Youtor
5 AWS SageMaker Algorithms amp Their Use Cases
>> YOUR LINK HERE: ___ http://youtube.com/watch?v=PlX23jkf32Y
In this video, I will describe 5 SageMaker Built-in algorithms and their use cases. • 1. SAGEMAKER LINEAR LEARNER (CLASSIFICATION/REGRESSION) • Linear Learner is a supervised machine learning algorithm in SageMaker that is used to fit a line to the training data. • Linear Learner could be used for both classification and regression tasks • Linear Learner is simple and fast compared to other training algorithms. • Application: The algorithm could be used to predict continuous outputs such as employee salary. It could also be used to perform classification tasks such as predicting whether a patient has diabetes or not. • 2. SAGEMAKER XGBOOST (CLASSIFICATION/REGRESSION) • XGBoost or Extreme Gradient Boosting algorithm is one of the most famous and powerful algorithms in SageMaker to perform both regression and classification tasks. • XGBoost is a supervised learning algorithm and implements gradient boosted trees algorithm. The algorithm work by combining an ensemble of predictions from several weak models. • Recently, XGBoost is the go-to algorithm for most developers and has won several Kaggle competitions. • Application: The algorithm could be used for classification such as predict whether a credit card transaction is fraudulent or not. It could also be used to perform regression to predict sales or demand. • 3. PRINCIPAL COMPONENT ANALYSIS (PCA) (UNSUPERVISED/DIMENSIONALITY REDUCTION) • PCA is an unsupervised machine learning algorithm that is used for dimensionality reduction. The algorithm work by mapping the original features to fewer number of “components” while preserving the original information contained in the data. • Application: PCA can be used for dimensionality reduction applications such as image compression. • PCA could also be used to spot patterns in high dimensional datasets such as analyzing bank customer data. • 4. DEEPAR ALGORITHM (TIME-SERIES FORECASTING) • Amazon SageMaker DeepAR is a one-dimensional time series forecasting algorithm that works in supervised fashion and uses Recurrent Neural Networks (RNN). • One key feature of DeepAR is the ability to train a single time series forecasting model based on several time series. • DeepAR can work well in challenging problems when companies are introducing totally new product to the market with not history (Cold Start Problem). • Application: DeepAR algorithm could be leveraged to predict future sales so retailers can stock their products and plan accordingly. DeepAR could be used to generate point forecast (Number of products to be sold next month is 100,000) and probabilistic forecast (Number of products to be sold next month is between 30,000 and 100,000 with 85% probability). • 5. SAGEMAKER IMAGE CLASSIFICATION • Image classification algorithm is a supervised machine learning algorithm that can classify multi-class images. SageMaker image classification algorithm uses convolutional neural network known as ResNet to perform classification. • The network could be trained from scratch or starting from a pretrained network using transfer learning. The algorithm does not specify the location of the object in the image so there is no bounding box. • Application: SageMaker image classification algorithm can be used for self driving cars to take in an image as an input and generate the corresponding label as an output such as: Stop Sign, deer crossing, yield,..etc • I hope you'll enjoy this video! Please subscribe for more videos. • Thanks, • Ryan • #Sagemaker #AWS #Machinelearning #pca #deepar #linearlearner
#############################
