Applying machine learning to problems can be a difficult task because of all the different models that are offered. Discover how to evaluate and select machine learning models and apply machine learning to a problem.
There are numerous options available to scale and encode features and labels in data sets to get the best out of machine learning (ML) algorithms. In this 10-video course, explore techniques such as standardizing, nomalizing, and one-hot encoding. Learners begin by learning how to use Pandas library to load a data set in the form of a CSV file and perform exploratory analysis on its features. Then use scikit-learn's Binarizer to transform the continuous data in a series to binary values; apply the MiniMaxScaler on a data set to get two similar columns to have the same range of values; and standardize multiple columns in data sets with scikit-learn's StandardScaler. Examine differences between the Normalizer and other scaling techniques, and learn how to represent values in a column as a proportion of the maximum absolute value by using the MaxAbScaler. Finally, discover how to use Pandas library to one-hot encode one or more features of your data set and distinguish between this technique and label encoding. The concluding exercise involves building ML training sets.
This 7-video course helps learners discover how to implement machine learning scaling techniques such as standardizing and min-max scaling on continuous data and one-hot encoding on categorical features to improve performance of linear regression models. In the first tutorial, you will use Pandas library to load a CSV file into a data frame and analyze its contents by using Pandas and Matplotlib. You will then learn how to create a linear regression model with scikit-learn to predict the sale price of a house and evaluate this model by using metrics such as mean squared error and r-square. Next, learners will examine the application of min-max scaling on continuous fields and one-hot encoding on the categorical columns of a data set. Then analyze effects of preprocessing by recognizing benefits of scaling and encoding data sets by evaluating the performance of a regression model built with preprocessed data. Also, learn how to use scikit-learn's StandardScaler on a data set's continuous features and compare its effects with that of min-max scaling. The concluding exercise involves preprocessing data for regression.
In this course, learners can explore how to implement machine learning scaling techniques such as standardizing and normalizing on continuous data and label encoding on the target, in order to get the best out of machine learning algorithms. Examine dimensionality reduction by using Principal Component Analysis (PCA). Start this 6-video course by using Pandas library to load a CSV data set into a data frame and scale continuous features by using a standard scaler. You will then learn how to build and evaluate a support vector classifier in scikit-learn; use Pandas and Seaborn to generate a heat map; and spot the correlations between features in a data set. Discover how to apply the technique of PCA to reduce the number of dimensions in your input data and obtain the explained variance of each principal component. In the course's final tutorial, you will explore how to apply normalization and PCA on data sets and build a classification model with the principal components of scaled data. The concluding exercise involves processing data for classification.
From self-driving cars to predicting stock prices, machine learning has an exciting range of applications. BigML, due to its ease of use, makes these algorithms widely accessible. This course outlines machine learning fundamentals and how these are applied in BigML. You'll start by examining various machine learning algorithm categories and the kinds of problems they're used to solve. You'll then investigate the classification problem and the process involved in training and evaluating such models. Next, you'll examine linear regression and how this can help predict a continuous value. Moving on, you'll explore the concept of unsupervised learning and its application in clustering, Principal Component Analysis (PCA), and generating associations. Finally, you'll recognize how all of this comes together when using BigML to significantly simplify the building and maintenance of your machine learning models.
BigML not only provides ease-of-use, but it also offers flexibility in how you work with your data. This course serves as a hands-on introduction to BigML and its vast array of features. You'll start by exploring the different ways data can be loaded into the platform and how these can be transformed into datasets to train and test a machine learning model. You'll gain practical experience with some of the tools available to help you better understand your data - from histograms and scatterplots to visualizations of value distribution. Moving on, you'll build a fundamental classification model, a decision tree, which takes employee details and predicts whether they'll stay or leave in the next year. Finally, you'll investigate some possible configurations for this model.
