Download as PDF, PPTX
Uploaded byCarlos Paredes
Machine_Learning_with_MATLAB_Seminar_Latest.pdf
Machine learning and deep learning techniques can be used to analyze diverse types of data such as images, text, signals and more. Deep learning uses neural networks to learn directly from raw data, enabling applications like object recognition, speech recognition, and analyzing time series signals. Deep learning has become popular due to labeled public datasets, increased GPU acceleration, and pre-trained models that provide a starting point for new problems.
More Related Content
Similar to Machine_Learning_with_MATLAB_Seminar_Latest.pdf
Machine_Learning_with_MATLAB_Seminar_Latest.pdf
- 1. 1 © 2019 TheMathWorks, Inc. Machine Learning and Deep Learning with MATLAB Sergio E. Obando Quintero, Ph.D. Application Engineer Source: https://xkcd.com/1838/
- 2. 2 Machine learning inaction CamVid Dataset 1. Segmentation and Recognition Using Structure from Motion Point Clouds, ECCV 2008 2. Semantic Object Classes in Video: A High-Definition Ground Truth Database, Pattern Recognition Letters
- 3. 3 Machine Learning isEverywhere Automobile Industrial Automation CES & Aero Defense Energy & Finance Tire Wear Detect Oversteer Engine Health (Pred Maintenance) Forecasting & Risk Analysis Overlay metrology improvement Building energy use optimization Portfolio Allocation Telecom customer churn prediction
- 4. 4 Artificial Intelligence Machine Learning AI,Machine Learning and Deep Learning Deep Learning Timeline 1950s Today 1980s Application Breadth Automated Driving Speech Recognition Robotics Object Recognition Bioinformatics Recommender Systems Spam Detection Fraud Detection Weather Forecasting Algorithmic Trading Sentiment Analysis Medical Diagnosis Health Monitoring Computer Board Games Machine Translation Knowledge Representation Perception Reasoning Interactive Programs Expert Systems
- 5. 5 What is MachineLearning? learn complex non- linear relationships Solution is too complex for hand written rules or equations update as more data becomes available Solution needs to adapt with changing data learn efficiently from very large data sets Solution needs to scale Speech Recognition Object Recognition Engine Health Monitoring Weather Forecasting Energy Load Forecasting Stock Market Prediction IoT Analytics Taxi Availability Airline Flight Delays Ability to learn from data without being explicitly programmed
- 6. 6 Real-life saving pattern-detectionapplication! Read more... From Traditional Monitoring ▪ Outdated technology ▪ Not suitable for children ▪ Considerable patient effort To Respiratory Digital Health ▪ Sensors, apps, etc. to monitor and control symptoms ▪ Convenient effortless ▪ Objective symptom measurement
- 7. 7 Goals Part I ▪Overview of machine learning ▪ Machine learning models & techniques available in MATLAB ▪ Using MATLAB to streamline your machine learning workflow
- 8. 8 Machine Learning Machine learninguses data and produces a program to perform a task Traditional Approach Machine Learning Approach 𝑚𝑜𝑑𝑒𝑙 = < 𝑴𝒂𝒄𝒉𝒊𝒏𝒆 𝑳𝒆𝒂𝒓𝒏𝒊𝒏𝒈 𝑨𝒍𝒈𝒐𝒓𝒊𝒕𝒉𝒎 >(𝑠𝑒𝑛𝑠𝑜𝑟_𝑑𝑎𝑡𝑎, 𝑎𝑐𝑡𝑖𝑣𝑖𝑡𝑦) Computer Program Machine Learning 𝑚𝑜𝑑𝑒𝑙: Inputs → Outputs Hand Written Program Formula or Equation If X_acc > 0.5 then “SITTING” If Y_acc < 4 and Z_acc > 5 then “STANDING” … 𝑌𝑎𝑐𝑡𝑖𝑣𝑖𝑡𝑦 = 𝛽1𝑋𝑎𝑐𝑐 + 𝛽2𝑌𝑎𝑐𝑐 + 𝛽3𝑍𝑎𝑐𝑐 + … Task: Human Activity Detection
- 9. 9 Steps Challenge Access andExplore data Data diversity Numeric, Images, Signals, Text – not always tabular Preprocess Data Lack of domain tools Filtering and feature extraction Feature selection and transformation Develop Predictive Models Time consuming Train and compare several models to find the “best” Select optimal parameters and avoid overfitting Integrate Analytics with Systems Platform diversity Translate analytics to production Deploy on different target platforms Iterate Challenges in Machine Learning
- 10. 10 Different Types ofMachine Learning Machine Learning Supervised Learning Classification Regression Unsupervised Learning Clustering Group and interpret data based only on input data Develop predictive model based on both input and output data Type of Learning Categories of Algorithms • Discover a good internal representation • Learn a low dimensional representation • Output is a real number (temperature, stock prices). • Output is a choice between classes • (True, False) (Red, Blue, Green)
