The hashed column feature generator function is a vectorizer that can be used on large or variable length rows, or fixed lenght rows with lots of categorical values, to produce a smaller fixed length numerical feature vector.
A new function of type feature_hasher named <id> can be created as follows:
mldb.put("/v1/functions/"+<id>, {
"type": "feature_hasher",
"params": {
"numBits": <int>,
"mode": <HashingMode>
}
})with the following key-value definitions for params:
| Field, Type, Default | Description |
|---|---|
numBits | Number of bits to use for the hash. The number of resulting buckets will be \(2^{\text{numBits}}\). |
mode | Hashing mode to use. Controls what gets hashed. |
The mode field controls what gets hashed:
columns: This mode should be used if you're hashing a generally sparse
bag of words, where words are in the columns and the value represents
a count. Only the column names will be hashed.columnsAndValues: This mode should be used if you're hashing dense data,
where fixed columns can take different values. The contatenation of both the
column name and the cell value will be hashed.Functions of this type have a single input value called columns which is a row and
a single output value called hash which is a row of size \(2^{\text{numBits}}\).