This page describes a list of Hivemall functions. See also a list of generic Hivemall functions for more general-purpose functions such as array and map UDFs.

Regression

  • train_arow_regr(array<int|bigint|string> features, float target [, constant string options]) - a standard AROW (Adaptive Reguralization of Weight Vectors) regressor that uses y - w^Tx for the loss function.
    SELECT 
  feature,
  argmin_kld(weight, covar) as weight
FROM (
  SELECT 
     train_arow_regr(features,label) as (feature,weight,covar)
  FROM 
     training_data
 ) t 
GROUP BY feature
    Reference: K. Crammer, A. Kulesza, and M. Dredze, "Adaptive Regularization of Weight Vectors", In Proc. NIPS, 2009.

  • train_arowe2_regr(array<int|bigint|string> features, float target [, constant string options]) - a refined version of AROW (Adaptive Reguralization of Weight Vectors) regressor that usages adaptive epsilon-insensitive hinge loss |w^t - y| - epsilon * stddev for the loss function

    SELECT 
  feature,
  argmin_kld(weight, covar) as weight
FROM (
  SELECT 
     train_arowe2_regr(features,label) as (feature,weight,covar)
  FROM 
     training_data
 ) t 
GROUP BY feature

  • train_arowe_regr(array<int|bigint|string> features, float target [, constant string options]) - a refined version of AROW (Adaptive Reguralization of Weight Vectors) regressor that usages epsilon-insensitive hinge loss |w^t - y| - epsilon for the loss function

    SELECT 
  feature,
  argmin_kld(weight, covar) as weight
FROM (
  SELECT 
     train_arowe_regr(features,label) as (feature,weight,covar)
  FROM 
     training_data
 ) t 
GROUP BY feature

  • train_pa1_regr(array<int|bigint|string> features, float target [, constant string options]) - PA-1 regressor that returns a relation consists of (int|bigint|string) feature, float weight.

    SELECT 
 feature,
 avg(weight) as weight
FROM 
 (SELECT 
     train_pa1_regr(features,label) as (feature,weight)
  FROM 
     training_data
 ) t 
GROUP BY feature

    Reference: Koby Crammer et.al., Online Passive-Aggressive Algorithms. Journal of Machine Learning Research, 2006.

  • train_pa1a_regr(array<int|bigint|string> features, float target [, constant string options]) - Returns a relation consists of (int|bigint|string) feature, float weight.

  • train_pa2_regr(array<int|bigint|string> features, float target [, constant string options]) - Returns a relation consists of (int|bigint|string) feature, float weight.

  • train_pa2a_regr(array<int|bigint|string> features, float target [, constant string options]) - Returns a relation consists of (int|bigint|string) feature, float weight.

  • train_regressor(list<string|int|bigint> features, double label [, const string options]) - Returns a relation consists of <string|int|bigint feature, float weight>

    Build a prediction model by a generic regressor

Classification

Binary classification

  • kpa_predict(@Nonnull double xh, @Nonnull double xk, @Nullable float w0, @Nonnull float w1, @Nonnull float w2, @Nullable float w3) - Returns a prediction value in Double

  • train_arow(list<string|int|bigint> features, int label [, const string options]) - Returns a relation consists of <string|int|bigint feature, float weight, float covar>

    Build a prediction model by Adaptive Regularization of Weight Vectors (AROW) binary classifier

    Reference: K. Crammer, A. Kulesza, and M. Dredze, "Adaptive Regularization of Weight Vectors", In Proc. NIPS, 2009.

  • train_arowh(list<string|int|bigint> features, int label [, const string options]) - Returns a relation consists of <string|int|bigint feature, float weight, float covar>

    Build a prediction model by AROW binary classifier using hinge loss

  • train_classifier(list<string|int|bigint> features, int label [, const string options]) - Returns a relation consists of <string|int|bigint feature, float weight>

    Build a prediction model by a generic classifier

  • train_cw(list<string|int|bigint> features, int label [, const string options]) - Returns a relation consists of <string|int|bigint feature, float weight, float covar>

    Build a prediction model by Confidence-Weighted (CW) binary classifier

  • train_kpa(array<string|int|bigint> features, int label [, const string options]) - returns a relation <h int, hk int, float w0, float w1, float w2, float w3>

  • train_pa(list<string|int|bigint> features, int label [, const string options]) - Returns a relation consists of <string|int|bigint feature, float weight>

    Build a prediction model by Passive-Aggressive (PA) binary classifier

  • train_pa1(list<string|int|bigint> features, int label [, const string options]) - Returns a relation consists of <string|int|bigint feature, float weight>

    Build a prediction model by Passive-Aggressive 1 (PA-1) binary classifier

  • train_pa2(list<string|int|bigint> features, int label [, const string options]) - Returns a relation consists of <string|int|bigint feature, float weight>

    Build a prediction model by Passive-Aggressive 2 (PA-2) binary classifier

  • train_perceptron(list<string|int|bigint> features, int label [, const string options]) - Returns a relation consists of <string|int|bigint feature, float weight>

    Build a prediction model by Perceptron binary classifier

  • train_scw(list<string|int|bigint> features, int label [, const string options]) - Returns a relation consists of <string|int|bigint feature, float weight, float covar>

