主动学习实验功能开发

ruoyi-modules/management-platform/src/main/java/com/ruoyi/platform/domain/ActiveLearn.java

@@ -20,47 +20,68 @@ public class ActiveLearn {
     @ApiModelProperty(value = "实验描述")
     private String description;
 
-    @ApiModelProperty(value = "数据集")
-    private String dataset;
+    @ApiModelProperty(value = "任务类型:classification, regression")
+    private String taskType;
+
+    @ApiModelProperty(value = "模型文件路径")
+    private String modelPath;
 
-    @ApiModelProperty(value = "数据集csv文件中哪几列是预测目标列，逗号分隔")
-    private String targetColumns;
+    @ApiModelProperty(value = "模型类名称")
+    private String modelClassName;
 
     @ApiModelProperty(value = "分类算法")
-    private String classifierType;
+    private String classifierAlg;
+
+    @ApiModelProperty(value = "回归算法")
+    private String regressorAlg;
+
+    @ApiModelProperty(value = "dataset文件路径")
+    private String dataset;
 
-    @ApiModelProperty(value = "每次查询个数")
-    private Integer queryBatchSize;
+    @ApiModelProperty(value = "dataset类名")
+    private String datasetClassName;
 
-    @ApiModelProperty(value = "停止判则")
-    private String stoppingCriterion;
+    @ApiModelProperty(value = "数据集文件路径")
+    private String datasetPath;
 
-    @ApiModelProperty(value = "stopping_criterion为num_of_queries时传入，查询次数")
-    private Integer numOfQueries;
+    @ApiModelProperty(value = "数据量")
+    private Integer dataSize;
 
-    @ApiModelProperty(value = "stopping_criterion为cost_limit时传入，成本限制")
-    private Double costLimit;
+    @ApiModelProperty(value = "是否随机打乱")
+    private Boolean shuffle;
 
-    @ApiModelProperty(value = "stopping_criterion为percent_of_unlabel时传入，未标记比例")
-    private Double percentOfUnlabel;
+    @ApiModelProperty(value = "训练集数据量")
+    private Integer trainSize;
 
-    @ApiModelProperty(value = "stopping_criterion为time_limit时传入，时间限制")
-    private Double timeLimit;
+    @ApiModelProperty(value = "初始训练数据量")
+    private Integer nInitial;
 
-    @ApiModelProperty(value = "查询策略")
+    @ApiModelProperty(value = "查询次数")
+    private Integer nQueries;
+
+    @ApiModelProperty(value = "每次查询数据量")
+    private Integer nInstances;
+
+    @ApiModelProperty(value = "查询策略:uncertainty_sampling, uncertainty_batch_sampling, max_std_sampling, expected_improvement, upper_confidence_bound")
     private String queryStrategy;
 
-    @ApiModelProperty(value = "实验次数")
-    private Integer numOfExperiment;
+    @ApiModelProperty(value = "loss文件路径")
+    private String lossPath;
+
+    @ApiModelProperty(value = "loss类名")
+    private String lossClassName;
+
+    @ApiModelProperty(value = "多少轮查询保存一次模型参数")
+    private Integer nCheckpoint;
 
-    @ApiModelProperty(value = "测试集比率")
-    private Double testRatio;
+    @ApiModelProperty(value = "batch_size")
+    private Integer batchSize;
 
-    @ApiModelProperty(value = "初始使用标记数据比率")
-    private Double initialLabelRate;
+    @ApiModelProperty(value = "epochs")
+    private Integer epochs;
 
-    @ApiModelProperty(value = "指标")
-    private String performanceMetric;
+    @ApiModelProperty(value = "学习率")
+    private Float lr;
 
     private Integer state;
 

ruoyi-modules/management-platform/src/main/java/com/ruoyi/platform/service/impl/ActiveLearnServiceImpl.java

