framework for recurrect neural networks #254
Open
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
# Please enter a commit message to explain why this merge is necessary, # especially if it merges an updated upstream into a topic branch. # # Lines starting with '#' will be ignored, and an empty message aborts # the commit.
lewardo
commented
Jan 7, 2024
There was a problem hiding this comment.
great catch on the part of @tremblap,, the first prediction was being discarded and the prediction was skipping points !
depending on the input size there can still be issues with output prediction, but if the input length is appropriate given the training lengths then the output is consistent.
We tested a linear time series and it successfully predicted the correct points so there is definitely consistence convergence (I believe @tremblap still has the patch if anyone would like it)
…lly check the max length dynamically zero pad ceil(log10(maxlen)) zero padding would do but maybe we should store a new "maxlength" key
# Conflicts: # FlucomaClients.cmake # include/clients/nrt/DataSeriesClient.hpp # include/data/FluidJSON.hpp
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Outdated suggestions cannot be applied.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.
@tremblap @weefuzzy @AlexHarker
Now that the
dataseriesobject is functional, fun algorithms such as the various flavours of recurrent neural networks can be tinkered with!The recurrent structure is written in a way to facilitate writing new recurrent networks (e.g. vanilla RNNs, GRUs, adaptive computation LSTMs etc.), currently only an LSTM is being implemented using the framework, but using the
Recurwrapper (as inspired by the Flux.jl ML library in Julia) writing new algorithms is much easier - I will also write up a guide to writing new recurrent algos as part of this pr once it is fully presentable.The client is not yet implemented, however there algorithm does compile and regress onto time series successfully - this pr is mostly for now to have a place to discuss potential UI challenges with the client and any other back and forth as needed!