/fastai_v1_2018

Notebooks used during the Practical Deep Learning for Coders course at USF

Primary LanguageJupyter Notebook

Fast.AI - Deep Learning part 1

L1: Image classification

  • Transfer Learning
  • Learning Rate Finder
  • Labelling

L2: Data cleaning and production; SGD from scratch

  • Build image classification model
  • from image collection to productionizing

L3: Data blocks; Multi-label classification; Segmentation

  • Datablock API
  • Multi-label classification
  • Segmentation

L4: NLP; Tabular data; Collaborative filtering; Embeddings

  • NLP, sentiment analysis using ULMFiT
  • Tabular Data
  • Collaborative Filtering
  • Embedding Layer

L5: Back propagation; Accelerated SGD; Neural net from scratch

  • Backpropagation -- train a neural net from scratch

L6: Regularization; Convolutions; Data ethics

  • Dropout
  • Data Augmentation
  • Batch Normalization
  • Class activated map in convolutions.

L7: ResNets, UNets, GANs, RNNs

  • skip connection
  • ResNet
  • Super-Resolution (U-Nets), perceptual loss
  • GANs