@@ -42,8 +42,16 @@ public class ActiveLearnServiceImpl implements ActiveLearnService {
         String username = SecurityUtils.getLoginUser().getUsername();
         activeLearn.setCreateBy(username);
         activeLearn.setUpdateBy(username);
+
         String datasetJson = JacksonUtil.toJSONString(activeLearnVo.getDataset());
         activeLearn.setDataset(datasetJson);
+        String modelJson = JacksonUtil.toJSONString(activeLearnVo.getModelPath());
+        activeLearn.setModelPath(modelJson);
+        String datasetPathJson = JacksonUtil.toJSONString(activeLearnVo.getDatasetPath());
+        activeLearn.setDatasetPath(datasetPathJson);
+        String lossPathJson = JacksonUtil.toJSONString(activeLearnVo.getLossPath());
+        activeLearn.setLossPath(lossPathJson);
+
         activeLearnDao.save(activeLearn);
         return activeLearn;
     }
@@ -59,8 +67,16 @@ public class ActiveLearnServiceImpl implements ActiveLearnService {
         BeanUtils.copyProperties(activeLearnVo, activeLearn);
         String username = SecurityUtils.getLoginUser().getUsername();
         activeLearn.setUpdateBy(username);
+
         String datasetJson = JacksonUtil.toJSONString(activeLearnVo.getDataset());
         activeLearn.setDataset(datasetJson);
+        String modelJson = JacksonUtil.toJSONString(activeLearnVo.getModelPath());
+        activeLearn.setModelPath(modelJson);
+        String datasetPathJson = JacksonUtil.toJSONString(activeLearnVo.getDatasetPath());
+        activeLearn.setDatasetPath(datasetPathJson);
+        String lossPathJson = JacksonUtil.toJSONString(activeLearnVo.getLossPath());
+        activeLearn.setLossPath(lossPathJson);
+
         activeLearnDao.edit(activeLearn);
 
         return "修改成功";
@@ -71,9 +87,18 @@ public class ActiveLearnServiceImpl implements ActiveLearnService {
         ActiveLearn activeLearn = activeLearnDao.getActiveLearnById(id);
         ActiveLearnVo activeLearnVo = new ActiveLearnVo();
         BeanUtils.copyProperties(activeLearn, activeLearnVo);
+        if (StringUtils.isNotEmpty(activeLearn.getDataset())) {
+            activeLearnVo.setDatasetPath(JsonUtils.jsonToMap(activeLearn.getDatasetPath()));
+        }
+        if (StringUtils.isNotEmpty(activeLearn.getModelPath())) {
+            activeLearnVo.setModelPath(JsonUtils.jsonToMap(activeLearn.getModelPath()));
+        }
         if (StringUtils.isNotEmpty(activeLearn.getDataset())) {
             activeLearnVo.setDataset(JsonUtils.jsonToMap(activeLearn.getDataset()));
         }
+        if (StringUtils.isNotEmpty(activeLearn.getLossPath())) {
+            activeLearnVo.setLossPath(JsonUtils.jsonToMap(activeLearn.getLossPath()));
+        }
         return activeLearnVo;
     }
 
@@ -101,7 +126,11 @@ public class ActiveLearnServiceImpl implements ActiveLearnService {
 
         ActiveLearnVo activeLearnParam = new ActiveLearnVo();
         BeanUtils.copyProperties(activeLearn, activeLearnParam);
+        activeLearnParam.setDatasetPath(JsonUtils.jsonToMap(activeLearn.getDatasetPath()));
         activeLearnParam.setDataset(JsonUtils.jsonToMap(activeLearn.getDataset()));
+        activeLearnParam.setModelPath(JsonUtils.jsonToMap(activeLearn.getModelPath()));
+        activeLearnParam.setLossPath(JsonUtils.jsonToMap(activeLearn.getLossPath()));
+
         String param = JsonUtils.objectToJson(activeLearnParam);
 
         // todo 调argo转换接口

ruoyi-modules/management-platform/src/main/java/com/ruoyi/platform/vo/ActiveLearnVo.java