Examine fundamentals of machine learning (ML) and how Pandas ML can be used to build ML models in this 7-video course. The working of Support Vector Machines to perform classification of data are also covered. Begin by learning about different kinds of machine learning algorithms, such as regression, classification, and clustering, as well as their specific applications. Then look at the process involved in learning relationships between input and output during the training phase of ML. This leads to an introduction to Pandas ML, and the benefits of combining Pandas, scikit-learn, and XGBoost into a single library to ease the task of building and evaluating ML models. You will learn about Support Vector Machines, which are a supervised machine learning algorithm, and how they are used to find a hyperplane to divide data points into categories. Learners then study the concept of overfitting in machine learning, and the problems associated with a model overfitted to training data. and how to mitigate the issue. The course concludes with an exercise in machine learning and classification.
In this 8-video course, explore the fundamentals of regression and clustering and discover how to use a confusion matrix to evaluate classification models. Begin by examining application of a confusion matrix and how it can be used to measure the accuracy, precision, and recall of a classification model. Then study an introduction to regression and how it works. Next, take a look at the characteristics of regression such as simplicity and versatility, which have led to widespread adoption of this technique in a number of different fields. Learn to distinguish between supervised learning techniques such as regression and classifications, and unsupervised learning methods such as clustering. You will look at how clustering algorithms are able to find data points containing common attributes and thus create logical groupings of data. Recognize the need to reduce large data sets with many features into a handful of principal components with the PCA (Principal Component Analysis) technique. Finally, conclude the course with an exercise recalling concepts such as precision and recall, and use cases for unsupervised learning.
Classification, regression, and clustering are some of the most commonly used machine learning (ML) techniques and there are various algorithms available for these tasks. In this 10-video course, learners can explore their application in Pandas ML. First, examine how to load data from a CSV (comma-separated values) file into a Pandas data frame and prepare the data for training a classification model. Then use the scikit-learn library to build and train a LinearSVC classification model and evaluate its performance with available model evaluation functions. You will explore how to install Pandas ML and define and configure a ModelFrame, then compare training and evaluation in Pandas ML with equivalent tasks in scikit-learn. Learn how to build a linear regression model by using Pandas ML. Then evaluate a regression model by using metrics such as r-square and mean squared error, and visualize its performance with Matplotlib. Work with ModelFrames for feature extraction and encoding, and configure and build a clustering model with the K-Means algorithm, analyzing data clusters to determine unique characteristics. Finally, complete an exercise on regression, classification, and clustering.
In this 12-video course, you will examine the different uses of data science tools and the overall platform, as well as the benefits and challenges of machine learning deployment. The first tutorial explores what automation is and how it is implemented. This is followed by a look at the tasks and processes best suited for automation. This leads learners into exploring automation design, including what Display Status is, and also the Human-Computer Collaboration automation design principle. Next, you will examine the Human Intervention automation design principle; automated testing in software design and development; and also the role of task runners in software design and development. Task runners are used to automate repeatable tasks in the build process. Delve into DevOps and automated deployment in software design, development, and deployment. Finally, you will examine process automation using robotics, and in the last tutorial in the course, recognize how modern robotics and AI designs are applied. The concluding exercise involves recognizing automation and robotics design application.
In this 19-video course, learners explore deep learning with Keras, including how to create and use neural networks with Keras for machine learning solutions. Begin with an overview of what neural networks are and their main components, followed by an introduction to Keras and its guiding principles. Observe how to configure Microsoft Cognitive Toolkit (CNTK) as your Keras backend; install and configure Keras; identify and work with both types of models available in Keras; and recognize features of commonly-used Keras layers and when to use them. Use Keras to make regression classifications and image classifications; Keras metrics to judge a model's performance; and Jupyter Notebooks with Keras. Next, download and load a data set from MNIST or CIFAR-10; explore data sets in Keras; prepare your data in Keras by defining input and target tensors, and compile the model in Keras. Then train and test your neural network; evaluate and score the performance of neural networks in Keras, and make predictions using your data set in Keras. The closing exercise involves using a neural network to make predictions.