    Build a prediction model by Soft Confidence-Weighted (SCW-1) binary classifier

  • train_scw2(list<string|int|bigint> features, int label [, const string options]) - Returns a relation consists of <string|int|bigint feature, float weight, float covar>

    Build a prediction model by Soft Confidence-Weighted 2 (SCW-2) binary classifier

Multiclass classification

  • train_multiclass_arow(list<string|int|bigint> features, {int|string} label [, const string options]) - Returns a relation consists of <{int|string} label, {string|int|bigint} feature, float weight, float covar>

    Build a prediction model by Adaptive Regularization of Weight Vectors (AROW) multiclass classifier

  • train_multiclass_arowh(list<string|int|bigint> features, int|string label [, const string options]) - Returns a relation consists of <int|string label, string|int|bigint feature, float weight, float covar>

    Build a prediction model by Adaptive Regularization of Weight Vectors (AROW) multiclass classifier using hinge loss

  • train_multiclass_cw(list<string|int|bigint> features, {int|string} label [, const string options]) - Returns a relation consists of <{int|string} label, {string|int|bigint} feature, float weight, float covar>

    Build a prediction model by Confidence-Weighted (CW) multiclass classifier

  • train_multiclass_pa(list<string|int|bigint> features, {int|string} label [, const string options]) - Returns a relation consists of <{int|string} label, {string|int|bigint} feature, float weight>

    Build a prediction model by Passive-Aggressive (PA) multiclass classifier

  • train_multiclass_pa1(list<string|int|bigint> features, {int|string} label [, const string options]) - Returns a relation consists of <{int|string} label, {string|int|bigint} feature, float weight>

    Build a prediction model by Passive-Aggressive 1 (PA-1) multiclass classifier

  • train_multiclass_pa2(list<string|int|bigint> features, {int|string} label [, const string options]) - Returns a relation consists of <{int|string} label, {string|int|bigint} feature, float weight>

    Build a prediction model by Passive-Aggressive 2 (PA-2) multiclass classifier

  • train_multiclass_perceptron(list<string|int|bigint> features, {int|string} label [, const string options]) - Returns a relation consists of <{int|string} label, {string|int|bigint} feature, float weight>

    Build a prediction model by Perceptron multiclass classifier

  • train_multiclass_scw(list<string|int|bigint> features, {int|string} label [, const string options]) - Returns a relation consists of <{int|string} label, {string|int|bigint} feature, float weight, float covar>

    Build a prediction model by Soft Confidence-Weighted (SCW-1) multiclass classifier

  • train_multiclass_scw2(list<string|int|bigint> features, {int|string} label [, const string options]) - Returns a relation consists of <{int|string} label, {string|int|bigint} feature, float weight, float covar>

    Build a prediction model by Soft Confidence-Weighted 2 (SCW-2) multiclass classifier

Matrix factorization

  • bprmf_predict(List<Float> Pu, List<Float> Qi[, double Bi]) - Returns the prediction value

  • mf_predict(array<double> Pu, array<double> Qi[, double Bu, double Bi[, double mu]]) - Returns the prediction value

  • train_bprmf(INT user, INT posItem, INT negItem [, String options]) - Returns a relation <INT i, FLOAT Pi, FLOAT Qi [, FLOAT Bi]>

  • train_mf_adagrad(INT user, INT item, FLOAT rating [, CONSTANT STRING options]) - Returns a relation consists of <int idx, array<float> Pu, array<float> Qi [, float Bu, float Bi [, float mu]]>

  • train_mf_sgd(INT user, INT item, FLOAT rating [, CONSTANT STRING options]) - Returns a relation consists of <int idx, array<float> Pu, array<float> Qi [, float Bu, float Bi [, float mu]]>

Factorization machines

  • ffm_predict(float Wi, array<float> Vifj, array<float> Vjfi, float Xi, float Xj) - Returns a prediction value in Double

  • fm_predict(Float Wj, array<float> Vjf, float Xj) - Returns a prediction value in Double

  • train_ffm(array<string> x, double y [, const string options]) - Returns a prediction model

  • train_fm(array<string> x, double y [, const string options]) - Returns a prediction model

Recommendation

  • train_slim( int i, map<int, double> r_i, map<int, map<int, double>> topKRatesOfI, int j, map<int, double> r_j [, constant string options]) - Returns row index, column index and non-zero weight value of prediction model

Anomaly detection

  • changefinder(double|array<double> x [, const string options]) - Returns outlier/change-point scores and decisions using ChangeFinder. It will return a tuple <double outlier_score, double changepoint_score [, boolean is_anomaly [, boolean is_changepoint]]

  • sst(double|array<double> x [, const string options]) - Returns change-point scores and decisions using Singular Spectrum Transformation (SST). It will return a tuple <double changepoint_score [, boolean is_changepoint]>

Topic modeling

  • lda_predict(string word, float value, int label, float lambda[, const string options]) - Returns a list which consists of <int label, float prob>

  • plsa_predict(string word, float value, int label, float prob[, const string options]) - Returns a list which consists of <int label, float prob>

  • train_lda(array<string> words[, const string options]) - Returns a relation consists of <int topic, string word, float score>

  • train_plsa(array<string> words[, const string options]) - Returns a relation consists of <int topic, string word, float score>