@@ -21,49 +21,68 @@ public class ActiveLearnVo {
     @ApiModelProperty(value = "实验描述")
     private String description;
 
-    /**
-     * 对应数据集
-     */
+    @ApiModelProperty(value = "任务类型:classification, regression")
+    private String taskType;
+
+    @ApiModelProperty(value = "模型文件路径")
+    private Map<String,Object> modelPath;
+
+    @ApiModelProperty(value = "模型类名称")
+    private String modelClassName;
+
+    @ApiModelProperty(value = "分类算法")
+    private String classifierAlg;
+
+    @ApiModelProperty(value = "分类算法")
+    private String regressorAlg;
+
+    @ApiModelProperty(value = "dataset文件路径")
     private Map<String,Object> dataset;
 
-    @ApiModelProperty(value = "数据集csv文件中哪一列是预测目标列，逗号分隔")
-    private String targetColumns;
+    @ApiModelProperty(value = "dataset类名")
+    private String datasetClassName;
 
-    @ApiModelProperty(value = "分类算法：logistic_regression（逻辑回归），decision_tree（决策树），random_forest（随机森林），SVM（支持向量机），naive_bayes（朴素贝叶斯），GBM（梯度提升机）")
-    private String classifierType;
+    @ApiModelProperty(value = "数据集")
+    private Map<String,Object> datasetPath;
 
-    @ApiModelProperty(value = "每次查询个数")
-    private Integer queryBatchSize;
+    @ApiModelProperty(value = "数据量")
+    private Integer dataSize;
 
-    @ApiModelProperty(value = "停止判则：num_of_queries（查询次数），percent_of_unlabel（未标记样本比例），time_limit（时间限制）")
-    private String stoppingCriterion;
+    @ApiModelProperty(value = "是否随机打乱")
+    private Boolean shuffle;
 
-    @ApiModelProperty(value = "stopping_criterion为num_of_queries时传入，查询次数")
-    private Integer numOfQueries;
+    @ApiModelProperty(value = "训练集数据量")
+    private Integer trainSize;
 
-//    @ApiModelProperty(value = "stopping_criterion为cost_limit时传入，成本限制")
-//    private Double costLimit;
+    @ApiModelProperty(value = "初始训练数据量")
+    private Integer nInitial;
 
-    @ApiModelProperty(value = "stopping_criterion为percent_of_unlabel时传入，未标记比例")
-    private Double percentOfUnlabel;
+    @ApiModelProperty(value = "查询次数")
+    private Integer nQueries;
 
-    @ApiModelProperty(value = "stopping_criterion为time_limit时传入，时间限制")
-    private Double timeLimit;
+    @ApiModelProperty(value = "每次查询数据量")
+    private Integer nInstances;
 
-    @ApiModelProperty(value = "查询策略：Uncertainty(不确定性)，QBC（委员会查询），Random（随机），GraphDensity（图密度）")
+    @ApiModelProperty(value = "查询策略:uncertainty_sampling, uncertainty_batch_sampling, max_std_sampling, expected_improvement, upper_confidence_bound")
     private String queryStrategy;
 
-    @ApiModelProperty(value = "实验次数")
-    private Integer numOfExperiment;
+    @ApiModelProperty(value = "loss文件路径")
+    private Map<String,Object> lossPath;
+
+    @ApiModelProperty(value = "loss类名")
+    private String lossClassName;
+
+    @ApiModelProperty(value = "多少轮查询保存一次模型参数")
+    private Integer nCheckpoint;
 
-    @ApiModelProperty(value = "测试集比率")
-    private Double testRatio;
+    @ApiModelProperty(value = "batch_size")
+    private Integer batchSize;
 
-    @ApiModelProperty(value = "初始使用标记数据比率")
-    private Double initialLabelRate;
+    @ApiModelProperty(value = "epochs")
+    private Integer epochs;
 
