Differences

This shows you the differences between two versions of the page.

--- data_mining:neural_network:transfer_learning [2018/05/06 19:35] – phreazer
+++ data_mining:neural_network:transfer_learning [2018/05/25 21:13] (current) – [Transfer learning] phreazer
@@ Line 2: / Line 2: @@
 Using pre-trained models / their trained weights as a starting point to train a model for a different data set.
+Use pre-trained net, initialize last layers with random weights and e.g. create own softmax layer, **freeze** parameters in previous layers.
+Options: Train new layers of network, or even more layers.
+Prereqs:
+  * Task A and B have same input x
+  * Much data for existing model/task A, for new model only few data
+  * Low level features from A could be helpful for task B
+  * Pre-trained model needs to generalize
+Another trick:
+  * Precompute output of frozen layers for all samples (save computation time later)
+For larger set of samples:
+  * Only freeze first layers, train last few layers (and replace softmax)
+For large set of samples:
+  * Use weights as initialization, then train (and replace softmax)
+===== Image-based NNs =====
+Pre-trained models which generalize can be used as a starting model, e.g. https://github.com/KaimingHe/deep-residual-networks ResNet models.
+For CNNs more generic features are usually contained in the **earlier** layers.