Preprocessing

  • add_bias(feature_vector in array<string>) - Returns features with a bias in array<string>

  • add_feature_index(ARRAY[DOUBLE]: dense feature vector) - Returns a feature vector with feature indices

  • extract_feature(feature_vector in array<string>) - Returns features in array<string>

  • extract_weight(feature_vector in array<string>) - Returns the weights of features in array<string>

  • feature(<string|int|long|short|byte> feature, <number> value) - Returns a feature string

  • feature_index(feature_vector in array<string>) - Returns feature indices in array<index>

  • sort_by_feature(map in map<int,float>) - Returns a sorted map

Data amplification

  • amplify(const int xtimes, *) - amplify the input records x-times

  • rand_amplify(const int xtimes [, const string options], *) - amplify the input records x-times in map-side

Feature binning

  • build_bins(number weight, const int num_of_bins[, const boolean auto_shrink = false]) - Return quantiles representing bins: array<double>

  • feature_binning(array<features::string> features, map<string, array<number>> quantiles_map) - returns a binned feature vector as an array<features::string> FUNC(number weight, array<number> quantiles) - returns bin ID as int

    WITH extracted as (
  select 
    extract_feature(feature) as index,
    extract_weight(feature) as value
  from
    input l
    LATERAL VIEW explode(features) r as feature
),
mapping as (
  select
    index, 
    build_bins(value, 5, true) as quantiles -- 5 bins with auto bin shrinking
  from
    extracted
  group by
    index
),
bins as (
   select 
    to_map(index, quantiles) as quantiles 
   from
    mapping
)
select
  l.features as original,
  feature_binning(l.features, r.quantiles) as features
from
  input l
  cross join bins r

> ["name#Jacob","gender#Male","age:20.0"] ["name#Jacob","gender#Male","age:2"]
> ["name#Isabella","gender#Female","age:20.0"]    ["name#Isabella","gender#Female","age:2"]

Feature format conversion

  • conv2dense(int feature, float weight, int nDims) - Return a dense model in array<float>

  • quantify(boolean output, col1, col2, ...) - Returns an identified features

  • to_dense_features(array<string> feature_vector, int dimensions) - Returns a dense feature in array<float>

  • to_libsvm_format(array<string> feautres [, double/integer target, const string options]) - Returns a string representation of libsvm

    Usage:
 select to_libsvm_format(array('apple:3.4','orange:2.1'))
 > 6284535:3.4 8104713:2.1
 select to_libsvm_format(array('apple:3.4','orange:2.1'), '-features 10')
 > 3:2.1 7:3.4
 select to_libsvm_format(array('7:3.4','3:2.1'), 5.0)
 > 5.0 3:2.1 7:3.4

  • to_sparse_features(array<float> feature_vector) - Returns a sparse feature in array<string>

Feature hashing

  • array_hash_values(array<string> values, [string prefix [, int numFeatures], boolean useIndexAsPrefix]) returns hash values in array<int>

  • feature_hashing(array<string> features [, const string options]) - returns a hashed feature vector in array<string>

    select feature_hashing(array('aaa:1.0','aaa','bbb:2.0'), '-libsvm');
 ["4063537:1.0","4063537:1","8459207:2.0"]

select feature_hashing(array('aaa:1.0','aaa','bbb:2.0'), '-features 10');
 ["7:1.0","7","1:2.0"]

select feature_hashing(array('aaa:1.0','aaa','bbb:2.0'), '-features 10 -libsvm');
 ["1:2.0","7:1.0","7:1"]

  • mhash(string word) returns a murmurhash3 INT value starting from 1

  • prefixed_hash_values(array<string> values, string prefix [, boolean useIndexAsPrefix]) returns array<string> that each element has the specified prefix

  • sha1(string word [, int numFeatures]) returns a SHA-1 value

Feature paring

  • feature_pairs(feature_vector in array<string>, [, const string options]) - Returns a relation <string i, string j, double xi, double xj>

  • polynomial_features(feature_vector in array<string>) - Returns a feature vectorhaving polynomial feature space

  • powered_features(feature_vector in array<string>, int degree [, boolean truncate]) - Returns a feature vector having a powered feature space

Ranking

  • bpr_sampling(int userId, List<int> posItems [, const string options])- Returns a relation consists of <int userId, int itemId>

  • item_pairs_sampling(array<int|long> pos_items, const int max_item_id [, const string options])- Returns a relation consists of <int pos_item_id, int neg_item_id>

  • populate_not_in(list items, const int max_item_id [, const string options])- Returns a relation consists of <int item> that item does not exist in the given items

Feature scaling

  • l1_normalize(ftvec string) - Returned a L1 normalized value

  • l2_normalize(ftvec string) - Returned a L2 normalized value

  • rescale(value, min, max) - Returns rescaled value by min-max normalization

  • zscore(value, mean, stddev) - Returns a standard score (zscore)

Feature selection

  • chi2(array<array<number>> observed, array<array<number>> expected) - Returns chi2_val and p_val of each columns as <array<double>, array<double>>

  • snr(array<number> features, array<int> one-hot class label) - Returns Signal Noise Ratio for each feature as array<double>

Feature transformation and vectorization

  • add_field_indices(array<string> features) - Returns arrays of string that field indices (<field>:<feature>)* are augmented

  • binarize_label(int/long positive, int/long negative, ...) - Returns positive/negative records that are represented as (..., int label) where label is 0 or 1

  • categorical_features(array<string> featureNames, feature1, feature2, .. [, const string options]) - Returns a feature vector array<string>

  • ffm_features(const array<string> featureNames, feature1, feature2, .. [, const string options]) - Takes categorical variables and returns a feature vector array<string> in a libffm format <field>:<index>:<value>

  • indexed_features(double v1, double v2, ...) - Returns a list of features as array<string>: [1:v1, 2:v2, ..]

