 189
- dimensionality reduction, 188
- encoding vector, 189
- gold datasets, 270
- patterns, 189
- PCA (Principal Component Analysis), 188
- reconstruction error, 197–198
- timeseries, 247
- training datasets, 151
- validation datasets, 151
- decision logic analytics
- models, data‐driven, 18
- physics‐based analytics, 17–18
- rules‐based analytics, 17–18
- decision trees, 54
- Deep Models, VGGNet, 145–149
- DeepMind, 69
- dependent variables, 13
- deployment, ML model, 252–253
- descriptive analytics, 13–14, 25, 246–247
- desktop applications, 203
- Windows Remote Desktop for Windows, 208
- DevOps, containers and, 211
- diagnostic analytics, 14–15
- digital cameras, 74
- dimensionality reduction, 188
- Django, 224
- DL (Deep Learning), 11, 61
- CNN (Convolutional Neural Networks), 131
- images, fashion, classifying, 136–142
- models, 114, 131–132
- deployment, Keras, 282–286
- MLP (multi‐layered perceptron), 131–132
- networks, 112
- Adam optimizer, 125
- batch normalization layer, 135–136
- convolution layer, 133–135
- Cross‐Entropy Loss function, 126
- dropout layer, 135
- Keras, 121–126
- pooling layer, 135
- PyTorch, 121–126
- TensorFlow, 121–126
- patterns, 131
- unstructured data, 20
- Docker, 210
- containers, 233–237
- Docker files, 233–237
- images, pushing to repository, 238
- microservices, building, 223–228
- documents, 91
- DQN (Deep Q Networks), 65, 68–69
- dropout, 128
- dropout network layer, 135
- E
- eager execution, 172
- Ecker, Alexander S., 170, 293
- edge processing, 3
- ML model deployment on edge devices, 254–263
- Einstein (SalesForce), 245
- EIP (Enterprise Integration Patterns), 212
- Ellison, David, 293
- encoder, 188
- end‐to‐end earning, 73
- engineering, feature engineering, 32
- ERP (Enterprise Resource Planning), 212
- ESB (Enterprise Service Bus), 212
- ExpressJS, 224
- F
- Facebook, 200
- AI (artificial intelligence), research pages, 290
- fashion image classification, 136–142
- FBLearner flow, 245
- feature engineering, 32, 132
- feed‐forward architecture, 160
- FFT (Fast Fourier Transform), 105
- FitBit, motion tracking, 20
- Flash, 224
- Fowler, Martin, 294
- FPGAs (field‐programmable gate arrays), 5
- G
- gaming data, 73
- GAN (Generative Adversarial Networks), 6, 181–185
- Gatys, Leon A., 170, 293
- GCP (Google Cloud Platform), 200
- General Electric, 289
- generative models, image generation, 180
- Gensim, 99
- GET call (HTTP), 225
- GIF (Graphics Interchange Format), 77
- GIGO (Garbage In Garbage Out), 25, 246–248
- GitHub, 294
- GLM (generalized linear model), 274
- globalization, cloud computing and, 205
- gold dataset, 270
- Goodfellow, Ian J., 293
- Google
- AI (artificial intelligence), research pages, 290
- AutoML, 245, 266
- Colab Notebook, 180
- Colaboratory, 200, 271
- Home, 104
- machine learning course, 291
- Maps, 16
- TPU, 5
- GPS, 3
- GPU (graphic processing unit), 201
- gradient descent
- linear regression, 40–43
- neural networks, 117–119
- Batch Gradient Descent, 119–120
- SGD (stochastic gradient descent), 119–120
- granularity, 80
- ground truth, 53
- ML model lifecycle, 245–246
- H
- H20
- AI (artificial intelligence), research pages, 290
- ML regression model, 272–275
- Hadoop, 208
- hardware, software‐defined, 200
- HDF (Hierarchical Data Format), 159
