V 0.7.0
Deep Learning
- Refactored computation graph to work on operation level
- Created methods for arithmetic and basic matrix operations
- Defined general methods for operations backward and forward pass
- Created general definition of a
value in the computation graph
- Flushed out logic for how loss functions will be added to computation graph
- Created general union abstraction around layer types for neural nets
- Implemented basic multi layer perceptron with newly refactored computation graph
- Added layer implementations for activation and linear concepts
- Refactored neural network library
- Added network methods for training and adding layers
Graph Library
- Added integration tests for GMUL file interpretation
- Added row limit for data frame structure
- Integration tests for how many rows a data frame can load
Pipeline
- Refactored pipeline workflow
- Separated build and test steps for core package
- Provided binaries in commit history