This course explores research being done in machine learning and deep learning. Topics covered include neural networks and deep neural networks. First, learners examine how to prevent neural networks from overfitting. You will explore research on multilabel learning algorithms, multilabel classification, and multiple-output classifications, which are variants of the standard classification problem. Then examine deep learning algorithms, the enhanced performance of deeper neural networks that are more adept at automatic feature extraction. Next, ut facial alignment, regression tree ensembles, and deep features for scene recognition. Review ELM (Extreme Learning Machine), and how it is used to perform regression and multiclass classification.
Explore machine learning reinforcement learning, along with the essential components of reinforcement learning that will assist in the development of critical algorithms for decisionmaking, in this 10-video course. You will examine how to achieve continuous improvement in performance of machines or programs over time, along with key differences between reinforcement learning and machine learning paradigm. Learners will observe how to depict the flow of reinforcement learning by using agent, action, and environment. Next, you will examine different scenarios of state changes and transition processes applied in reinforcement learning. Then examine the reward hypothesis, and learn to recognize the role of rewards in reinforcement learning. You will learn that all goals can be described by maximization of the expected cumulative rewards. Continue by learning the essential steps applied by agents in reinforcement learning to make decisions. You will explore the types of reinforcement learning environments, including deterministic, observable, discrete or continuous, and single-agent or multi-agent. Finally, you will learn how to install OpenAI Gym and OpenAl Universe.
This 9-video course explores how to implement machine learning reinforcement learning by examining the terminology, including agents, the environment, state, and policy. This course demonstrates how to implement reinforcement learning by using Keras and Python; how to ensure that you can build a model; and how to launch and use Ubuntu, and VI editor to do score calculations. First, learn the role of the Markov decision process in which the agent observes the environment, with output consisting of a reward and the next state, and then acts upon it. You will explore Q-learning, a model-free reinforcement learning technique, an asynchronous dynamic programming approach, and will learn about the Q-learning rule, and Deep Q-learning. Next, learn the steps to install TensorFlow for reinforcement learning, as well as framework, which is used for reinforcement learning provided by OpenAI. Then learn how to implement TensorFlow for reinforcement learning. Finally, you will learn to implement Q-learning using Python, and then utilize capabilities of OpenAl Gym and FrozenLake.
Microsoft Cognitive Toolkit provides powerful machine learning and deep learning algorithms for developing AI. Knowing which problems are easier to solve using Microsoft CNTK over other frameworks helps AI practitioners decide on the best software stack for a given application. In this course, you'll explore advanced techniques for working with Microsoft CNTK and identify which cases benefit most from MS CNTK. You'll examine how to load and use external data using CNTK and how to use its imperative and declarative APIs. You'll recognize how to carry out common AI development tasks using CNTK, such as working with epochs and batch sizes, model serialization, model visualization, feedforward neural networks, and machine learning model evaluation. Finally, you'll implement a series of practical AI projects using Python and MS CNTK.
Microsoft Cognitive Toolkit (CNTK) is an open source framework for distributed deep learning suitable for commercial applications. It's primarily used to develop neural networks but can also be used for machine learning and cognitive computing. It supports multiple languages and can easily be used in the cloud. These factors make CNTK a good fit for various AI projects. In this course, you'll explore the basic concepts required to work with Microsoft CNTK. You'll compare other frameworks with CNTK, examine the process of creating machine learning and deep learning models with CNTK, and learn how it can be used with several cloud services. You'll move on to learn where to access CNTK documentation, community, and installation guidelines. Finally, you'll use CNTK to predict diabetes using retina scans.
Explore artificial neural networks (ANNs), their essential components, tools, and frameworks for their implementation in machine learning solutions. In this 9-video course, you will discover recurrent neural networks (RNNs) and how they are implemented. Key concepts covered here include perceptrons and the computational role they play in ANNs; learning features and characteristics of ANNs and how components are used to build a model; and learning prominent tools and frameworks used to implement sequence models and ANNs. Next, you will learn about sequence modeling as it pertains to language models; RNNs and their capabilities and components; and how to specify RNN types and their implementation features. Learners will then explore the concept of linear and nonlinear functions and classify how they are used with perceptrons; explore the concept of backpropagation and usage of backpropagation algorithm in neural networks; and examine the concept of activation functions and how linear and nonlinear activations are utilized in neural networks. Finally, you will see how to implement perceptrons with Python, and how to use modeling tools and architectures and applications of sequence models.
