Deployment bug-fix #1036
Open
Deployment bug-fix #1036
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
…terface for simplenn. (Missing: DAG, unit tests.)
…ries). RMSProp is unpublished, this implementation is based on Karpathy's lecture notes: http://cs231n.github.io/neural-networks-3/
…roved documentation on learning rates.
…t tests; minor documentation tweaks
…crease of update magnitudes to 0
Conflicts: examples/cnn_train.m examples/cnn_train_dag.m
… instead of the MATLAB default ones
Loss improvements Approved-by: Joao Henriques <[email protected]>
…r vl_nn* functions) An empty derivative input was required in forward mode when passing additional parameters (e.g. vl_nnrelu(x, [], 'leak', 0.1), while in other functions the derivative is omitted.
At the end, the original purpose of having a network all the time in a valid state does almost never works as the network fails on rebuild. The rethrow makes it harder to debug issues.
Loss.m: number of samples for spatial maps now handled correctly (instead of crashing). vl_nnpdist.m: Handles correctly aggregate for 4D tensors.
With aggregate -> single scalar, numerically the algorithm is even worse. Decreased the size of the input tensor.
Minimal training/custom IMDB example with toy data, and tensor visualization function * added cnn_toy_data example, and vl_tshow to visualize tensors * fix vl_tshow for ndims < 4 * decouple toy image generation from IMDB creation, more realistic use-case * vl_tshow: allow setting parent axes and other properties * vl_tshow: replace hacky montage+padding with imagesc+grid, better scaling * moved files to examples/custom_imdb/
Improvements to vl_argparse: 'field.subfield' syntax, ignore missing defaults (all optional) Approved-by: Samuel <[email protected]> Approved-by: Andrea Vedaldi <[email protected]>
fixes dropout explanation updates vl_nndropout doc string to define rate as the probability that a variable is dropped. Approved-by: Andrea Vedaldi <[email protected]>
Originally used only in debug mode. The bilinearsampler_gpu needs cleanup.
Makefile: upgrades default CUDA and MATLAB
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Previous layer's biases should be multiplied by (gamma / sigma) as well