- Hinton, Geoffrey, 43
- HTML (HyperText Markup Language), 203
- Ajax, 204–205
- CGI‐Scripts, 204
- HTML 2.0, 205
- Java Servlets, 204
- JavaScript, 204–205
- PHP, 204
- stylesheets, 226
- templates, 226
- HTTP (HyperText Transfer Protocol), 204
- hyper‐parameters, 30, 36, 42, 143–145
- AutoML, 251–252
- ML model, 251–252
- Hypervisor, 209–210
- I
- IaaS (Infrastructure‐as‐a‐Service), 207, 208
- IBM Watson, 8
- AI (artificial intelligence), research pages, 290
- IDE (integrated development environments), 200
- ILSVRC (ImageNet Large Scale Visual Recognition Challenge), 143
- image processing libraries, 77
- ImageNet, 143–145
- images
- array operations, 81
- computer vision, 77
- digital cameras, 74
- Docker, pushing to repository, 238
- fashion, classifying, 136–142
- generating
- GAN (generative adversarial networks), 180, 181–185
- generative models, 180
- neural networks, 181–185
- probability distribution, 180
- GIF (Graphics Interchange Format), 77
- granularity, 80
- image processing, 77
- JPG/JPEG (Joint Photography Experts Group), 77
- PNG (Portable Network Graphics), 77
- resolution, 80
- scanners, 74
- Inception, 144
- independent variables, 12–13
- industrial IoT, 289
- Industry 4.0, 4, 289
- Instagram, 200
- integers, 24
- Intel, AI (artificial intelligence), research pages, 290
- IoT (Internet of Things)
- connectivity, built‐in, 3
- edge processing, 3
- overview, 2–3
- J
- Jain, Kunal, 291
- Java
- JAR files, 208
- microservices, 224
- Servlets, HTML content, 204
- JavaScript, 204–205
- Johnson, Reid A., 188
- JPG/JPEG (Joint Photography Experts Group), 77
- JSON (JavaScript Object Notation), 269
- Jupyter, 202
- Jupyter Notebooks, 292
- JupyterHub, 287
- K
- K‐fold cross‐validation, 127
- K‐Means, 28, 34
- Kafka, 268
- Kaggle, 26, 291
- Karmarkar, Abhijit, 294
- Katacoda, 293, 294
- KDnuggets, 293
- Keras, 121–126, 172, 292
- data augmentation, 149–150
- HDF (Hierarchical Data Format), 159
- NLP mode, 228–232
- KNN (K‐Nearest Neighbors), 53–54
- knowledge representation, 8–9
- KPIs (key performance indicators), 269
- Kubeflow, 244, 287
- Kubernetes, 214
- apps, deploying as microservices, 238–240
- extensions, 244
- microservices, building, 223–228
- Minikube, 214
- ML model lifecycle, 244
- nodes, 215
- plug‐ins, 244
- pods, 216
- VMs (virtual machines), 214
- L
- Lecun, Yann, 74
- lemmatization, 94
- Lewis, James, 294
- Li, Fei Fei, 143
- linear regression, 37–40
- Linux
- logging, 286–287
- logistic regression, 53
- LSTM (Long Short‐Term Memory), 160, 250
- M
- Machine Learning Group, 188
- MAE (mean absolute error), 119
- Marr, Bernard, 289
- MDM (Master Data Management), 271
- Mesosphere, 294
- metrics
- precision and, 49
- recall and, 49–50
- microservices, 199
- app deployment on Kubernetes, 238–240
- architecture, containers and, 212–214
- boilerplate code, 225
- CaaS and, 211–212
- Docker and, 223–228
- ExpressJS, 224
- HTML
- stylesheets, 226
- templates, 226
- Java, 224
- Kubernates and, 223–228
- NodeJS, 224
- Python, 224–228
- Microsoft Azure, 200
- Microsoft Azure Studio, 266
- Minikube, 214–215
- Mirza, Mehdi, 293
- ML (machine learning), 10
- community, 26–27
- courses, 290–292
- data acquisition, 267–270
- data cleansing, 270–271
- datasets, gold datasets, 270