In this 7-video course, learners can explore the concepts of language modeling, natural language processing (NLP), and sequence generation for NLP. Prominent machine learning modeling algorithms such as vanishing gradient problem, gated recurrent units (GRUs), and long short-term memory (LSTM) network are also covered. Key concepts studied in this course include language models, one of the most important parts of NLP. and how to implement NLP along with its essential components; learning the process and approach of generating sequence for NLP; and vanishing gradient problem implementation approaches to overcome the problem of taking longer times to achieve convergence. Then, learn about features and characteristics of GRUs used to resolve issues with vanishing gradient problems, and learn the problems and drawbacks of implementing short-term memory and LSTM as modeling solutions. In the concluding exercise, learners will review the essential components and prominent applications of language modeling and specify some of the solutions for vanishing gradient problems.
Explore the concept of artificial neural networks (ANNs) and components of neural networks, and examine the concept of learning and training samples used in supervised, unsupervised, and reinforcement learning in this 10-video course. Other topics covered in this course include network topologies, neuron activation mechanism, training sets, pattern recognition, and the need for gradient optimization procedure for machine learning. You will begin the course with an overview of ANN and its components, then examine the artificial network topologies that implement feedforward, recurrent, and linked networks. Take a look at the activation mechanism for neural networks, and the prominent learning samples that can be applied in neural networks. Next, compare supervised learning samples, unsupervised learning samples, and reinforcement learning samples, and then view training samples and the approaches to building them. Explore training sets and pattern recognition and, in the final tutorial, examine the need for gradient optimization in neural networks. The exercise involves listing neural network components, activation functions, learning samples, and gradient descent optimization algorithms.
Learners will examine the concepts of perception, layers of perception, and backpropagation, and discover how to implement recurrent neural network by using Python, TensorFlow, and Caffe2 in this 10-video course. Begin by taking a look at the essential features and processes of implementing perception and backpropagation in machine learning neural networks. Next, you will compare single-layer perception and multilayer perception and describe the need for layer management. You will learn about the steps involved in building recurrent neural network models; building recurrent neural networks with Python and TensorFlow; implementing long short-term memory (LSTM) by using TensorFlow, and building recurrent neural networks with Caffe2. Caffe is a deep learning framework. Building deep learning language models using Keras-an open source neural network library-will be explored in the final tutorial of the course. The concluding exercise entails implementing recurrent neural networks by using TensorFlow and Caffe2 and building deep learning language models by using Keras.
Learners can explore the concepts of convolutional neural network (CNN); the underlying architecture, principles, and methods needed to build a CNN; and its implementation in a deep neural network. In this 12-video course, you will examine visual perception, and the ability to interpret the surrounding environment by using light in the visible spectrum. First, learn about CNN architecture; how to analyze the essential layers; and the impact of an initial choice of layers. Next, you will learn about nonlinearity in the first layer, and the need for several pooling techniques. Then learn how to implement a convolutional layer and sparse interaction. Examine the hidden layers of CNN, which are convolutional layers, ReLU (rectified linear unit) layers, or activation functions, the pooling layers, the fully connected layer, and the normalization layer. You will examine machine learning semantic segmentation to understand an image at the pixel level, and its implementation using Texton Forest and a random based classifier. Finally, this course examines Gradient Descent and its variants.
This course explores machine learning convolutional neural networks (CNNs), which are popular for implementation in image and audio processing. Learners explore AI (artificial intelligence), and the issues surrounding implementation, how to approach organizational talent and strategy, and how to prepare for AI architecture in this 8-video course. You will learn to use the Google Colab tool, and to implement image recognition classifier by using CNN, Keras, and TensorFlow. Next, learn to install and implement a model, and use it for image classification. You will examine the artificial neural network ResNet (residual neural network), and how it builds on constructs known from pyramidal cells and cerebral cortex. You will also study PyTorch, an open-source machine learning library that enables fast, flexible experimentation, and efficient production through a hybrid front end, and learn to use the PyTorch ecosystem tool to develop and implement neural networks. Finally, this course demonstrates how to create a data set by using Training CNN by using PyTorch to categorize garments.
