v0.15.0

Improves

• Add Multiclass BSGD #548
• Add RDF #554
• Add ABOD #557
• Add CLIQUE #562
• Add PROCLUS #566
• Add ORCLUS #572
• Improve ONNX #576
• Add FINDIT #584
• Add ZTP #586
• Implement numpy like broadcasting #594
• Add DBCLASD #600
• Add property in CalculatedGraph #602
• Apply some parameter from other layer in NNs #606
• Add toDot method in computational graph #614
• Add some ONNX layer and combine binary operations in one place #616
• Add and improve reduce layers #618

Breaking changes

• Change some export to default #558
• Remove some properties and fix some bugs #570
• Change ONNX importer to return array of object #588
• Change error class and add tests #593
• Remove LossLayer #605
• Change power layer to two input operator #608
• Accept SAME_UPPER and SAME_LOWER at auto_pad #610
• Separate classifier from HMM model #612

Minor changes

• Use includes instead of indexOf for existence checks #590
• Remove dependency and improve bit #598
• Increase retry times of some tests #620

Others

• Add workerIdleMemoryLimit option #596
• Update version to 0.15.0 #626

Full Changelog: v0.14.0...v0.15.0

v0.14.0

Improve

• Add ILK/SILK #447
• Remove shifting process in first order online learning #457
• Remove shifting process in PA #460
• Improve ComputationalGraph #465
• Randomly select data in (Kernelized) Pegasos #467
• Add Growing SOM #473
• Add neuralnetwork layer and onnx layer #475
• Add COLL #477
• Add BPA #482
• Add LTSA #484
• Add BOGD #488
• Accept different pad value in averagepool #500
• Add method to generate maze in grid environment #513
• Improve graph #525
• Implement Schur decomposition #527
• Implement Gomoku RL environment #535
• Add MADALINE #537
• Add Blackjack environment #543
• Improve graph #545

Breaking change

• Do not export qt function #448
• Change to get new data with fit only #468
• Remove unnecessary arguments from the HMM model #491
• Fix some eslint error/warning #523
• Change name of linear layer to identity layer #533

Bug fix

• Fix to update parameter k #456
• Fix the index of y in finding process #462
• Fix the index of y in finding process #464
• Fix correlation matrix of LTSA #486
• Don't import NeuralnetworkException from layer class #495
• Fix calculation when channel dim is 1 #496
• Fix power transformation calculation #498
• Fix some bugs and add tests #504
• Fix LSA process if the argument is default #506

Minor improve

• Decrease learning rate of COLL #479
• Add JSDoc for neuralnetwork layers #521

Others

• Make test mode in onnx.sh #444
• Update dependencies #469
• Add JSDoc for Edge class of graph #472
• Increase retry times #492
• Change exit code if some scripts are failed in onnx.sh #502
• Add tests #515
• Minify HTML files #518
• Add devDependencies #520
• Include NOTICE.md in the package #538
• Update version to 0.14.0 #546

Full Changelog: v0.13.0...v0.14.0

v0.13.0

Improve

• Add Selective sampling Winnow #380
• Add Forgetron #385
• Add Projectron and Projectron++ #389
• Add IKNN #394
• Add Weighted KNN #397
• Add MIRA #400
• Improve graph class #404
• Implement ONNX importing #408
• Add Banditron #412
• Add Ballseptron #414
• Accept function at kernel parameter #422
• Add evaluation indicator #425
• Add Stoptron #429
• Add BSGD #433
• Add Tighter perceptron #436
• Add Tightest perceptron #438

Bug fix

• Fix kernel function #387
• Fix IKNN #396
• Fix export statement in onnx_importer.js #410
• Fix regularized incomplete beta function #440

Minor changes

• Fix JSDoc of Selective sampling Winnow #382
• Simplify Mean shift and LOF implementation #398
• Fix JSDoc of ONNX #416
• Add script for onnx #419
• Create all onnx file in one python script #420
• Improve some tests #424
• Update README.md #427
• Update version to 0.13.0 #441

