【HTM 算法】第一个例子 Online Prediction Framework (OPF)

叁歲伎倆 2023-05-30 09:09 97阅读 0赞

**HTM** **Python**

安装好 NuPIC 后，第一个例子。安装教程 [前往][Link 1]  
官网地址：[http://nupic.docs.numenta.org/1.0.5/quick-start/opf.html][http_nupic.docs.numenta.org_1.0.5_quick-start_opf.html]  
总共三个文件，其相对路径关系如图所示：

*  opf
    
     *  data
        
         *  gymdata.csv
     *  examples
        
         *  test.py
     *  params
        
         *  model.yaml

### Python 代码 ###

新建一个py文件，复制官网文档 [该例][http_nupic.docs.numenta.org_1.0.5_quick-start_opf.html] 代码如下：

import csv
    import datetime
    import os
    import yaml
    from itertools import islice
    
    from nupic.frameworks.opf.model_factory import ModelFactory
    
    _NUM_RECORDS = 3000
    _EXAMPLE_DIR = os.path.dirname(os.path.abspath(__file__))
    _INPUT_FILE_PATH = os.path.join(_EXAMPLE_DIR, os.pardir, "data", "gymdata.csv")
    _PARAMS_PATH = os.path.join(_EXAMPLE_DIR, os.pardir, "params", "model.yaml")
    
    
    # create model according to the file of params/model.yaml
    def createModel():
      with open(_PARAMS_PATH, "r") as f:
        modelParams = yaml.safe_load(f)
      return ModelFactory.create(modelParams)
    
    
    # run Hotgym. 
    def runHotgym(numRecords):
      model = createModel()
      model.enableInference({ "predictedField": "consumption"})
      # open the csv file as fin
      with open(_INPUT_FILE_PATH) as fin:
        reader = csv.reader(fin)
        # skip three line useless data
        headers = reader.next()
        reader.next()
        reader.next()
    	
        results = []
        # for each line 
        for record in islice(reader, numRecords):
          # Create a dictionary with field names as keys, row values as values.
          modelInput = dict(zip(headers, record))
          # Convert string consumption to float value.
          modelInput["consumption"] = float(modelInput["consumption"])
          # Convert timestamp string to Python datetime.
          modelInput["timestamp"] = datetime.datetime.strptime(
            modelInput["timestamp"], "%m/%d/%y %H:%M")
          # Push the data into the model and get back results.
          result = model.run(modelInput)
          bestPredictions = result.inferences["multiStepBestPredictions"]
          allPredictions = result.inferences["multiStepPredictions"]
          # Confidence values are keyed by prediction value in multiStepPredictions. 
          oneStep = bestPredictions[1]
          oneStepConfidence = allPredictions[1][oneStep]
          fiveStep = bestPredictions[5]
          fiveStepConfidence = allPredictions[5][fiveStep]
          result = (oneStep, oneStepConfidence * 100,
                    fiveStep, fiveStepConfidence * 100)
          print "1-step: {:16} ({:4.4}%)\t 5-step: {:16} ({:4.4}%)".format(*result)
          results.append(result)
        return results
    
    
    if __name__ == "__main__":
      runHotgym(_NUM_RECORDS)

### 添加一个yaml文件 ###

需要注意：如果不修改上面的代码，则需要在该py文件的上级目录下新建文件夹，起名为 `params` ，并且在此params目录下新建一个yaml文件，起名为 `model.yaml` ，填写内容如下：

model: HTMPrediction
    version: 1
    
    aggregationInfo:
      fields:
      - [consumption, mean]
      microseconds: 0
      milliseconds: 0
      minutes: 0
      months: 0
      seconds: 0
      hours: 1
      days: 0
      weeks: 0
      years: 0
    
    predictAheadTime: null
    
    modelParams:
      inferenceType: TemporalMultiStep
      sensorParams:
        verbosity: 0
        encoders:
          consumption:
            fieldname: consumption
            name: consumption
            resolution: 0.88
            seed: 1
            type: RandomDistributedScalarEncoder
          timestamp_timeOfDay:
            fieldname: timestamp
            name: timestamp_timeOfDay
            timeOfDay: [21, 1]
            type: DateEncoder
          timestamp_weekend:
            fieldname: timestamp
            name: timestamp_weekend
            type: DateEncoder
            weekend: 21
        sensorAutoReset: null
    
      spEnable: true
    
      spParams:
        inputWidth: 946
        columnCount: 2048
        spVerbosity: 0
        spatialImp: cpp
        globalInhibition: 1
        localAreaDensity: -1.0
        numActiveColumnsPerInhArea: 40
        seed: 1956
        potentialPct: 0.85
        synPermConnected: 0.1
        synPermActiveInc: 0.04
        synPermInactiveDec: 0.005
        boostStrength: 3.0
    
      tmEnable: true
    
      tmParams:
        verbosity: 0
        columnCount: 2048
        cellsPerColumn: 32
        inputWidth: 2048
        seed: 1960
        temporalImp: cpp
        newSynapseCount: 20
        initialPerm: 0.21
        permanenceInc: 0.1
        permanenceDec: 0.1
        maxAge: 0
        globalDecay: 0.0
        maxSynapsesPerSegment: 32
        maxSegmentsPerCell: 128
        minThreshold: 12
        activationThreshold: 16
        outputType: normal
        pamLength: 1
    
