Data Normalization. See ml_data for details
◆ ml_column_mode
The main mode of each entry value; determines how it is stored and how it is translated and what functionality it works with.
Definition at line 32 of file ml_data_column_modes.hpp.
◆ ml_missing_value_action
◆ column_mode_enum_to_name()
For error reporting, returns a name of the mode based on the column mode value.
◆ mode_has_fixed_size()
Returns true if the underlying type always results in constant size pattern, and false otherwise.
Definition at line 66 of file ml_data_column_modes.hpp.
◆ mode_is_categorical()
Returns true if the underlying type is treated as a categorical variable, and false otherwise.
Definition at line 47 of file ml_data_column_modes.hpp.
◆ mode_is_indexed()
Returns true if the underlying type is indexed, and false otherwise. This differs form the is_categorical in that dictionaries are not treated as pure categorical variables, as they have values associated with them, but they are indexed.
Definition at line 87 of file ml_data_column_modes.hpp.
◆ _column_indexer_first_level_lookup_size_n_bits
constexpr int turi::ml_data_internal::_column_indexer_first_level_lookup_size_n_bits = 8 |
|
static |
Use a two-level hash table to store the index mappings. The first level is constant size and unlocked, determined by an n-bit hash. Each leaf in this one contains a hash table and lock. This significantly reduces lock contention. This _column_metadata_first_level_lookup_size_n_bits gives the number of bits used for this first lookup.
Definition at line 37 of file column_indexer.hpp.