Full Changelog: v0.12.0...v0.13.0

v0.12.0

Improve

• Accept tensor class in layers #297
• Improve conv layer #299
• Add pooling layers #303
• Analytically compute det and eigenvalues of 4th order matrix #305
• Add ADAMENN #308
• Improve A2C #327
• Add Semi-supervised Naive Bayes #331
• Add CLUES #334
• Add Winnow #337
• Add Voted perceptron #342
• Add learning rate parameter for Voted perceptron #344
• Add Kernelized perceptron #348
• Add Pegasos #354
• Add Selective sampling perceptron and its adaptive version and Selective sampling second-order Perceptron #355
• Add Margin perceptron (PAM) #361
• Add PAUM #363
• Add Kernelized pegasos #364
• Add Shifting perceptron and RBP #367
• Add Budget perceptron #368
• Add internal parameter to the gaussian kernel of the Kernelized perceptron #370
• Implement fixed-size cache version of Budget perceptron #372

Bug fix

• Fix and improve convolutional layer #296
• Fix broadcastOperate of tensor class #302
• Fix importance calculation of DecisionTree class #313
• Improve and fix dropout layer #318
• Change default threshold of Winnow #339

Breaking changes

• Separate averaged perceptron from perceptron #351
• Fix return type of the Mean shift fit function #359

Minor changes

• Improve test #314
• Improve isZero function of Matrix class #325
• Change some fixed values to properties #335
• Improve test #350
• Format and fix some tests #374
• Update version to 0.12.0 #375

Full Changelog: v0.11.0...v0.12.0

v0.11.0

Improve

• Add MONA #258
• Add NLMeans #264
• Improve reversi environment #266
• Add MLLE #269
• Add CHAMELEON #283
• Add multiclass perceptron #286
• Add NNBCA #288

Breaking changes

• Remove set property from almost all models #276

Bug fix

• Fix hessian of HLLE #284

Minor changes

• Use hasOwn instead of hasOwnProperty #247
• Improve generating sequential numbers #260
• Keep tree structure in DIANA clustering #261
• Update dependencies #274
• Improve tests #278
• Remove some unreachable codes #279
• Change return value of NNBCA #290

Full Changelog: v0.10.0...v0.11.0

v0.10.0

Improve

• Add RKOF #214

Breaking changes

• Remove Tree class dependency #216
• Remove Tree class #241

Bug fix

• Fix AdditiveCouplingLayer and add tests #224

Improve tests

• Improve some tests #230

Full Changelog: v0.9.0...v0.10.0

v0.9.0

Improve

• Speed up digamma function in VBGMM #155
• Add COF #160
• Add Mutual kNN #167
• Add some method in Matrix class #169
• Add ODIN #171
• Add LDOF #173
• Add LOCI #181
• Add INFLO #183
• Add LoOP #185
• Add C2P #187
• Add LDF #189
• Add KDEOS #192
• Add ENN #198
• Add RDOS #200

Breaking changes

• Improve FuzzyKNN #161
• Add SMI CPD and remove some CPD models #179

Bug fix

• Fix minkowski calculation #197

Minor changes

• Improve SNE/tSNE #158
• Fix return type of GTM predict function #177
• Fix parameter types #202

Tests

• Test faster #156
• Add tests #175
• Increate retry number #203

Full Changelog: v0.8.0...v0.9.0

v0.8.0

Improve

• Return loss value in some NN models #148

Breaking changes

• Remove unused property #147

Minor changes

• Increase the number of retries #143
• Add description of JSDoc #146
• Show slow test #150

Full Changelog: v0.7.2...v0.8.0

v0.7.2

Minor changes

• Fix breaker env and to be able to play manually #111
• Use value instead _value #117

Full Changelog: v0.7.1...v0.7.2

v0.7.1

Big fix

• Fix reshape layer #101

Full Changelog: v0.7.0...v0.7.1