  • onehot_encoding(PRIMITIVE feature, ...) - Compute onehot encoded label for each feature

    WITH mapping as (
  select 
    m.f1, m.f2 
  from (
    select onehot_encoding(species, category) m
    from test
  ) tmp
)
select
  array(m.f1[t.species],m.f2[t.category],feature('count',count)) as sparse_features
from
  test t
  CROSS JOIN mapping m;

["2","8","count:9"]
["5","8","count:10"]
["1","6","count:101"]

  • quantified_features(boolean output, col1, col2, ...) - Returns an identified features in a dense array<double>

  • quantitative_features(array<string> featureNames, feature1, feature2, .. [, const string options]) - Returns a feature vector array<string>

  • vectorize_features(array<string> featureNames, feature1, feature2, .. [, const string options]) - Returns a feature vector array<string>

Geospatial functions

  • haversine_distance(double lat1, double lon1, double lat2, double lon2, [const boolean mile=false])::double - return distance between two locations in km [or miles] using haversine formula

    Usage: select latlon_distance(lat1, lon1, lat2, lon2) from ...

  • lat2tiley(double lat, int zoom)::int - Returns the tile number of the given latitude and zoom level

  • lon2tilex(double lon, int zoom)::int - Returns the tile number of the given longitude and zoom level

  • map_url(double lat, double lon, int zoom [, const string option]) - Returns a URL string

    OpenStreetMap: http://tile.openstreetmap.org/${zoom}/${xtile}/${ytile}.png
Google Maps: https://www.google.com/maps/@${lat},${lon},${zoom}z

  • tile(double lat, double lon, int zoom)::bigint - Returns a tile number 2^2n where n is zoom level. FUNC(lat,lon,zoom) = xtile(lon,zoom) + ytile(lat,zoom) * 2^zoom

    refer https://wiki.openstreetmap.org/wiki/Slippy_map_tilenames for detail

  • tilex2lon(int x, int zoom)::double - Returns longitude of the given tile x and zoom level

  • tiley2lat(int y, int zoom)::double - Returns latitude of the given tile y and zoom level

Distance measures

  • angular_distance(ftvec1, ftvec2) - Returns an angular distance of the given two vectors

    WITH docs as (
  select 1 as docid, array('apple:1.0', 'orange:2.0', 'banana:1.0', 'kuwi:0') as features
  union all
  select 2 as docid, array('apple:1.0', 'orange:0', 'banana:2.0', 'kuwi:1.0') as features
  union all
  select 3 as docid, array('apple:2.0', 'orange:0', 'banana:2.0', 'kuwi:1.0') as features
) 
select
  l.docid as doc1,
  r.docid as doc2,
  angular_distance(l.features, r.features) as distance,
  distance2similarity(angular_distance(l.features, r.features)) as similarity
from 
  docs l
  CROSS JOIN docs r
where
  l.docid != r.docid
order by 
  doc1 asc,
  distance asc;

doc1    doc2    distance        similarity
1       3       0.31678355      0.75942624
1       2       0.33333337      0.75
2       3       0.09841931      0.91039914
2       1       0.33333337      0.75
3       2       0.09841931      0.91039914
3       1       0.31678355      0.75942624

  • cosine_distance(ftvec1, ftvec2) - Returns a cosine distance of the given two vectors

    WITH docs as (
  select 1 as docid, array('apple:1.0', 'orange:2.0', 'banana:1.0', 'kuwi:0') as features
  union all
  select 2 as docid, array('apple:1.0', 'orange:0', 'banana:2.0', 'kuwi:1.0') as features
  union all
  select 3 as docid, array('apple:2.0', 'orange:0', 'banana:2.0', 'kuwi:1.0') as features
) 
select
  l.docid as doc1,
  r.docid as doc2,
  cosine_distance(l.features, r.features) as distance,
  distance2similarity(cosine_distance(l.features, r.features)) as similarity
from 
  docs l
  CROSS JOIN docs r
where
  l.docid != r.docid
order by 
  doc1 asc,
  distance asc;

doc1    doc2    distance        similarity
1       3       0.45566893      0.6869694
1       2       0.5     0.6666667
2       3       0.04742068      0.95472616
2       1       0.5     0.6666667
3       2       0.04742068      0.95472616
3       1       0.45566893      0.6869694

  • euclid_distance(ftvec1, ftvec2) - Returns the square root of the sum of the squared differences: sqrt(sum((x - y)^2))

