This repository contains solutions to the exercises from the book Inside Deep Learning. The solutions are implemented with modern tools and frameworks to ensure clarity, reproducibility, and efficiency.
-
MLflow: Used for tracking experiments, managing model versions, and ensuring reproducibility. Logs key metrics, parameters, and results for each exercise, so you can visualize and analyze the results efficiently.
-
Optuna: Leveraged for hyperparameter optimization to find the best-performing models. Makes experimenting with different configurations both easy and effective.
- 1 The mechanics of learning
- 2 Fully connected networks
- 3 Convolutional neural networks
- 4 Recurrent neural networks
- 5 Modern training techniques
- 6 Common design building blocks
- 7 Autoencoding and self-supervision
- 8 Object detection
- 9 Generative adversarial networks
- 10 Attention mechanisms
- 11 Sequence-to-sequence
- 12 Network design alternatives to RNNs