- 11. 11 Different Types ofMachine Learning Machine Learning Supervised Learning Classification Regression Unsupervised Learning Clustering Group and interpret data based only on input data Develop predictive model based on both input and output data Type of Learning Categories of Algorithms • k-means, k-medoids, Hierarchical, Gaussian Mixture Models, Nearest Neighbors, Hidden Markov Models • Linear Regression, Generalized Linear Regression, Nonlinear Regression, Mixed-Effects, Stepwise, Gaussian Process Regression • Decision Trees, Nearest Neighbor, Support Vector Machines, Ensemble Methods, Discriminant Analysis
- 12. 12 Machine Learning Workflow IntegrateAnalytics with Systems Desktop Apps Enterprise Scale Systems Embedded Devices and Hardware Files Databases Sensors Access and Explore Data Develop Predictive Models Model Creation e.g. Machine Learning Model Validation Parameter Optimization Preprocess Data Working with Messy Data Data Reduction/ Transformation Feature Extraction
- 13. 13 MODEL PREDICTION Machine Learning Workflow Train:Iterate till you find the best model Predict: Integrate trained models into applications MODEL SUPERVISED LEARNING CLASSIFICATION REGRESSION PREPROCESS DATA SUMMARY STATISTICS PCA FILTERS CLUSTER ANALYSIS LOAD DATA PREPROCESS DATA SUMMARY STATISTICS PCA FILTERS CLUSTER ANALYSIS NEW DATA
- 14. 14 Classification: Human ActivityRecognition Objective: Train a classifier to classify human activity from sensor data Data: Approach: – Extract features from raw sensor signals – Train and compare classifiers – Test results on new sensor data Predictors 3-axial Accelerometer and Gyroscope data Response Activity:
- 15. 15 PREDICTION MODEL Machine Learning Workflowfor Classification Example Train: Iterate till you find the best model Predict: Integrate trained models into applications MODEL SUPERVISED LEARNING CLASSIFICATION REGRESSION PREPROCESS DATA SUMMARY STATISTICS PCA FILTERS CLUSTER ANALYSIS LOAD DATA PREPROCESS DATA SUMMARY STATISTICS PCA FILTERS CLUSTER ANALYSIS TEST DATA 1. Mean 2. Standard deviation 3. PCA Classification Learner 1. Mean 2. Standard deviation 3. PCA
- 16. 17 Training: Machine Learningwith MATLAB After this course you will be able to: ▪ Discover natural patterns in data ▪ Create predictive models ▪ Validate the predictions of a model ▪ Simplify and improve models
- 17. 18 Challenge Solution Data diversity Extensivedata support Work with signal, images, financial, textual, and others formats Lack of domain tools High-quality libraries Industry-standard algorithms for Finance, Statistics, Signal, Image processing & more Time consuming Interactive, app-driven workflows Focus on machine learning, not programing Select best model and easily fine-tune model parameter Platform diversity Run analytics anywhere Code generation for embedded targets Deploy to broad range of enterprise system architectures Flexible architecture for customized workflows Complete machine learning platform MATLAB Strengths for Machine Learning
- 18. 19 Key Takeaways Use appsin your workflow to quickly compare and select candidate algorithms Use programmatic workflows for fine-tuning model parameters to achieve robust performance Use feature selection to get rid of unnecessary features and prevent overfitting Use automatic code generation to rapidly deploy your analytics to embedded targets
- 19. 20 Integrate Analytics withSystems MATLAB Runtime C, C ++ HDL PLC Embedded Hardware CUDA MATLAB Analytics run anywhere C/C++ ++ Excel Add-in Java Hadoop/ Spark .NET MATLAB Production Server Standalone Application Enterprise Systems Python Web Apps
- 20. 21 Machine Learning forEdge Analytics and Code Deployment ▪ SVM Class. ▪ Linear Class. ▪ Linear Regr. ▪ Generalized Linear Regr. ▪ Decision trees ▪ Ensembles for Class. ▪ Ensembles for Regr. ▪ SVM Regr. ▪ KNN Class. ▪ Gaussian Process Regr. ▪ Discriminant Analysis Deploy trained models as standalone C/C++ code
- 21. 23 © 2019 TheMathWorks, Inc. Deep Learning A Practical Approach in MATLAB
- 22. 24 What is DeepLearning?