-    @ApiModelProperty(value = "指标：accuracy_score，roc_auc_score，get_fps_tps_thresholds，hamming_loss，one_error，coverage_error")
-    private String performanceMetric;
+    @ApiModelProperty(value = "学习率")
+    private Float lr;
 
     private Integer state;
 

ruoyi-modules/management-platform/src/main/resources/mapper/managementPlatform/ActiveLearnDaoMapper.xml

@@ -2,18 +2,18 @@
 <!DOCTYPE mapper PUBLIC "-//mybatis.org//DTD Mapper 3.0//EN" "http://mybatis.org/dtd/mybatis-3-mapper.dtd">
 <mapper namespace="com.ruoyi.platform.mapper.ActiveLearnDao">
     <insert id="save">
-        insert into active_learn(name, description, dataset, target_columns, classifier_type, query_batch_size,
-                                 stopping_criterion,
-                                 num_of_queries, cost_limit, percent_of_unlabel, time_limit, query_strategy,
-                                 num_of_experiment, test_ratio,
-                                 initial_label_rate, performance_metric, create_by, update_by)
-        values (#{activeLearn.name}, #{activeLearn.description}, #{activeLearn.dataset}, #{activeLearn.targetColumns},
-                #{activeLearn.classifierType}, #{activeLearn.queryBatchSize}, #{activeLearn.stoppingCriterion},
-                #{activeLearn.numOfQueries},
-                #{activeLearn.costLimit}, #{activeLearn.percentOfUnlabel}, #{activeLearn.timeLimit},
+        insert into active_learn(name, description, task_type, model_path, model_class_name, classifier_alg,
+                                 regressor_alg, dataset, dataset_class_name, dataset_path, data_size,
+                                 shuffle, train_size, n_initial, n_queries, n_instances, query_strategy,
+                                 loss_path, loss_class_name, n_checkpoint, batch_size, epochs, lr, create_by, update_by)
+        values (#{activeLearn.name}, #{activeLearn.description}, #{activeLearn.taskType}, #{activeLearn.modelPath},
+                #{activeLearn.modelClassName}, #{activeLearn.classifierAlg}, #{activeLearn.regressorAlg},
+                #{activeLearn.dataset}, #{activeLearn.datasetClassName}, #{activeLearn.datasetPath},
+                #{activeLearn.dataSize}, #{activeLearn.shuffle}, #{activeLearn.trainSize},
+                #{activeLearn.nInitial}, #{activeLearn.nQueries}, #{activeLearn.nInstances},
                 #{activeLearn.queryStrategy},
-                #{activeLearn.numOfExperiment}, #{activeLearn.testRatio}, #{activeLearn.initialLabelRate},
-                #{activeLearn.performanceMetric},
+                #{activeLearn.lossPath}, #{activeLearn.lossClassName}, #{activeLearn.nCheckpoint},
+                #{activeLearn.batchSize}, #{activeLearn.epochs}, #{activeLearn.lr},
                 #{activeLearn.createBy}, #{activeLearn.updateBy})
     </insert>
 
