TensorFlow 2.0 Practical

Build your first artificial neural network by constructing a simple neuron with weights and a bias to map temperature in C to Fahrenheit, trained with TensorFlow 2.0.

Explore TensorFlow 2.0 practical project 1 by visualizing Celsius–Fahrenheit data with seaborn, creating training data, and building a simple dense neural network trained with Adam and mean squared error.

Develop a TensorFlow 2.0 regression model in Google Colab to map temperature to revenue, training and evaluating a neural network and comparing with linear regression.

Explore the bias-variance trade-off in machine learning by comparing linear regression and a high-order polynomial, emphasizing training versus testing performance and regularization in TensorFlow 2.0 practical.

Study regression metrics to evaluate model performance in TensorFlow 2.0, including MAE, MSE, RMSE, MAPE, and R-squared, plus the adjusted version. Understand residuals and predictions to assess fit.

Build and evaluate a TensorFlow 2.0 neural network on Google Colab using bike rental data features to train and test a regression model.

Visualize the dataset by formatting dates and plotting weekly, monthly, quarterly counts with seaborn pair plots and heatmap. Prepare numerical features for training and testing in TensorFlow 2.0 Practical.

Encode categorical features with one-hot encoding, merge with numerical data, drop unused columns, normalize the target, and split 80/20 for training and testing to train and evaluate the model.

Evaluate and visualize a TensorFlow 2.0 model using Keras, training and validation loss, apply cross validation, check for overfitting, and assess performance with RMSE, MSE, MAE, R-squared, and predictions.

Explore predicting house prices with an artificial neural network in TensorFlow 2.0 using Google Colab, visualization, and a two-stage regression approach that improves R squared from 0.578 to about 0.87.

Learn to predict house prices with TensorFlow 2.0 by loading and visualizing data, selecting features such as bedrooms, bathrooms, and square footage, and scaling inputs and outputs.

Learn to read a confusion matrix for classification, identify true/false positives and negatives, and apply accuracy, error rate, precision, and recall through a cancer-patient example.

Clean and engineer data by dropping date and rating, encoding variation with get_dummies, and concatenating results, then apply countvectorizer tokenization to reviews to create features (X) and output (Y).

Build and evaluate a diabetes prediction model in TensorFlow 2.0 on Google Colab, covering data import, visualization, feature scaling, train-test split, model training, and performance reporting.

Explore the diabetes dataset with TensorFlow 2.0, visualize with seaborn (count plot, pair plot, heatmap), and prepare data with cleaning, scaling, and a train-test split.

Apply convolutions with kernels to extract features and create feature maps, then use relu, pool, flatten, and feed into a fully connected network for TensorFlow 2.0.

Explore how relu adds non-linearity to feature maps by zeroing negatives and passing positives, boosting sparsity and training effectiveness in TensorFlow 2.0 cnn architectures and lenet.

Explore pooling and downsampling in CNNs, using 2x2 max pooling with stride 2 to reduce a 4x4 feature map, then flatten for a dense network that classifies digits.

Load the fashion mnist dataset from a csv, split into training and testing arrays, and visualize 28x28 grayscale samples with a 15 by 15 image grid using numpy and matplotlib.

Evaluate a CNN on test data to achieve about 91% accuracy, explore validation data and cross-validation to improve generalization, and visualize predictions, confusion matrix, and a detailed classification report.

Explore the Lenet architecture, with convolutions, max pooling, and downsampling, and learn to implement Lenet-5 in TensorFlow 2.0 using Keras.

Train a fashion-MNIST classifier with convolution layers in TensorFlow 2.0, save and deploy the model on a server via TensorFlow Serving, and perform inference on uploaded images.

Save and explore a trained model with metagraph and signature, then serve it via TensorFlow Serving on port 8501 and perform three-sample JSON inferences.

Learn to use tensor board to visualize training progress, track accuracy and loss, and view the network architecture graph, including layer details and weight distributions.