- 23. 25 Deep learning isa type of machine learning in which a model learns to perform tasks directly from image, time-series or text data. Deep learning is usually implemented using a neural network architecture.
- 24. 26 Example 1: Objectrecognition using deep learning Training (GPU) Millions of images from 1000 different categories Prediction Real-time object recognition using a webcam connected to a laptop
- 25. 27 Example 2: Detectionand localization using deep learning Regions with Convolutional Neural Network Features (R-CNN) Semantic Segmentation using SegNet
- 26. 28 Example 3: Analyzingsignal data using deep learning Signal Classification using LSTMs Speech Recognition using CNNs
- 27. 29 Why deep learning? FashionMNIST: The “Hello, World!” of deep learning Transfer learning with CNNs (optional) Semantic segmentation (optional) Deep learning with time series data Ground Truth Labeling for datasets Everything else in deep learning… Agenda
- 28. 30 Diverse Applications ofDeep Learning Iris Recognition – 99.4% accuracy1 1. Source: An experimental study of deep convolutional features for iris recognition Signal Processing in Medicine and Biology Symposium (SPMB), 2016 IEEE Shervin Minaee ; Amirali Abdolrashidiy ; Yao Wang; An experimental study of deep convolutional features for iris recognition 2. "A Real-Time Pedestrian Detector using Deep Learning for Human-Aware Navigation" David Ribeiro, Andre Mateus, Jacinto C. Nascimento, and Pedro Miraldo 3. Deep Joint Rain Detection and Removal from a Single Image" Wenhan Yang, Robby T. Tan, Jiashi Feng, Jiaying Liu, Zongming Guo, and Shuicheng Yan Human Aware Navigation for Robots2 Rain Detection and Removal3 MatConvnet MatCaffe MatConvnet
- 29. 31 Can you tellthe difference? Japanese or Blenheim Spaniel? Blenheim Spaniel Japanese Spaniel Source: ILSVRC ImageNet dataset
- 30. 32 Why is DeepLearning So Popular Now? Source: ILSVRC Top-5 Error on ImageNet Human Accuracy
- 31. 33 Deep Learning Enablers •Labeled public datasets • Increased GPU acceleration • World-class models AlexNet PRETRAINED MODEL Caffe I M P O R T E R ResNet-50 PRETRAINED MODEL TensorFlow- Keras I M P O R T E R VGG-16 PRETRAINED MODEL GoogLeNet PRETRAINED MODEL ResNet-101 PRETRAINED MODEL Inception-v3 M O D E L S
- 32. 34 Machine Learning vsDeep Learning Deep learning performs end-to-end learning by learning features, representations and tasks directly from images, text and sound Deep learning algorithms also scale with data – traditional machine learning saturates Machine Learning Deep Learning
- 33. 35 Deep Learning Workflow Files Databases Sensors ACCESSAND EXPLORE DATA DEVELOP PREDICTIVE MODELS Hardware-Accelerated Training Hyperparameter Tuning Network Visualization LABEL AND PREPROCESS DATA Data Augmentation/ Transformation Labeling Automation Import Reference Models INTEGRATE MODELS WITH SYSTEMS Desktop Apps Enterprise Scale Systems Embedded Devices and Hardware
- 34. 36 Why deep learning? FashionMNIST: The “Hello, World!” of deep learning Transfer learning with CNNs (optional) Semantic segmentation (optional) Deep learning with time series data Ground Truth Labeling for datasets Everything else in deep learning… Agenda
- 35. 37 Fashion-MNIST Dataset What? A collectionif items such as bags, shoes, etc. Why? Benchmark machine learning algorithms How many? 60,000 training images 10,000 test images Best results? 96.3% accuracy Sources: https://github.com/zalandoresearch/fashion-mnist