    WITH docs as (
  select 1 as docid, array('apple:1.0', 'orange:2.0', 'banana:1.0', 'kuwi:0') as features
  union all
  select 2 as docid, array('apple:1.0', 'orange:0', 'banana:2.0', 'kuwi:1.0') as features
  union all
  select 3 as docid, array('apple:2.0', 'orange:0', 'banana:2.0', 'kuwi:1.0') as features
) 
select
  l.docid as doc1,
  r.docid as doc2,
  euclid_distance(l.features, r.features) as distance,
  distance2similarity(euclid_distance(l.features, r.features)) as similarity
from 
  docs l
  CROSS JOIN docs r
where
  l.docid != r.docid
order by 
  doc1 asc,
  distance asc;

doc1    doc2    distance        similarity
1       2       2.4494898       0.28989795
1       3       2.6457512       0.2742919
2       3       1.0     0.5
2       1       2.4494898       0.28989795
3       2       1.0     0.5
3       1       2.6457512       0.2742919

  • hamming_distance(integer A, integer B) - Returns Hamming distance between A and B

    select 
  hamming_distance(0,3) as c1, 
  hamming_distance("0","3") as c2 -- 0=0x00, 3=0x11
;

c1      c2
2       2

  • jaccard_distance(integer A, integer B [,int k=128]) - Returns Jaccard distance between A and B

    select 
  jaccard_distance(0,3) as c1, 
  jaccard_distance("0","3") as c2, -- 0=0x00, 0=0x11
  jaccard_distance(0,4) as c3
;

c1      c2      c3
0.03125 0.03125 0.015625

  • kld(double mu1, double sigma1, double mu2, double sigma2) - Returns KL divergence between two distributions

  • manhattan_distance(list x, list y) - Returns sum(|x - y|)

    WITH docs as (
  select 1 as docid, array('apple:1.0', 'orange:2.0', 'banana:1.0', 'kuwi:0') as features
  union all
  select 2 as docid, array('apple:1.0', 'orange:0', 'banana:2.0', 'kuwi:1.0') as features
  union all
  select 3 as docid, array('apple:2.0', 'orange:0', 'banana:2.0', 'kuwi:1.0') as features
) 
select
  l.docid as doc1,
  r.docid as doc2,
  manhattan_distance(l.features, r.features) as distance,
  distance2similarity(angular_distance(l.features, r.features)) as similarity
from 
  docs l
  CROSS JOIN docs r
where
  l.docid != r.docid
order by 
  doc1 asc,
  distance asc;

doc1    doc2    distance        similarity
1       2       4.0     0.75
1       3       5.0     0.75942624
2       3       1.0     0.91039914
2       1       4.0     0.75
3       2       1.0     0.91039914
3       1       5.0     0.75942624

  • minkowski_distance(list x, list y, double p) - Returns sum(|x - y|^p)^(1/p)

    WITH docs as (
  select 1 as docid, array('apple:1.0', 'orange:2.0', 'banana:1.0', 'kuwi:0') as features
  union all
  select 2 as docid, array('apple:1.0', 'orange:0', 'banana:2.0', 'kuwi:1.0') as features
  union all
  select 3 as docid, array('apple:2.0', 'orange:0', 'banana:2.0', 'kuwi:1.0') as features
) 
select
  l.docid as doc1,
  r.docid as doc2,
  minkowski_distance(l.features, r.features, 1) as distance1, -- p=1 (manhattan_distance)
  minkowski_distance(l.features, r.features, 2) as distance2, -- p=2 (euclid_distance)
  minkowski_distance(l.features, r.features, 3) as distance3, -- p=3
  manhattan_distance(l.features, r.features) as manhattan_distance,
  euclid_distance(l.features, r.features) as euclid_distance
from 
  docs l
  CROSS JOIN docs r
where
  l.docid != r.docid
order by 
  doc1 asc,
  distance1 asc;

doc1    doc2    distance1       distance2       distance3       manhattan_distance      euclid_distance
1       2       4.0     2.4494898       2.1544347       4.0     2.4494898
1       3       5.0     2.6457512       2.2239802       5.0     2.6457512
2       3       1.0     1.0     1.0     1.0     1.0
2       1       4.0     2.4494898       2.1544347       4.0     2.4494898
3       2       1.0     1.0     1.0     1.0     1.0
3       1       5.0     2.6457512       2.2239802       5.0     2.6457512

  • popcnt(a [, b]) - Returns a popcount value

    select 
  popcnt(3),
  popcnt("3"),  -- 3=0x11
  popcnt(array(1,3));

2       2       3

Locality-sensitive hashing

  • bbit_minhash(array<> features [, int numHashes]) - Returns a b-bits minhash value

  • minhash(ANY item, array<int|bigint|string> features [, constant string options]) - Returns n different k-depth signatures (i.e., clusterid) for each item <clusterid, item>

  • minhashes(array<> features [, int numHashes, int keyGroup [, boolean noWeight]]) - Returns minhash values

Similarity measures

  • angular_similarity(ftvec1, ftvec2) - Returns an angular similarity of the given two vectors