- deployment, automated, 278–279
- development, lifecycle, 243–263
- FitBit motion tracking, 20
- hyper‐parameter tuning, 278–279
- JSON (JavaScript Object Notation), 269
- metrics
- precision, 49
- recall, 49–50
- model development, 275–277
- neural networks, 42, 43
- platform (See platforms)
- reinforcement, 31
- supervised
- classification, 30–31, 43–48, 52–56
- Cost function, 30
- Error function, 30
- gradient descent, 30
- hyper‐parameters, 30
- learning rate, 30
- model training, 30
- models, 29
- neural networks, 29
- positives, 29
- probability, 31
- regression, 30
- weights, 29
- Talend, 270–271
- Tamr, 270–271
- training at scale, 277
- unsupervised
- anomaly detection, 29
- clustering, 27–28
- dimensionality reduction, 28
- problems, 33–36
- ML model
- AutoML, 251–252
- AWS SageMaker, 245
- Azure ML Studio, 245
- building, 248–251
- CNN (Convolutional Neural Network), 250
- data
- cleansing, 246–248
- collection, 246–248
- feature engineering, 247–248
- preparation, 246–248
- timeseries, 247
- deployment, 252–253
- automated, 279–286
- edge devices, 254–263
- NVIDIA, 254–255
- development, 275–277
- Einstein, 245
- FBLearner flow, 245
- feedback, 253–254
- Google AutoML, 245
- ground truth, 245–246
- hyper‐parameters, 251–252
- lifecycle, 244–263
- LSTM (Long Short‐Term Memory), 250
- pipeline, 244
- problem definition, 245–246
- RNN (recurrent neural network), 250
- TensorFlow‐Serving, 253
- training, 248–251
- updates, 253–254
- validation, 251–252
- MLP (multi‐layered perceptron), 120, 131–132
- model‐based RL, 63–64
- model‐free RL, 63–64
- models, 23
- architectures, 144
- audio data sequence‐to‐sequence models, 144
- data‐driven, 18
- Deep Models, VGGNet, 145–149
- DL (Deep Learning), 114, 131–132
- deployment, Keras, 282–286
- MLP (multi‐layered perceptron), 131–132
- edge processing, 254–263
- generative models, image generation, 180
- GLM (generalized linear model), 274
- training, 25
- monitoring, 286–287
- Moore's Law of electronics, 4–5
- MSE (mean squared error), 117–118, 119
- MVC (Model‐View‐Controller), 212
- N
- NER (Named Entity Recognition), 95–97
- networks
- batch normalization layer, 135–136
- convolution layer, 133–135
- dropout layer, 135
- pooling layer, 135
- neural networks, 42, 43, 112–117., See also CNN (Convolutional Neural Networks)
- activations, 113, 114
- artificial, 112
- back‐propagation, 117–119
- bias neurons, 116, 117
- blackboxes, 115
- CNN (Convolutional Neural Networks), 250
- computational graph, 114
- dataflow, 114
- Deep Learning framework
- Adam optimizer, 125
- Cross‐Entropy Loss function, 126
- Keras, 121–126
- PyTorch, 121–126
- TensorFlow, 121–126
- dense layer, 115
- feature engineering, 132
- feed‐forward architecture, 115
- fully‐connected layer, 115
- gradient descent, 117–119
- Batch Gradient Descent, 119–120
- SGD (stochastic gradient descent), 119–120
- human brain comparison, 112–113
- image generation
- discriminator, 181–185
- generator, 181–185
- MLP (multi‐layered perceptron), 120
- RNN (recurrent neural network), 160–166, 250
- shallow, 114
- weight values, 115–116, 117
- neural style transfer, 169–180
- Ng, Andrew, 5, 289, 290, 292
- NLP (Natural Language Processing), 10–11, 91–97, 228, 292
- algorithms, 90–91
- data cleansing, 92–94