- 36. 38 Convolutional Neural Networks(CNN) “Deep” in deep learning refers to number of layers
- 37. 39 Long Short TermMemory Networks ▪ Recurrent Neural Network that carries a memory cell throughout the process ▪ Sequence Problems c0 C1 Ct
- 38. 40 Types of Datasets Numeric Data TimeSeries/ Text Data Image Data Machine Learning or LSTM LSTM or CNN CNN
- 39. 41 Why deep learning? FashionMNIST: The “Hello, World!” of deep learning Transfer learning with CNNs (optional) Semantic segmentation (optional) Deep learning with time series data Ground Truth Labeling for datasets Everything else in deep learning… Agenda
- 40. 42 Why deep learning? FashionMNIST: The “Hello, World!” of deep learning Transfer learning with CNNs (optional) Semantic segmentation (optional) Deep learning with time series data Ground Truth Labeling for datasets Everything else in deep learning… Agenda
- 41. 43 Two Approaches forDeep Learning 2. Fine-tune a pre-trained model (transfer learning) 1. Train a Deep Neural Network from Scratch
- 42. 44 Transfer Learning Workflow Earlylayers learn low- level features (edges, blobs, colors) Last layers learn task- specific features 1 million images 1000s classes Load pretrained network
- 43. 45 Transfer Learning Workflow Earlylayers that learned low-level features (edges, blobs, colors) Last layers that learned task specific features 1 million images 1000s classes Load pretrained network Fewer classes Learn faster New layers learn features specific to your data Replace final layers
- 44. 46 Transfer Learning Workflow Earlylayers that learned low-level features (edges, blobs, colors) Last layers that learned task specific features 1 million images 1000s classes Load pretrained network Fewer classes Learn faster New layers to learn features specific to your data Replace final layers 100s images 10s classes Training images Training options Train network
- 45. 47 Fewer classes Learn faster Newlayers learn features specific to your data Replace final layers Transfer Learning Workflow Early layers that learned low-level features (edges, blobs, colors) Last layers that learned task specific features 1 million images 1000s classes Load pretrained network Fewer classes Learn faster New layers to learn features specific to your data Replace final layers 100s images 10s classes Training images Training options Train network Test images Trained Network Predict and assess network accuracy
- 46. 48 Transfer Learning Workflow Earlylayers that learned low-level features (edges, blobs, colors) Last layers that learned task specific features 1 million images 1000s classes Load pretrained network Fewer classes Learn faster New layers to learn features specific to your data Replace final layers 100s images 10s classes Training images Training options Train network Test images Trained Network Predict and assess network accuracy Test images Trained Network Predict and assess network accuracy Probability Boat Plane Car Train Deploy results
- 47. 49 Transfer Learning Workflow Probability Boat Plane Car Train Deployresults Early layers that learned low-level features (edges, blobs, colors) Last layers that learned task specific features 1 million images 1000s classes Load pretrained network Fewer classes Learn faster New layers to learn features specific to your data Replace final layers 100s images 10s classes Training images Training options Train network Test images Trained Network Predict and assess network accuracy
