Even though the implementation of SmartMemory still needs a little work to avoid crashes in the case of misuse by the user,
I will consider it done. Still have to remove comments.
Added multi-threaded lzw compression.
A few compression methods allow the use of a BinarySet* to avoid an additional copy.
Note that even though neural networks now adjust bias along with weights and they both follow typical conventions,
the resulting network will likely not be correct for unknown reasons.
closes #52 closes #107 closes #98
Even though the implementation of SmartMemory still needs a little work to avoid crashes in the case of misuse by the user, I will consider it done. Still have to remove comments.
Added multi-threaded lzw compression. A few compression methods allow the use of a BinarySet* to avoid an additional copy.
Note that even though neural networks now adjust bias along with weights and they both follow typical conventions, the resulting network will likely not be correct for unknown reasons.