Learners can explore fundamental concepts of biological and artificial neural networks, computational models that can be implemented with neural networks, and how to implement neural networks with Python, in this 12-video course. Begin with a look at characteristics of machine learning biological neural networks that inspired artificial neural networks. Then explore components of biological neural networks and the signal processing mechanism. Next, take a look at the essential components of the structure of artificial neural networks; learn to recognize the layered architecture of neural networks; and observe how to classify various computational models that can be implemented by using neural networks paradigm. Examine neurons connectivity, by describing the interconnection between neurons involving weights and fixed weights. This leads on to threshold functions in neural networks and the basic logic gates of AND, OR, and XNOR. Implement neural networks by using Python and the core libraries provided by Python for neural networks; create a neural network model using Python, Keras, and TensorFlow, and finally, view prominent neural network use cases. The concluding exercise involves implementing neural networks.
Discover the basics of perceptrons, including single- layer and multilayer, and the roles of linear and nonlinear functions in this 10-video course. Learners will explore how to implement perceptrons and perceptron classifiers by using Python for machine learning solutions. Key concepts covered in this course include perceptrons, single-layer and multilayer perceptrons, and the computational role they play in artificial neural networks; learning the algorithms that can be used to implement single-layer perceptron training models; and exploring multilayer perceptrons and illustrating the algorithmic difference from single-layer perceptrons. Next, you will learn to classify the role of linear and nonlinear functions in perceptrons; learn how to implement perceptrons by using Python; and learn approaches and benefits of using the backpropagation algorithm in neural networks. Then learn the uses of linear and nonlinear activation functions in artificial neural networks; learn to implement a simple perceptron classifier using Python; and learn the benefits of using the backpropagation algorithm in neural networks and implement perceptrons and perceptron classifiers by using Python.
In this 13-video course, learners can explore how to work with machine learning frameworks and Python to implement training algorithms for neural networks. You will learn the concept and characteristics of perceptrons, a single layer neural network that aggregates the weighted sum of inputs, and returns either zero or one, and neural networks. You will then explore some of the prominent learning rules that to apply in neural networks, and the concept of supervised and unsupervised learning. Learn several types of neural network algorithms, and several training methods. Next, you will learn how to prepare and curate data by using Amazon SageMaker, and how to implement an artificial neural network training process using Python, and other prominent and essential learning algorithms to train neural networks. You will learn to use Python to train artificial neural networks, and how to use Backpropagation in Keras to implement multilayer perceptrons or neural networks. Finally, this course demonstrates how to implement regularization in multilayer perceptrons by using Keras.
This 15-video course explores how to design advanced machine learning algorithms by using training patterns, pattern association, the Hebbian learning rule, and competitive learning. First, learners examine the concepts and characteristics of online and offline training techniques in implementing artificial neural networks, and different training patterns in teaching inputs that are used in implementing artificial neural networks. You will learn to manage training samples, and how to use Google Colab to implement overfitting and underfitting scenarios by using baseline models. You will examine regularization techniques to use in training artificial neural networks. This course then demonstrates how to train previously-built neural network models using Python, and the prominent training algorithms to implement pattern associations. Next, learn the architecture and algorithm associated with learning vector quantization; the essential phases involved in implementing Hebbian learning; how to implement Hebbian learning rule using Python; and the steps involved in implementing competitive learning. Finally, you will examine prominent techniques to use to optimize neural networks, and how to debug neural networks.
Explore essential machine learning components used to learn, train, and build neural networks and prominent clustering and classification algorithms in this 12-video course. The use of hyperparameters and perceptrons in artificial neuron networks (ANNs) is also covered. Learners begin by studying essential ANN components required to process data, and also different paradigms of learning used in ANN. Examine essential clustering techniques that can be applied on ANN, and the roles of the essential components that are used in building neural networks. Next, recall the approach of generating deep neural networks from perceptrons; learn how to classify differences between models and hyperparameters and specify the approach of tuning hyperparameters. You will discover types of classification algorithm that can be used in neural networks, and features of essential deep learning frameworks for building neural networks. Explore how to choose the right neural network framework for neural network implementations from the perspective of usage scenarios and fitment model, and define computational models that can be used to build neural network models. The concluding exercise concerns ANN training and classification.