      clParams:
        verbosity: 0
        regionName: SDRClassifierRegion
        alpha: 0.1
        steps: '1,5'
        maxCategoryCount: 1000
        implementation: cpp
    
      trainSPNetOnlyIfRequested: false

### 添加一个 csv 文件 ###

需要注意：如果不修改上面的代码，则需要在该py文件的上级目录下新建文件夹，起名为 `data` ，并且在此 data目录下新建一个csv文件，起名为 `gymdata.csv` ，填写内容如下：

timestamp,consumption
    datetime,float
    T,
    7/2/10 0:00,21.2
    7/2/10 1:00,16.4
    7/2/10 2:00,4.7
    7/2/10 3:00,4.7
    7/2/10 4:00,4.6
    7/2/10 5:00,23.5
    7/2/10 6:00,47.5
    7/2/10 7:00,45.4
    7/2/10 8:00,46.1
    7/2/10 9:00,41.5
    7/2/10 10:00,43.4
    7/2/10 11:00,43.8
    7/2/10 12:00,37.8
    7/2/10 13:00,36.6
    7/2/10 14:00,35.7
    7/2/10 15:00,38.9
    7/2/10 16:00,36.2

### 运行 ###

在该例中py文件所在目录下输入命令如下(根据实际文件名更改)：

$ sudo python test.py

### 效果展示 ###

运行后可以看到控制台输出内容如下：

1-step:             21.2 (100.0%)	 5-step:             21.2 (100.0%)
    1-step:             16.4 (99.8%)	 5-step:             16.4 (99.8%)
    1-step:              4.7 (99.6%)	 5-step:              4.7 (99.6%)
    1-step:              4.7 (99.6%)	 5-step:              4.7 (99.6%)
    1-step:              4.6 (99.4%)	 5-step:              4.6 (99.4%)
    1-step:             23.5 (99.4%)	 5-step:             23.5 (99.4%)
    1-step:             47.5 (99.21%)	 5-step:             47.5 (99.21%)
    1-step:             45.4 (99.06%)	 5-step:             45.4 (99.06%)
    1-step:             46.1 (98.87%)	 5-step:             46.1 (98.87%)
    1-step:             41.5 (98.87%)	 5-step:             41.5 (98.87%)
    1-step:             43.4 (98.68%)	 5-step:             43.4 (98.68%)
    1-step:             43.8 (33.89%)	 5-step:             43.8 (90.87%)
    1-step:             43.8 (46.41%)	 5-step:             37.8 (87.24%)
    1-step:             36.6 (78.89%)	 5-step:            45.61 (46.18%)
    1-step:            45.61 (40.95%)	 5-step:             47.5 (46.79%)
    1-step:            45.61 (87.64%)	 5-step:             43.8 (95.56%)
    1-step:             38.9 (50.13%)	 5-step:             43.8 (16.84%)

### 常见错误与解决方法 ###

1.  保证 python 版本为 2.x
2.  如果提示错误如下：
    
        Traceback (most recent call last):
          File "test.py", line 4, in <module>
            import yaml
        ImportError: No module named yaml
    
    可能是没有安装这个yaml包，但是更有可能是需要添加sudo ，比如说输入 `python test.py` 提示以上错误，可以考虑使用 `sudo python test.py` 。

### 略加修改与测试 ###

读python源码，发现只输出step1 和step5，所以我想输出step3。

1.  修改 `params/model.yaml` 文件，查找其中的 1,5的位置，改成1,3,5
    
    修改后相应内容如下：
    
        clParams:
            verbosity: 0
            regionName: SDRClassifierRegion
            alpha: 0.1
            steps: '1,3,5'
            maxCategoryCount: 1000
            implementation: cpp
2.  修改python代码，找到oneStep 与 fiveStep，类似地，定义一个变量threeStep，基本过程也是一样，需要修改数组中下标，输出也是需要进行类似修改。总而言之代码如下：
    
        oneStep = bestPredictions[1]
              oneStepConfidence = allPredictions[1][oneStep]
              treeStep = bestPredictions[3]
              treeStepConfidence = allPredictions[3][treeStep]
              fiveStep = bestPredictions[5]
              fiveStepConfidence = allPredictions[5][fiveStep]
              result = (oneStep, oneStepConfidence * 100,
                        treeStep, treeStepConfidence * 100,
                        fiveStep, fiveStepConfidence * 100)
              print "1-step: {:16} ({:4.4}%)\t 3-step: {:16} ({:4.4}%)\t 5-step: {:16} ({:4.4}%)".format(*result)
    
    测试效果如下：
    
    ![在这里插入图片描述][watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9zbWlsZXlhbi5ibG9nLmNzZG4ubmV0_size_16_color_FFFFFF_t_70]
    
    分析部分将会记录在另一篇博客。

> Smileyan
> 
> 2019年11月8日 20:58

[Link 1]: https://blog.csdn.net/smileyan9/article/details/102919366
[http_nupic.docs.numenta.org_1.0.5_quick-start_opf.html]: http://nupic.docs.numenta.org/1.0.5/quick-start/opf.html
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9zbWlsZXlhbi5ibG9nLmNzZG4ubmV0_size_16_color_FFFFFF_t_70]: https://img-blog.csdnimg.cn/20191109115030882.png?x-oss-process=image/watermark,type_ZmFuZ3poZW5naGVpdGk,shadow_10,text_aHR0cHM6Ly9zbWlsZXlhbi5ibG9nLmNzZG4ubmV0,size_16,color_FFFFFF,t_70