V 0.5.0
Data Structures
- Implement a hash map library
- Create POC, avoid collisions at all costs
- Throw exceptions for key lookup that don't exist
- Improvement and fix for the adjacency list data structure
- Detect unused slots in the graph
- Don't allocate more nodes that aren't needed for both unweighted, weighted and directed/un-directed graphs
- Use array of pointers for node_list structure, use arrow deference for head pointer
- Matrix library
- Add method for getting columns of matrix
to_cols
- Allow option for batching a matrix and selecting a certain amount or rows
- Add hash map as attribute for data frame
- Replicate similar methods for getting cols and rows
- Create a prototype of GML (graph markup language)
- Create a regular expression for parsing this type of file format
- Create an interpreter that can serialize files
- Will not do weighted graphs in this version
Deep Learning
- Refactored computation graph to forward and backward all node types
- Added refactored computation graph representation to network module
Graph Library
- Make all random walk methods use the walk structure
- Easier to use on the python front end
- Create general graph structure that makes use of data structures
- Include the adjacency list and matrices as inherited attributes
- Add attribute for keeping track of labeled nodes
- Label propagation with multiple classes
- Should work with more than two labels
- Discovered that graph markup serialization doesn't work here