- H5 binary files, 228
- Keras, 228–232
- lemmatization, 94
- NLU (Natural Language Understanding), 91
- stemming, 94
- NLTK (Natural Language Tool Kit), 91
- NLU (Natural Language Understanding), 91
- NodeJS, 224
- nodes, Kubernetes, 215
- NumPy (Numerical Python), 121–126
- NVIDIA, 254–255
- AI (artificial intelligence), research pages, 290
- GPU, 5
- O
- one‐hot encoding, 97–99
- OpenCV, 77
- images
- array operations, 81
- granularity, 80
- loading as array, 77
- resolution, 80
- tutorials, 291
- video, 89–90
- OTA (over‐the‐air) updates, 200
- overfitting, 57–61, 126–128
- Ozair, Sherjil, 293
- P
- PaaS (Platform‐as‐a‐Service), 199, 207, 208
- Pandas data frame, 34, 49
- parameters, hyper‐parameters, 30, 36, 42
- patterns, 23–24
- PCA (Principal Component Analysis), 102, 188
- PHP (hypertext preprocessor), HTML content, 204
- physics‐based analytics, 17–18
- Pichai, Sundar, 11
- platforms, 265–266
- Amazon SageMaker, 266, 270
- Google AutoML, 266
- Microsoft Azure Studio, 266
- PNG (Portable Network Graphics), 77
- podcasts, 291
- pods, Kubernetes, 216
- pooling network layer, 135
- POS (parts of speech), 95–97
- POST call (HTTP), 225
- Pouget‐Abadie, Jean, 293
- predictions, VGGNet, 145–149
- predictive analytics, 15, 25
- prescriptive analytics, 16–17
- pressure waves (sound), 105
- Prizma, 170, 179
- probability distribution, image generation, 180
- problem solving, 31–32
- ML model lifecycle, 245–246
- processing growth, 4–5
- project management, scrum, 201
- Python, 202
- microservice building, 224–228
- Q
- Q‐Learning, 65, 66–68
- QlikView, 13
- R
- Ramsundar, Bharath, 292
- Random Forest algorithm, 56
- real numbers, 24
- reconstruction errors, 197–198
- references
- AI (artificial intelligence), 289–290
- models as microservices, 294
- modern software, 293
- big data, 289–290
- Deep Learning
- machine learning, 290–291
- development lifecycle, 294
- platform, 294
- unstructured data, 291–292
- regression, 30
- linear, 37–40
- gradient descent and, 40–43
- logistic, 53
- regularization, 128
- reinforcement learning, 31
- Renelle, Tyler, 291
- repositories, Docker images, 238
- ResNet, 144
- resolution, 80
- resources, ML community, 26
- RFID tags, 3
- RL (Reinforcement Learning), 62
- DQN (Deep Q Networks), 65, 68–69
- model free, 65–66
- exploitation, 65–66
- exploration, 66
- greedy policy, 66
- random policy, 66
- model‐based, 63–64
- Q‐Learning, 65, 66–68
- SARSA, 65
- RNN (Recurrent Neural Networks), 73, 91, 160–166, 250
- Rosebrock, Adrian, 291
- rules‐based analytics, 17–18
- S
- SaaS (Software‐as‐a‐Service), 199, 207, 209
- Salesforce Einstein, AI (artificial intelligence) research pages, 290
- SARSA (state‐action‐reward‐state‐action), 65
- scanners, 74
- Scikit‐Learn, 292
- scrum, 201
- SDLC (software development lifecycle), 245
- self‐driving cars, 9
- sequence of points, 25
- sequence‐to‐sequence models, 104
- server‐side scripting, web applications, 204–205
- SGD (stochastic gradient descent), 119–120, 144
- shallow networks, 114
- Sisense, 13
- SOA (Services Oriented Architecture), 199, 212
- social media, 200
- software
- development, AI with modern software, 202–203
- modern, 200–202
- sprints, 201
- software‐defined hardware, 200
- sound. See audio data
- sprints, 201
- SQL (Structured Query Language), 267–268