@@ -26,47 +26,68 @@
             <if test="activeLearn.description != null and activeLearn.description !=''">
                 description = #{activeLearn.description},
             </if>
+            <if test="activeLearn.taskType != null and activeLearn.taskType !=''">
+                task_type = #{activeLearn.taskType},
+            </if>
+            <if test="activeLearn.modelPath != null and activeLearn.modelPath !=''">
+                model_path = #{activeLearn.modelPath},
+            </if>
+            <if test="activeLearn.modelClassName != null and activeLearn.modelClassName !=''">
+                model_class_name = #{activeLearn.modelClassName},
+            </if>
+            <if test="activeLearn.classifierAlg != null and activeLearn.classifierAlg !=''">
+                classifier_alg = #{activeLearn.classifierAlg},
+            </if>
+            <if test="activeLearn.regressorAlg != null and activeLearn.regressorAlg !=''">
+                regressor_alg = #{activeLearn.regressorAlg},
+            </if>
             <if test="activeLearn.dataset != null and activeLearn.dataset !=''">
                 dataset = #{activeLearn.dataset},
             </if>
-            <if test="activeLearn.targetColumns != null and activeLearn.targetColumns !=''">
-                target_columns = #{activeLearn.targetColumns},
+            <if test="activeLearn.datasetClassName != null and activeLearn.datasetClassName !=''">
+                dataset_class_name = #{activeLearn.datasetClassName},
             </if>
-            <if test="activeLearn.classifierType != null and activeLearn.classifierType !=''">
-                classifier_type = #{activeLearn.classifierType},
+            <if test="activeLearn.datasetPath != null and activeLearn.datasetPath !=''">
+                dataset_path = #{activeLearn.datasetPath},
             </if>
-            <if test="activeLearn.queryBatchSize != null">
-                query_batch_size = #{activeLearn.queryBatchSize},
+            <if test="activeLearn.dataSize != null">
+                data_size = #{activeLearn.dataSize},
             </if>
-            <if test="activeLearn.stoppingCriterion != null and activeLearn.stoppingCriterion !=''">
-                stopping_criterion = #{activeLearn.stoppingCriterion},
+            <if test="activeLearn.shuffle != null">
+                shuffle = #{activeLearn.shuffle},
             </if>
-            <if test="activeLearn.numOfQueries != null">
-                num_of_queries = #{activeLearn.numOfQueries},
+            <if test="activeLearn.trainSize != null">
+                train_size = #{activeLearn.trainSize},
             </if>
-            <if test="activeLearn.costLimit != null">
-                cost_limit = #{activeLearn.costLimit},
+            <if test="activeLearn.nInitial != null">
+                n_initial = #{activeLearn.nInitial},
             </if>
-            <if test="activeLearn.percentOfUnlabel != null">
-                percent_of_unlabel = #{activeLearn.percentOfUnlabel},
+            <if test="activeLearn.nQueries != null">
+                n_queries = #{activeLearn.nQueries},
             </if>
-            <if test="activeLearn.timeLimit != null">
-                time_limit = #{activeLearn.timeLimit},
+            <if test="activeLearn.nInstances != null">
+                n_instances = #{activeLearn.nInstances},
             </if>
             <if test="activeLearn.queryStrategy != null and activeLearn.queryStrategy !=''">
                 query_strategy = #{activeLearn.queryStrategy},
             </if>
-            <if test="activeLearn.numOfExperiment != null">
-                num_of_experiment = #{activeLearn.numOfExperiment},
+            <if test="activeLearn.lossPath != null and activeLearn.lossPath !=''">
+                loss_path = #{activeLearn.lossPath},
+            </if>
+            <if test="activeLearn.lossClassName != null and activeLearn.lossClassName !=''">
+                loss_class_name = #{activeLearn.lossClassName},
+            </if>
+            <if test="activeLearn.nCheckpoint != null">
+                n_checkpoint = #{activeLearn.nCheckpoint},
             </if>
-            <if test="activeLearn.testRatio != null">
-                test_ratio = #{activeLearn.testRatio},
+            <if test="activeLearn.batchSize != null">
+                batch_size = #{activeLearn.batchSize},
             </if>
-            <if test="activeLearn.initialLabelRate != null">
-                initial_label_rate = #{activeLearn.initialLabelRate},
+            <if test="activeLearn.epochs != null">
+                epochs = #{activeLearn.epochs},
             </if>
-            <if test="activeLearn.performanceMetric != null and activeLearn.performanceMetric !=''">
-                performance_metric = #{activeLearn.performanceMetric},
+            <if test="activeLearn.lr != null">
+                lr = #{activeLearn.lr},
             </if>
             <if test="activeLearn.updateBy != null and activeLearn.updateBy !=''">
                 update_by = #{activeLearn.updateBy},