- 48. 50 Example: Food classifierusing deep transfer learning 5 Category Classifier Caesar salad French fries Burgers Pizza Sushi
- 49. 51 Transfer Learning Workflow Probability Boat Plane Car Train Deployresults Early layers that learned low-level features (edges, blobs, colors) Last layers that learned task specific features 1 million images 1000s classes Load pretrained network Fewer classes Learn faster New layers to learn features specific to your data Replace final layers 100s images 10s classes Training images Training options Train network Test images Trained Network Predict and assess network accuracy
- 50. 52 Why Perform TransferLearning ▪ Requires less data and training time ▪ Reference models (like AlexNet, VGG-16, VGG-19, Inception-v3) are great feature extractors ▪ Leverage best network types from top researchers (list of all models) AlexNet PRETRAINED MODEL Caffe I M P O R T E R ResNet-50 PRETRAINED MODEL TensorFlow- Keras I M P O R T E R VGG-16 PRETRAINED MODEL GoogLeNet PRETRAINED MODEL ResNet-101 PRETRAINED MODEL Inception-v3 M O D E L S
- 51. 53 Import the LatestModels for Transfer Learning Pretrained Models* ▪ AlexNet ▪ VGG-16 ▪ VGG-19 ▪ GoogLeNet ▪ Inception-v3 ▪ ResNet50 ▪ ResNet-101 ▪ Inception-resnet-v2 ▪ SqueezeNet ▪ MobileNet (coming soon) * single line of code to access model Import Models from Frameworks ▪ Caffe Model Importer ▪ TensorFlow-Keras Model Importer ▪ ONNX - Importer/ Exporter AlexNet PRETRAINED MODEL Caffe I M P O R T E R ResNet-50 PRETRAINED MODEL TensorFlow- Keras I M P O R T E R VGG-16 PRETRAINED MODEL GoogLeNet PRETRAINED MODEL ResNet-101 PRETRAINED MODEL Inception-v3 M O D E L S
- 52. 54 Interoperability with DeepLearning Frameworks ▪ Import and export models using the Open Neural Network Exchange (ONNX) format ▪ Model importers – Caffe – TensorFlow-Keras ▪ Access pretrained models with a single line of code Sep ‘17 Mar ‘18 Mar ‘17 Sep ‘16 Sep ‘18 ONNX PyTorch MATLAB MXNet Caffe2 TensorFlow Chainer Core ML Cognitive Toolkit Keras- Tensorflow Caffe
- 53.
- 54. 56 What we coveredin this section… ▪ Transfer learning – Easily modify existing networks with one-line commands ▪ Access large datasets – Using the imageDatastore ▪ Visualize and Analyze Networks – Using the Network Analyzer ▪ Make training faster – Freezing the layers ▪ Improve training results – Data augmentation with augmentedImageDataStore – Automatic parameter selection with Bayesian hyperparameter tuning
- 55. 58 Why deep learning? FashionMNIST: The “Hello, World!” of deep learning Transfer learning with CNNs (optional) Semantic segmentation (optional) Deep learning with time series data Ground Truth Labeling for datasets Everything else in deep learning… Agenda
- 56. 59 Why deep learning? FashionMNIST: The “Hello, World!” of deep learning Transfer learning with CNNs (optional) Semantic segmentation (optional) Deep learning with time series data Ground Truth Labeling for datasets Everything else in deep learning… Agenda
- 57. 60 What is semanticsegmentation?
- 58. 61 Original Image ROI detection Pixelclassification
- 59. 62 Semantic Segmentation CamVid Dataset 1.Segmentation and Recognition Using Structure from Motion Point Clouds, ECCV 2008 2. Semantic Object Classes in Video: A High-Definition Ground Truth Database ,Pattern Recognition Letters
- 60.
- 61.