    WITH docs as (
  select 1 as docid, array('apple:1.0', 'orange:2.0', 'banana:1.0', 'kuwi:0') as features
  union all
  select 2 as docid, array('apple:1.0', 'orange:0', 'banana:2.0', 'kuwi:1.0') as features
  union all
  select 3 as docid, array('apple:2.0', 'orange:0', 'banana:2.0', 'kuwi:1.0') as features
) 
select
  l.docid as doc1,
  r.docid as doc2,
  angular_similarity(l.features, r.features) as similarity
from 
  docs l
  CROSS JOIN docs r
where
  l.docid != r.docid
order by 
  doc1 asc,
  similarity desc;

doc1    doc2    similarity
1       3       0.68321645
1       2       0.6666666
2       3       0.9015807
2       1       0.6666666
3       2       0.9015807
3       1       0.68321645

  • cosine_similarity(ftvec1, ftvec2) - Returns a cosine similarity of the given two vectors

    WITH docs as (
  select 1 as docid, array('apple:1.0', 'orange:2.0', 'banana:1.0', 'kuwi:0') as features
  union all
  select 2 as docid, array('apple:1.0', 'orange:0', 'banana:2.0', 'kuwi:1.0') as features
  union all
  select 3 as docid, array('apple:2.0', 'orange:0', 'banana:2.0', 'kuwi:1.0') as features
) 
select
  l.docid as doc1,
  r.docid as doc2,
  cosine_similarity(l.features, r.features) as similarity
from 
  docs l
  CROSS JOIN docs r
where
  l.docid != r.docid
order by 
  doc1 asc,
  similarity desc;

doc1    doc2    similarity
1       3       0.5443311
1       2       0.5
2       3       0.9525793
2       1       0.5
3       2       0.9525793
3       1       0.5443311

  • dimsum_mapper(array<string> row, map<int col_id, double norm> colNorms [, const string options]) - Returns column-wise partial similarities

  • distance2similarity(float d) - Returns 1.0 / (1.0 + d)

  • euclid_similarity(ftvec1, ftvec2) - Returns a euclid distance based similarity, which is 1.0 / (1.0 + distance), of the given two vectors

    WITH docs as (
  select 1 as docid, array('apple:1.0', 'orange:2.0', 'banana:1.0', 'kuwi:0') as features
  union all
  select 2 as docid, array('apple:1.0', 'orange:0', 'banana:2.0', 'kuwi:1.0') as features
  union all
  select 3 as docid, array('apple:2.0', 'orange:0', 'banana:2.0', 'kuwi:1.0') as features
) 
select
  l.docid as doc1,
  r.docid as doc2,
  euclid_similarity(l.features, r.features) as similarity
from 
  docs l
  CROSS JOIN docs r
where
  l.docid != r.docid
order by 
  doc1 asc,
  similarity desc;

doc1    doc2    similarity
1       2       0.28989795
1       3       0.2742919
2       3       0.5
2       1       0.28989795
3       2       0.5
3       1       0.2742919

  • jaccard_similarity(A, B [,int k]) - Returns Jaccard similarity coefficient of A and B

    WITH docs as (
  select 1 as docid, array('apple:1.0', 'orange:2.0', 'banana:1.0', 'kuwi:0') as features
  union all
  select 2 as docid, array('apple:1.0', 'orange:0', 'banana:2.0', 'kuwi:1.0') as features
  union all
  select 3 as docid, array('apple:2.0', 'orange:0', 'banana:2.0', 'kuwi:1.0') as features
) 
select
  l.docid as doc1,
  r.docid as doc2,
  jaccard_similarity(l.features, r.features) as similarity
from 
  docs l
  CROSS JOIN docs r
where
  l.docid != r.docid
order by 
  doc1 asc,
  similarity desc;

doc1    doc2    similarity
1       2       0.14285715
1       3       0.0
2       3       0.6
2       1       0.14285715
3       2       0.6
3       1       0.0

Evaluation

  • auc(array rankItems | double score, array correctItems | int label [, const int recommendSize = rankItems.size ]) - Returns AUC

  • average_precision(array rankItems, array correctItems [, const int recommendSize = rankItems.size]) - Returns MAP

  • f1score(array[int], array[int]) - Return a F1 score

  • fmeasure(array|int|boolean actual, array|int| boolean predicted [, const string options]) - Return a F-measure (f1score is the special with beta=1.0)

  • hitrate(array rankItems, array correctItems [, const int recommendSize = rankItems.size]) - Returns HitRate

  • logloss(double predicted, double actual) - Return a Logrithmic Loss

  • mae(double predicted, double actual) - Return a Mean Absolute Error

  • mrr(array rankItems, array correctItems [, const int recommendSize = rankItems.size]) - Returns MRR

  • mse(double predicted, double actual) - Return a Mean Squared Error

  • ndcg(array rankItems, array correctItems [, const int recommendSize = rankItems.size]) - Returns nDCG

  • precision_at(array rankItems, array correctItems [, const int recommendSize = rankItems.size]) - Returns Precision

  • r2(double predicted, double actual) - Return R Squared (coefficient of determination)

  • recall_at(array rankItems, array correctItems [, const int recommendSize = rankItems.size]) - Returns Recall

  • rmse(double predicted, double actual) - Return a Root Mean Squared Error

Sketching

  • approx_count_distinct(expr x [, const string options]) - Returns an approximation of count(DISTINCT x) using HyperLogLogPlus algorithm

  • bloom(string key) - Constructs a BloomFilter by aggregating a set of keys

    CREATE TABLE satisfied_movies AS 
  SELECT bloom(movieid) as movies
  FROM (
    SELECT movieid
    FROM ratings
    GROUP BY movieid
    HAVING avg(rating) >= 4.0
  ) t;

  • bloom_and(string bloom1, string bloom2) - Returns the logical AND of two bloom filters

    SELECT bloom_and(bf1, bf2) FROM xxx;

  • bloom_contains(string bloom, string key) or FUNC(string bloom, array<string> keys) - Returns true if the bloom filter contains all the given key(s). Returns false if key is null.