- structured data, 71–74, 112
- clickstreams, 73
- columnar data, 72
- credit card fraud detection, 188–198
- datasets
- autoencoders, 188–189
- decoding, 189
- dimensionality reduction, 188
- encoding vector, 189
- patterns, 189
- PCA, 188
- reconstruction error, 197–198
- gaming data, 73
- tabular data, 72
- timeseries data, 72
- weblogs, 73
- stylesheets (HTML), 226
- supervised ML
- classification, 30–31, 43–44
- Cost function, 30
- Error function, 30
- gradient descent, 30
- hyper‐parameters, 30
- learning rate, 30
- model training, 30
- models, 29
- neural networks, 29
- positives, 29
- probability, 31
- regression, 30
- weights, 29
- T
- Tableau, 13
- tabular data, 72
- Talend, 270–271
- Tamr, 270–271
- templates (HTML), 226
- TensorFlow, 172, 292, 293
- TensorFlow‐Serving, 253
- textual data, 90–91
- chatbots, 91
- documents, 91
- NER (Named Entity Recognition), 95–97
- NLP (Natural Language Processing), 91–97
- POS (parts of speech), 95–97
- RNN (Recurrent Neural Networks), 91
- TF‐IDF (term frequency‐inverse document frequency), 99
- word embeddings, 97–103
- word frequencies, 99
- Word2Vec model, 102
- TF‐IDF (term frequency‐inverse document frequency), 99
- TF‐Job, 287
- TF‐Serving (TensorFlow‐Serving), 280, 287
- The Starry Night (van Gogh), 171
- timeseries data, 72
- tokens, 91
- topic modeling, 99
- TPU (Tensor Processing Unit), 255–263
- training, models, 25
- training datasets, 151
- transfer learning, 149–150
- classification and, 150–160
- Twitter, 200
- U
- ULB (Université Libre de Bruxelles), 188
- underfitting, 57–61, 126–128
- unstructured data, 71–74, 111–112
- end‐to‐end earning, 73
- features, extracting, 73
- tweets, 73
- Unsupervised Learning, 270
- unsupervised ML
- anomaly detection, 29
- clustering, 27–28
- DBSCAN, 28
- Hierarchical Clustering, 28
- K‐Means, 28
- dimensionality reduction, 28
- PCA (principal component analysis), 28
- problems, 33–36
- user interface, analytics, 271–275
- V
- validation, ML model, 251–252
- validation datasets, 151
- variance, 57–61
- Vasudevan, Gautam, 294
- vectors
- one‐hot encoding, 97–99
- words, 97
- VGG (Visual Geometry Group), 145–149
- VGGNet, 145–149
- video
- codecs, 89
- containers, 89
- OpenCV, 89–90
- surveillance, 10
- virtualization, 207
- VMs (virtual machines), 200
- Hypervisor, 209–210
- JAR files, 208
- Kubernetes, 214
- W
- Warde‐Farley, David, 293
- waterfall model, 201
- Waymo self‐driving car, 9
- web applications, 203
- Ajax, 204–205
- HTML (HyperText Markup Language), 203
- HTML 2.0, 205
- HTTP (HyperText Transfer Protocol), 204
- server‐side scripting, 204–205
- web browsers, applications, 203
- weblogs, 73
- WhatsApp, 200
- Windows Remote Desktop for Windows as a desktop, 208
- word embeddings
- arrays, 97
- TF‐IDF (term frequency‐inverse document frequency), 99
- topic modeling, 99
- vectors, 97
- word frequencies, 99
- Word2Vec model, 102
- word frequencies, 99
- Word2Vec model, 102
- X
- Xs (independent variables), 12–13
- Xu, Bing, 293
- Y
- YAML files, 239–240
- Ys (dependent variables), 13
- Yuan, Raymond, 172, 293
- Z
- Zadeh, Reza, 292
- zettabytes, 5
..................Content has been hidden....................
You can't read the all page of ebook, please click
here login for view all page.