Artificial Neural Networks (ANN) with Keras in Python and R

Install Python and Anaconda distribution to set up a data science environment; download the Anaconda distribution, install Python 3.7, choose your operating system, run the installer, and confirm internet connectivity.

Celebrate reaching this milestone as you join the top 50% of learners and explore playback, subtitles, offline download, Q&A, AI assistant, rating tips, and your certificate.

Open and launch the Jupyter notebook using Anaconda Navigator, Anaconda Prompt, or Command Prompt, then locate and open ipynb notebooks in your default directory.

Explore how to create a new Jupyter notebook with Python 3, run code in cells, and use markdown and raw formats to write readable comments, headings, and bullet lists.

Learn to use Jupyter cells: switch between editable and non-editable modes, recognize blue and green borders, switch between code and markdown with shortcuts (m, vi), and insert or delete cells.

Learn Python basics by performing arithmetic operations: addition, subtraction, multiplication, division, exponent, and modulus. The lesson covers operator precedence with parentheses, variable assignment, augmented (automatic) operators, and comparison operators.

Explore strings in Python basics, learning how to pass strings with quotes, assign variables, print results, and format output using format, plus length, replace, indexing, and slicing.

Explore lists in Python, where each element can be any type and lists are defined with square brackets. Practice indexing, printing, slicing (start, stop, step), and nesting lists within lists.

Learn to create and manipulate Python lists using range, start, stop, and step; sort with reverse, append, insert, and slice; and remove or delete elements with help.

Explore tuples and dictionaries in Python, noting that tuples are immutable and defined with parentheses, while lists use a square record and dictionaries use curly braces for key-value pairs.

Cover numpy basics for numerical computing: create arrays and matrices from lists, explore shapes and dtypes, perform indexing and slicing, and generate random numbers with arange, linspace, zeros, ones, diag.

Learn to manipulate data with the pandas library in Python, import CSV data into data frames, and navigate with index, iloc, and loc for precise data selection.

Explore the seaborn Python library by loading the iris dataset, inspecting its five features across 150 rows, describing statistics, and visualizing with scatter plots, distributions, and a pairplot.

Install and set up R and R Studio on your system, including downloading from the R project site, choosing the right installer, and using R Studio for streamlined scripting.

Master basics of r and r studio: run commands with control enter, use comments, assign variables with <-, and build vectors with c or 1:10; list and clear the workspace.

Learn how to install, load, and manage packages in R, including inbuilt and external packages, and visualize data with the DG plot package.

learn to use the R data set package to access inbuilt data, inspect the iris dataset with five variables and 150 observations, and load it into your workspace for analysis.

Learn manual data entry in ANN workflows: assign values to variables, use the concatenation operator, generate a five to fifty sequence with the sequence function, and input numbers with scan.

Import data from tab-delimited text and comma-separated csv files into the workspace, specify file paths, headers, and delimiters, and inspect resulting data frames.

Learn to compute region frequency distributions and create barplots in R, customizing color, orientation, borders, and titles, then export the chart as PNG.

Learn to create histograms in R to visualize age distributions using breaks for bin control and frequency. Export the chart as an image for presentations.

View perceptrons as binary input neurons with weights and a threshold, producing 0 or 1 output after summing products; illustrated by choosing shirts based on color, sleeve, and fabric.

Extend the perception from binary to real-valued inputs with bias by replacing the step activation with a sigmoid (logistic) function, yielding outputs between 0 and 1.

Learn to build a simple perceptron classifier in Python with scikit-learn using iris data, selecting two features, training, predicting, and evaluating accuracy, then inspect coefficients and intercept.

Explore basic terminologies of artificial neural networks, including neurons, parallel and sequential stacking, feed-forward and fully connected architectures, single and multiple outputs, and the transition to deep learning.

Explore how neural networks learn through gradient descent, using sigmoid neurons, forward and backward propagation, and iterative weight and bias updates to minimize error.

Explore how gradient descent minimizes cross-entropy loss through forward and backward propagation, updating weights and biases with a learning rate to train accurate classifiers.

Explore activation functions like sigmoid, tanh, and identified linear unit, and explain why they bound outputs and enable nonlinear patterns; cover softmax for multiclass and gradient descent variants, including epochs.

Explore how to set classification and regression hyperparameters in neural networks using Keras, including input neurons, hidden layers, activation functions, and loss functions like cross entropy and mean squared error.

Explore Keras as a high-level deep learning framework to define and train models, while leveraging backends like TensorFlow and other libraries; learn to install Keras with its TensorFlow backend.

Install and activate the keras and tensorflow backends in R, using the CPU-based setup, then prepare to load and split the dataset into training and testing data.

Load the fashion eminence database from keras, train a model on 60k training and 10k test images, each 28 by 28 grayscale, to classify into 10 fashion categories.

Normalize image pixel intensities by dividing by 255 to scale 0-255 to 0-1, using floating outputs. Split data into validation, training, and test sets to assess model performance.

Import the fashion dataset, already split into 60,000 training and 10,000 test images; normalize 28 by 28 grayscale pixel values by 255 for model training.

Learn to build ann models with Keras in Python and R using the sequential API for simple dense networks and the functional API for complex architectures.

Build a sequential keras network with 784-inputs (28×28), two hidden layers (300 and 100) with relu, a softmax 10-class output, seed 242 for reproducibility, random weights, and zero biases.

Compile and train the neural network using sparse categorical cross-entropy loss, an optimizer, and accuracy metrics, monitor training and validation loss and accuracy, and consider class weights for imbalance.

Evaluate model performance on a test set with Keras using the evaluate method to obtain loss and accuracy, then predict class probabilities and labels for new unseen data.