This 13-video course helps learners discover how to implement various neural networks scenarios by using Python, Keras, and TensorFlow for machine learning. Learn how to optimize, tune, and speed up the processes of artificial neural networks (ANN) and how to implement predictions with ANN is also covered. You will begin with a look at prominent building blocks involved in building a neural network, then recalling the concept and characteristics of evolutionary algorithms, gradient descent, and genetic algorithms. Learn how to build neural networks with Python and Keras for classification with Tensorflow as the backend. Discover how to build neural networks by using PyTorch; implement object image classification using neural network algorithms; and define and illustrate the use of learning rates to optimize deep learning. Examine various parameters and approaches of optimizing neural network speed; learn how to select hyperparameters and tune for dense networks by using Hyperas; and build linear models with estimators by using the capabilities of TensorFlow. Explore predicting with neural networks, temporal prediction optimization, and heterogenous prediction optimization. The concluding exercise involves building neural networks.
In this 13-video course, you will explore the capabilities and features of convolutional networks for machine learning that make it a recommended choice for visual recognition implementation. Begin by examining the architecture and the various layers of convolutional networks, including pooling layer, convo layer, normalization layer, and fully connected layer, and defining the concept and types of filters in convolutional networks along with their usage scenarios. Learn about the approach to maximizing filter activation with Keras; define the concept of feature map in convolutional networks and illustrate the approach of visualizing feature maps; and plot the map of the first convo layer for given images, then visualize the feature map output from every block in the visual geometry group (VGG) model. Look at optimization parameters for convolutional networks, and hyperparameters for tuning and optimizing convolutional networks. Learn about applying functions on pooling layer; pooling layer operations; implementing pooling layer with Python, and implementing convo layer with Python. The concluding exercise involves plotting feature maps.
Learners can explore the machine learning concept and classification of activation functions, the limitations of Tanh and the limitations of Sigmoid, and how these limitations can be resolved using the rectified linear unit, or ReLU, along with the significant benefits afforded by ReLU, in this 10-video course. You will observe how to implement ReLU activation function in convolutional networks using Python. Next, discover the core tasks used in implementing computer vision, and developing CNN models from scratch for object image classification by using Python and Keras. Examine the concept of the fully-connected layer and its role in convolutional networks, and also the CNN training process workflow and essential elements that you need to specify during the CNN training process. The final tutorial in this course involves listing and comparing the various convolutional neural network architectures. In the concluding exercise you will recall the benefits of applying ReLU in CNNs, list the prominent CNN architectures, and implement ReLU function in convolutional networks using Python.
Explore convolutional neural networks, their different types, and prominent use cases for machine learning, in this 10-video course. Learners will study the different layers and parameters of convolutional neural networks and their roles in implementing and addressing image recognition and classification problems. Key concepts covered in this course include the working mechanisms of convolutional neural networks, and the different types of convolutional neural networks that we can implement; and problems associated with computer vision, along with the prominent techniques to manage them. Next, you will learn about the role of neural networks and convolutional neural networks in implementing and addressing image recognition and classification problems; observe the prominent layers and parameters of convolutional neural networks for image classification; and learn to see the convolutional layer from a mathematical perspective, while recognizing the mathematical elements that enter into the convolution operations. Finally, learners will be shown how to build a convolutional neural network for image classification by using Python.