    WITH satisfied_movies as (
  SELECT bloom(movieid) as movies
  FROM (
    SELECT movieid
    FROM ratings
    GROUP BY movieid
    HAVING avg(rating) >= 4.0
  ) t
)
SELECT
  l.rating,
  count(distinct l.userid) as cnt
FROM
  ratings l 
  CROSS JOIN satisfied_movies r
WHERE
  bloom_contains(r.movies, l.movieid) -- includes false positive
GROUP BY 
  l.rating;

l.rating        cnt
1       1296
2       2770
3       5008
4       5824
5       5925

  • bloom_contains_any(string bloom, string key) or FUNC(string bloom, array<string> keys)- Returns true if the bloom filter contains any of the given key

    WITH data1 as (
  SELECT explode(array(1,2,3,4,5)) as id
),
data2 as (
  SELECT explode(array(1,3,5,6,8)) as id
),
bloom as (
  SELECT bloom(id) as bf
  FROM data1
)
SELECT 
  l.* 
FROM 
  data2 l
  CROSS JOIN bloom r
WHERE
  bloom_contains_any(r.bf, array(l.id))

  • bloom_not(string bloom) - Returns the logical NOT of a bloom filters

    SELECT bloom_not(bf) FROM xxx;

  • bloom_or(string bloom1, string bloom2) - Returns the logical OR of two bloom filters

    SELECT bloom_or(bf1, bf2) FROM xxx;

Ensemble learning

  • argmin_kld(float mean, float covar) - Returns mean or covar that minimize a KL-distance among distributions

    The returned value is (1.0 / (sum(1.0 / covar))) * (sum(mean / covar)

  • max_label(double value, string label) - Returns a label that has the maximum value

  • maxrow(ANY compare, ...) - Returns a row that has maximum value in the 1st argument

Bagging

  • voted_avg(double value) - Returns an averaged value by bagging for classification

  • weight_voted_avg(expr) - Returns an averaged value by considering sum of positive/negative weights

Decision trees and RandomForest

  • train_gradient_tree_boosting_classifier(array<double|string> features, int label [, string options]) - Returns a relation consists of <int iteration, int model_type, array<string> pred_models, double intercept, double shrinkage, array<double> var_importance, float oob_error_rate>

  • train_randomforest_classifier(array<double|string> features, int label [, const string options, const array<double> classWeights])- Returns a relation consists of <string model_id, double model_weight, string model, array<double> var_importance, int oob_errors, int oob_tests>

  • train_randomforest_regressor(array<double|string> features, double target [, string options]) - Returns a relation consists of <int model_id, int model_type, string model, array<double> var_importance, double oob_errors, int oob_tests>

  • decision_path(string modelId, string model, array<double|string> features [, const string options] [, optional array<string> featureNames=null, optional array<string> classNames=null]) - Returns a decision path for each prediction in array<string>

    SELECT
  t.passengerid,
  decision_path(m.model_id, m.model, t.features, '-classification')
FROM
  model_rf m
  LEFT OUTER JOIN
  test_rf t;
 | 892 | ["2 [0.0] = 0.0","0 [3.0] = 3.0","1 [696.0] != 107.0","7 [7.8292] <= 7.9104","1 [696.0] != 828.0","1 [696.0] != 391.0","0 [0.961038961038961, 0.03896103896103896]"] |

-- Show 100 frequent branches
WITH tmp as (
  SELECT
    decision_path(m.model_id, m.model, t.features, '-classification -no_verbose -no_leaf', array('pclass','name','sex','age','sibsp','parch','ticket','fare','cabin','embarked'), array('no','yes')) as path
  FROM
    model_rf m
    LEFT OUTER JOIN -- CROSS JOIN
    test_rf t
)
select
  r.branch,
  count(1) as cnt
from
  tmp l
  LATERAL VIEW explode(l.path) r as branch
group by
  r.branch
order by
  cnt desc
limit 100;

  • guess_attribute_types(ANY, ...) - Returns attribute types

    select guess_attribute_types(*) from train limit 1;
 Q,Q,C,C,C,C,Q,C,C,C,Q,C,Q,Q,Q,Q,C,Q

  • rf_ensemble(int yhat [, array<double> proba [, double model_weight=1.0]]) - Returns ensembled prediction results in <int label, double probability, array<double> probabilities>

  • tree_export(string model, const string options, optional array<string> featureNames=null, optional array<string> classNames=null) - exports a Decision Tree model as javascript/dot]

  • tree_predict(string modelId, string model, array<double|string> features [, const string options | const boolean classification=false]) - Returns a prediction result of a random forest in <int value, array<double> a posteriori> for classification and <double> for regression

XGBoost

  • train_xgboost(array<string|double> features, <int|double> target, const string options) - Returns a relation consists of <string model_id, array<string> pred_model>

