Examining-the-effect-of-hyperparameters-and-Exploring-the-relationship-between-hyperparameters.

Examining the effect of hyperparameters and Exploring the relationship between hyperparameters through experimentation: Building, Training, and Tuning an Image Classification Model with TensorFlow and PyTorch – A CIFAR-10 dataset Use case.

1. This notebook contains an introduction to the CIFAR-10 dataset for image classification, and shows how to build, train, and tune a model for image classification with TensorFlow from scratch with an emphasis on understanding the effect of different hyperparameters and their relationships through experimentation.
2. It introduces training with GPU and checks to confirm that TensorFlow is connected to your device’s GPU.
3. It introduces different ways of loading data in TensorFlow and building input data pipelines: Generate TensorFlow datasets, shuffle, batch, and prefetch for efficient data loading during training.
4. Introduces the Keras image pre-processing layers.
5. Building and choosing a baseline model to be improved upon.
6. Building convolutional neural network models with different layers like the dropout layers, batch normalization layers, maxpooling layers, flattening layers, activation layers, convolution layers, choice of activation functions, loss functions, and optimizers.
7. Discusses overfitting and underfitting; the signs of overfitting and underfitting using learning plots and how to solve them.
8. Examining the effect of hyperparameters and exploring the relationship between hyperparameters through experimentation Hyperparameter choice and introduction of Keras tuner for hyperparameter tuning.
9. A pytorch implementation of the final tuned model is also developed.
10. Finally, the TensorFlow resnet model is implemented to achieve a validation accuracy of over 90% on the CIFAR-10 dataset.