In this 12-video course, learners can explore machine learning problems that can be addressed with hyperparameters, and prominent hyperparameter tuning methods, along with problems associated with hyperparameter optimization. Key concepts covered here include the iterative workflow for machine learning problems, with a focus on essential measures and evaluation protocols; steps to improve performance of neural networks, along with impacts of data set sizes on neural network models and performance estimates; and impact of the size of training data sets on quality of mapping function and estimated performance of a fit neural network model. Next, you will learn the approaches of identifying overfitting scenarios and preventing overfitting by using regularization techniques; learn the impact of bias and variances on machine learning algorithms, and recall the approaches of fixing high bias and high variance in data sets; and see how to trade off bias variance by building and deriving an ideal learning curve by using Python. Finally, learners will observe how to test multiple models and select the right model by using Scikit-learn.
Learners can explore the concept of loss function, the different types of Loss function and their impact on neural networks, and the causes of optimization problems, in this 10-video course. Examine alternatives to optimization, the prominent optimizer algorithms and their associated properties, and the concept of learning rates in neural networks for machine learning solutions. Key concepts in this course include learning loss function and listing various types of loss function; recognizing impacts of the different types of loss function on neural networks models; and learning how to calculate loss function and score by using Python. Next, learners will learn to recognize critical causes of optimization problems and essential alternatives to optimization; recall prominent optimizer algorithms, along with their properties that can be applied for optimization; and how to perform comparative optimizer analysis using Keras. Finally, discover the relevance of learning rates in optimization and various approaches of improving learning rates; and learn the approach of finding learning rate by using RMSProp optimizer.
Explore how to create and optimize machine learning neural network models, scaling data, batch normalization, and internal covariate shift. Learners will discover the learning rate adaptation schedule, batch normalization, and using L1 and L2 regularization to manage overfitting problems. Key concepts covered in this 10-video course include the approach of creating deep learning network models, along with steps involved in optimizing networks, including deciding size and budget; how to implement the learning rate adaptation schedule in Keras by using SGD and specifying learning rate, epoch, and decay using Google Colab; and scaling data and the prominent data scaling methods, including data normalization and data standardization. Next, you will learn the concept of batch normalization and internal covariate shift; how to implement batch normalization using Python and TensorFlow; and the steps to implement L1 and L2 regularization to manage overfitting problems. Finally, observe how to implement gradient descent by using Python and the steps related to library import and data creation.
The versatility of BigML allows you to build supervised learning models without much complexity. In this course, you'll practice constructing a selection of supervised learning models using BigML. You'll start by building an ensemble of decision trees to perform binary classification. Next, you'll build a linear regression model to predict the values of homes in a particular region. You'll then train and evaluate a logistic regression model to illustrate how it can be used to solve similar problems to those solved using ensemble methods. Another BigML capability you'll explore is building a time series plot to make various forecasts. In each demonstration, you'll delve into some optional configurations for the model being trained. Lastly, you'll use the OptiML feature to find the optimal model for your data.
BigML includes various unsupervised learning models used to gain insights into your data. These insights can help make pivotal business decisions or act as a starting point to build supervised learning models. In this course, you'll build several unsupervised learning models and analyze the results they produce. You'll start by creating clusters from a dataset and examining how data points within a cluster share similarities. You'll move on to uncover associations in a dataset about items purchased on an e-commerce platform. Next, you'll apply topic modeling to extract the topics discussed in a collection of texts. Following this, you'll transform a dataset containing multiple fields into a handful of principal components using Principal Component Analysis, or PCA. Finally, you'll explore the detection of anomalies in your dataset.
This 11-video course explores the machine learning concepts of Bayesian methods and the implementation of Bayes' theorem and methods in machine learning. Learners can examine Bayesian statistics and analysis with a focus on probability distribution and prior knowledge distribution. Begin with a look at the concept of Bayesian probability and statistical inference, then move on to the concept of Bayesian theorem and its implementation in machine learning. Next, learn about the role of probability and statistics in Bayesian analysis from the perspective of frequentist probability and subjective probability paradigms. You will examine standard probability, continuous distribution, and discrete distribution, and recall the essential elements of Bayesian statistics including prior distribution, likelihood function, and posterior inference. Recognize the implementation of prominent Bayesian methods including inference, statistical modeling, influence of prior belief, and statistical graphics. Describe prior knowledge and compare the differences between non-informative prior distribution and informative prior distribution. The steps involved in Bayesian analysis, including modeling data, deciding prior distribution, likelihood construction, and posterior distribution are also covered. The concluding exercise focuses on Bayesian statistics and analysis.