    SELECT 
  train_xgboost(features, label, '-objective binary:logistic -iters 10') 
    as (model_id, model)
from (
  select features, label
  from xgb_input
  cluster by rand(43) -- shuffle
) shuffled;

  • xgboost_batch_predict(PRIMITIVE rowid, array<string|double> features, string model_id, array<string> pred_model [, string options]) - Returns a prediction result as (string rowid, array<double> predicted)

    select
  rowid, 
  array_avg(predicted) as predicted,
  avg(predicted[0]) as predicted0
from (
  select
    xgboost_batch_predict(rowid, features, model_id, model) as (rowid, predicted)
  from
    xgb_model l
    LEFT OUTER JOIN xgb_input r
) t
group by rowid;

  • xgboost_predict(PRIMITIVE rowid, array<string|double> features, string model_id, array<string> pred_model [, string options]) - Returns a prediction result as (string rowid, array<double> predicted)

    select
  rowid, 
  array_avg(predicted) as predicted,
  avg(predicted[0]) as predicted0
from (
  select
    xgboost_predict(rowid, features, model_id, model) as (rowid, predicted)
  from
    xgb_model l
    LEFT OUTER JOIN xgb_input r
) t
group by rowid;

  • xgboost_predict_one(PRIMITIVE rowid, array<string|double> features, string model_id, array<string> pred_model [, string options]) - Returns a prediction result as (string rowid, double predicted)

    select
  rowid, 
  avg(predicted) as predicted
from (
  select
    xgboost_predict_one(rowid, features, model_id, model) as (rowid, predicted)
  from
    xgb_model l
    LEFT OUTER JOIN xgb_input r
) t
group by rowid;

  • xgboost_predict_triple(PRIMITIVE rowid, array<string|double> features, string model_id, array<string> pred_model [, string options]) - Returns a prediction result as (string rowid, string label, double probability)

    select
  rowid,
  label,
  avg(prob) as prob
from (
  select
    xgboost_predict_triple(rowid, features, model_id, model) as (rowid, label, prob)
  from
    xgb_model l
    LEFT OUTER JOIN xgb_input r
) t
group by rowid, label;

  • xgboost_version() - Returns the version of xgboost

    SELECT xgboost_version();

Term Vector Model

  • bm25(double termFrequency, int docLength, double avgDocLength, int numDocs, int numDocsWithTerm [, const string options]) - Return an Okapi BM25 score in double. Refer http://hivemall.incubator.apache.org/userguide/ft_engineering/bm25.html for usage

  • tf(string text) - Return a term frequency in <string, float>

  • tfidf(double termFrequency, long numDocs, const long totalNumDocs) - Return a smoothed TFIDF score in double.

NLP

  • stoptags_exclude(array<string> excludeTags, [, const string lang='ja']) - Returns stoptags excluding given tags

    SELECT stoptags_exclude(array('名詞-固有名詞', '形容詞'))

  • tokenize_cn(String line [, const list<string> stopWords]) - returns tokenized strings in array<string>

  • tokenize_ja(String line [, const string mode = "normal", const array<string> stopWords, const array<string> stopTags, const array<string> userDict (or const string userDictURL)]) - returns tokenized strings in array<string>

    select tokenize_ja("kuromojiを使った分かち書きのテストです。第二引数にはnormal/search/extendedを指定できます。デフォルトではnormalモードです。");

> ["kuromoji","使う","分かち書き","テスト","第","二","引数","normal","search","extended","指定","デフォルト","normal"," モード"]

  • tokenize_ja_neologd(String line [, const string mode = "normal", const array<string> stopWords, const array<string> stopTags, const array<string> userDict (or const string userDictURL)]) - returns tokenized strings in array<string>

    select tokenize_ja_neologd("kuromojiを使った分かち書きのテストです。第二引数にはnormal/search/extendedを指定できます。デフォルトではnormalモードです。");

> ["kuromoji","使う","分かち書き","テスト","第","二","引数","normal","search","extended","指定","デフォルト","normal"," モード"]

  • tokenize_ko(String line [, const string mode = "discard" (or const string opts), const array<string> stopWords, const array<string> stopTags, const array<string> userDict (or const string userDictURL)]) - returns tokenized strings in array<string>

    select tokenize_ko("소설 무궁화꽃이 피었습니다.");

> ["소설","무궁","화","꽃","피"]

Others

  • hivemall_version() - Returns the version of Hivemall

    SELECT hivemall_version();

  • lr_datagen(options string) - Generates a logistic regression dataset

    WITH dual AS (SELECT 1) SELECT lr_datagen('-n_examples 1k -n_features 10') FROM dual;

  • bm25(double termFrequency, int docLength, double avgDocLength, int numDocs, int numDocsWithTerm [, const string options]) - Return an Okapi BM25 score in double. Refer http://hivemall.incubator.apache.org/userguide/ft_engineering/bm25.html for usage

  • tf(string text) - Return a term frequency in <string, float>

  • tfidf(double termFrequency, long numDocs, const long totalNumDocs) - Return a smoothed TFIDF score in double.

Apache Hivemall is an effort undergoing incubation at The Apache Software Foundation (ASF), sponsored by the Apache Incubator.

results matching ""

    No results matching ""