- 62. 65 Useful Tools forSemantic Segmentation ▪ Automatically create network structures – Using segnetLayers and fcnLayers ▪ Handle pixel labels – Using the pixelLabelImageDatastore and pixelLabelDatastore ▪ Evaluate network performance – Using evaluateSemanticSegmentation ▪ Examples and tutorials to learn concepts
- 63. 66 Why deep learning? FashionMNIST: The “Hello, World!” of deep learning Transfer learning with CNNs (optional) Semantic segmentation (optional) Deep learning with time series data Ground Truth Labeling for datasets Everything else in deep learning… Agenda
- 64. 67 Why deep learning? FashionMNIST: The “Hello, World!” of deep learning Transfer learning with CNNs (optional) Semantic segmentation (optional) Deep learning with time series data Ground Truth Labeling for datasets Everything else in deep learning… Agenda
- 65. 68 New Product: TextAnalytics Toolbox ▪ Text extraction from PDF and Microsoft Word files ▪ Text preprocessing and normalization ▪ TF-IDF and word frequency statistics ▪ Machine learning algorithms, including Latent Dirichlet Allocation (LDA) and Latent Semantic Analysis (LSA) ▪ Word-embedding training, and pretrained model import with word2vec, FastText, and GloVe ▪ Word cloud and text scatter plots Analyze and model text data
- 66. 70 Time Series Classification(Human Activity Recognition) Long short-term memory networks ▪ Dataset is accelerometer and gyroscope signals captured with a smartphone ▪ Data is a collection of time series with 9 channels Deep Learning sequence to one
- 67. 71 LSTM sequences Sequence 1Sequence 2 Sequence (n-1) 1 2 3 4 5 6 7 8 9 2.5 sec 2.5 sec 2.5 sec ……………………………........... ……………………………........... Standing Walking Sitting Time
- 68. 72 Why deep learning? FashionMNIST: The “Hello, World!” of deep learning Transfer learning with CNNs (optional) Semantic segmentation (optional) Deep learning with time series data Ground Truth Labeling for datasets Everything else in deep learning… Agenda
- 69. 73 Why deep learning? FashionMNIST: The “Hello, World!” of deep learning Transfer learning with CNNs (optional) Semantic segmentation (optional) Deep learning with time series data Ground Truth Labeling for datasets Everything else in deep learning… Agenda
- 70. 74 “I love tolabel and preprocess my data” ~ Said no engineer, ever.
- 71. 75 Data Apps for Labeling “Howdo I label my data?” New App for Ground Truth Labeling Label pixels and regions for semantic segmentation
- 72.
- 73. 77 Why deep learning? FashionMNIST: The “Hello, World!” of deep learning Transfer learning with CNNs (optional) Semantic segmentation (optional) Deep learning with time series data Ground Truth Labeling for datasets Everything else in deep learning… Agenda
- 74. 78 Why deep learning? FashionMNIST: The “Hello, World!” of deep learning Transfer learning with CNNs (optional) Semantic segmentation (optional) Deep learning with time series data Ground Truth Labeling for datasets Everything else in deep learning… Agenda
- 75. 79 Deploying Deep LearningModels for Inference GPU Coder Deep Learning Networks NVIDIA TensorRT & cuDNN Libraries ARM Compute Library Intel MKL-DNN Library
- 76. 80 Semantic Segmentation Running inMATLAB Generated Code from GPU Coder
- 77. 81 Desktop CPU Raspberry Piboard Deploying to CPUs GPU Coder Deep Learning Networks NVIDIA TensorRT & cuDNN Libraries
- 78. 82 Training Performance andScalability
- 79. 84 Deep Learning onCPU, GPU, Multi-GPU and Clusters Single CPU Single CPU Single GPU HOW TO TARGET? Single CPU, Multiple GPUs On-prem server with GPUs Cloud GPUs (AWS)
- 80. 85 Inference Performance andDeployment
- 81. 86 Using GPU Coderfor Deep Learning Access Data Preprocess Select Network Train Image Acquisition Tbx Image Processing Tbx Computer Vision System Tbx Neural Network Tbx Parallel Computing Tbx GPU Coder Deploy
- 82. 87 Using CUDA v9 andcuDNN v7 The inference is for single image performance.
- 83. 89 Overview of deeplearning deployment options “How do I deploy my model?” Deploy / Share ▪ Create Desktop Apps ▪ Run Enterprise Solution ▪ Generate C and C++ Code GPU Coder ▪ Target GPUs Introducing: GPU Coder- Convert to NVIDIA CUDA code ▪ Generate C and C++ Code
- 84.