Learners can examine the concept of Bayesian learning and the different types of Bayesian models in this 12-video course. Discover how to implement Bayesian models and computations by using different approaches and PyMC for your machine learning solutions. Learners start by exploring critical features of and difficulties associated with Bayesian learning methods, and then take a look at defining the Bayesian model and classifying single-parameter, multiparameter, and hierarchical Bayesian models. Examine the features of probabilistic programming and learn to list the popular probabilistic programming languages. You will look at defining Bayesian models with PyMC and arbitrary deterministic function and generating posterior samples with PyMC models. Next, learners recall the fundamental activities involved in the PyMC Bayesian data analysis process, including model checking, evaluation, comparison, and model expansion. Delve into the computation methods of Bayesian, including numerical integration, distributional approximation, and direct simulation. Also, look at computing with Markov chain simulation, and the prominent algorithms that can be used to find posterior modes based on the distribution approximation. The concluding exercise focuses on Bayesian modeling with PyMC.
This 11-video course explores advanced Bayesian computation models, as well as how to implement Bayesian modeling with linear regression, nonlinear, probabilistic, and mixture models. In addition, learners discover how to implement Bayesian inference models with PyMC3. First, learn how to build and implement Bayesian linear regression models by using Python for machine learning solutions. Examine prominent hierarchical linear models from the perspective of regression coefficients. Then view the concept of probability models and use of Bayesian methods for problems with missing data. You will discover how to build probability models by using Python, and examine coefficient shrinkage with nonlinear models, nonparametric models, and multivariate regression from nonlinear models. Examine fundamental concepts of Gaussian process models; the approaches of classification with mixture models and regression with mixture models; and essential properties of Dirichlet process models. Finally, learn how to implement Bayesian inference models in Python with PyMC3. The concluding exercise recalls hierarchical linear models from the perspective of regression coefficients, and asks learners to describe the approach of working with generalized linear models, and implement Bayesian inference by using PyMC3.
Keras is a deep learning package suitable for beginners. Although it is applied in multiple standard deep learning use cases, it is also used by commercial-grade products. To facilitate this, Keras provides additional, flexible options on top of the well-known Sequential API, which allow you to customize and create various neural networks. To utilize this, however, requires a more in-depth knowledge of the Keras framework. In this course, you'll develop the core skills needed to work with the Keras framework. You'll explore the advantages and disadvantages of using Keras over other frameworks, and examine how Keras can be used with TensorFlow. You'll move on to recognize how Keras is used for machine learning and deep learning. Finally, you'll implement two deep learning projects using the Keras framework.
Keras provides a quick way to implement, train, and evaluate robust neural networks in Python. Using Keras for AI development for prototyping AI is standard practice and AI practitioners need to know why and how to use Keras for particular AI implementations. In this course, you'll explore advanced techniques for working with the Keras framework. You'll recognize how Keras is different from other AI frameworks and identify cases in which it is advantageous to use Keras. You'll examine the functionality of the Keras Sequential model and Functional API and the role of multiple deep learning layers present in Keras. Finally, you will work with practical AI projects developed using Keras and troubleshoot common problems related to model training and evaluation.
Machine learning (ML) has become indispensable across all industries. With staggering amounts of data generated globally every second, it's impossible to make sense of it without using such advanced data analytics. The AWS Certified Machine Learning - Specialty certification is one of the most coveted yet challenging certs a data engineer or scientist can get. To pass the associated exam, candidates must demonstrate knowledge of various machine learning concepts and the ability to solve real-world business challenges. Use this course to prepare for acquiring this valuable certification. Get to grips with key data engineering and machine learning terminology, concepts, tools, tasks, and workflows. Then, dive into how the AWS Machine Learning platform is used for real-world applications. Upon completing this course, you'll recognize key ML concepts and how to prepare datasets, develop ML models, and optimize models for improved predictive accuracy.