machine-learning-zoomcamp
machine-learning-zoomcamp copied to clipboard
DataTalksClub
Timecodes for "ML Zoomcamp 8.5 - Transfer Learning"
I'll talk about it on the course launch
On Mon, Sep 5, 2022, 15:42 abhinit21 @.***> wrote:
What is this Issue label timecodes ? Can I contribute @amitfrancis https://github.com/amitfrancis
— Reply to this email directly, view it on GitHub https://github.com/alexeygrigorev/mlbookcamp-code/issues/156#issuecomment-1237052055, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAGVV3R6DJUFULDPQXL53ETV4X2CRANCNFSM6AAAAAAQE5MO7E . You are receiving this because you authored the thread.Message ID: @.***>
0:00:00 - Lesson 5: Transfer Learning with Pre-trained Neural Network 0:01:50 - Transfer learning: reusing convolutional layers for training. 0:03:41 - Pre-process input, read from directory, specify batch size. 0:05:41 - Exploring image dataset classes and generator. 0:07:31 - Multi-class classification with one-hot encoding. 0:09:31 - Creating a base model for training. 0:11:45 - Extracting vector representation from non-trainable model. 0:13:58 - Creating and testing a Keras model. 0:15:47 - Drawing 3D cube, converting to vector. 0:17:45 - Converting 3D/2D to 1D using pooling. 0:19:51 - Building neural network with functional style. 0:21:45 - Pooling and dense layer for predictions. 0:23:27 - Optimizers in Keras: Adam recommended default algorithm. 0:25:17 - Optimizing neural network with loss evaluation. 0:26:58 - Using row scores for stable predictions. 0:28:45 - Training model with fit method and epochs. 0:30:35 - Training model and analyzing accuracy fluctuations. 0:32:49 - Training accuracy reaches almost 1, overfitting observed.