- 85. 91 Designing and BuildingDeep Learning Models ▪ Edit and build deep networks (Deep Network Designer app) ▪ Visualize, analyze, and find problems in network architectures before training (Network Analyzer) ▪ Automate ground-truth labeling using apps – Image Labeler app – Video Labeler app – Audio Labeler app ▪ Monitor training progress with plots for accuracy, loss, validation metrics, and more ▪ Visualize and debug deep learning models + Deep Learning Toolbox Computer Vision System Toolbox Audio System Toolbox
- 86. 92 Define new deepneural network layers
- 87. 93 Object Detection Frameworks ▪Single line of code to train a detector ▪ Includes: – R-CNN – Fast R-CNN – Faster R-CNN
- 88. 94 Deep learning featuresoverview ▪ Classification ▪ Regression * ▪ Semantic segmentation ▪ Object detection * ▪ Scalability * – Multiple GPUs – Cluster or cloud ▪ Custom network layers * ▪ Import models * – Caffe – Keras/TensorFlow ▪ Data augmentation * ▪ Hyperparameter tuning * – Bayesian optimization ▪ Python MATLAB interface * ▪ LSTM networks * – Time series, signals, audio ▪ Custom labeling * – API for ground-truth labeling automation – Superpixels ▪ Data validation * – Training and testing * We can cover in more detail outside this presentation
- 89.
- 90. 96 What is ReinforcementLearning? ▪ What is Reinforcement Learning? – Type of machine learning that trains an ‘agent’ through repeated interactions with an environment ▪ How does it work? – Through a trial & error process that maximizes success
- 91. 97 Reinforcement Learning Applications ▪Why should you care about Reinforcement Learning? – It enables the use of deep learning for controls and decision-making applications Game Play Controls Robotics Autonomous driving
- 92. 102 Reinforcement Learning Toolbox Newin R2019a ▪ Built-in and custom algorithms for reinforcement learning ▪ Environment modeling in MATLAB and Simulink ▪ Deep Learning Toolbox support for designing policies ▪ Training acceleration through GPUs and cloud resources ▪ Deployment to embedded devices and production systems ▪ Reference examples for getting started
- 93.
- 94. 106 Learn More: MachineLearning Documentation: mathworks.com/machine-learning
- 95. 107 Learn More: BigData ▪ MATLAB Documentation – Strategies for Efficient Use of Memory – Resolving "Out of Memory" Errors ▪ Big Data with MATLAB – www.mathworks.com/discovery/big-data-matlab.html ▪ Tall Arrays in Action – https://www.mathworks.com/videos/matlab-tall-arrays-in-action-122883.html
- 96. 108 Get Training Accelerate yourlearning curve: - Customized curriculum - Learn best practices - Practice on real-world examples Options to fit your needs: - Self-paced (online) - Instructor led (online and in-person) - Customized curriculum (on-site)
- 97. 109 MATLAB Answers Blogs Cody File Exchange and more… Answers Blogs ThingSpeak Every month,over 2 million MATLAB & Simulink users visit MATLAB Central to get questions answered, download code and improve programming skills. MATLAB Central Community MATLAB Answers: Q&A forum; most questions get answered in only 60 minutes File Exchange: Download code from a huge repository of free code including tens of thousands of open source community files Cody: Sharpen programming skills while having fun Blogs: Get the inside view from Engineers who build and support MATLAB & Simulink ThingSpeak: Explore IoT Data And more for you to explore… Learn Connect Contribute
- 98. 110 Resources ▪ Machine LearningIntro Tech talks ▪ Machine Learning with MATLAB: – Overview – Cheat sheet – Introductory eBook – Mastering Machine Learning eBook – Try the Classification Learner App in a browser ▪ Deep learning onramp course
- 99. 111 MathWorks can helpyou do Deep Learning ▪ Guided evaluations with a MathWorks deep learning engineer ▪ Proof-of-concept projects ▪ Deep learning hands-on workshop ▪ Seminars and technical deep dives ▪ Deep learning onramp course ▪ Consulting services ▪ Training courses ▪ Technical support ▪ Advanced customer support ▪ Installation, enterprise, and cloud deployment ▪ MATLAB for Deep Learning Free resources More options
- 100.
- 101. 113 © 2019 TheMathWorks, Inc.