{ "cells": [ { "cell_type": "code", "execution_count": 1, "metadata": { "collapsed": true }, "outputs": [], "source": [ "from urllib import urlretrieve\n", "import csv\n", "import datetime\n", "from io import BytesIO\n", "import numpy as np\n", "import json\n", "import copy\n", "import random\n", "\n", "import uuid\n", "from IPython.display import HTML\n", "\n", "from nupic.encoders import MultiEncoder\n", "import nupic.research.TP" ] }, { "cell_type": "code", "execution_count": 2, "metadata": { "collapsed": false }, "outputs": [ { "data": { "application/javascript": [ "require.config({\n", " paths: {\n", " d3: '//d3js.org/d3.v3.min',\n", " hello: \"//mrcslws.com/stuff/segment-stories.2016.04.28\"\n", " }\n", "});" ], "text/plain": [ "" ] }, "metadata": {}, "output_type": "display_data" } ], "source": [ "%%javascript\n", "require.config({\n", " paths: {\n", " d3: '//d3js.org/d3.v3.min',\n", " hello: \"//mrcslws.com/stuff/segment-stories.2016.04.28\"\n", " }\n", "});" ] }, { "cell_type": "code", "execution_count": 3, "metadata": { "collapsed": true }, "outputs": [], "source": [ "def drawEverything(columnStatesChartBuilder, segmentLifetimesChartBuilder):\n", " elementId = str(uuid.uuid1())\n", " addChart = \"\"\"\n", "
\n", " \n", " \"\"\" % (elementId, elementId,\n", " columnStatesChartBuilder.output.getvalue().replace('\\r', '\\\\r').replace('\\n', '\\\\n'),\n", " segmentLifetimesChartBuilder.getOutput().replace('\\r', '\\\\r').replace('\\n', '\\\\n'))\n", " \n", " return HTML(addChart)" ] }, { "cell_type": "code", "execution_count": 4, "metadata": { "collapsed": true }, "outputs": [], "source": [ "def parseHotgym(hotgymRaw):\n", " parsed = []\n", "\n", " csvReader = csv.reader(hotgymRaw)\n", " csvReader.next()\n", " csvReader.next()\n", " csvReader.next()\n", " \n", " for row in csvReader:\n", " timestampStr, consumptionStr = row\n", " parsed.append({\n", " 'timestamp': datetime.datetime.strptime(timestampStr,\n", " \"%m/%d/%y %H:%M\"),\n", " 'consumption': float(consumptionStr)\n", " })\n", " \n", " return parsed\n", "\n", "def encodeHotgym(hotgym):\n", " encoder = MultiEncoder({\n", " 'consumption': {\n", " 'fieldname': 'consumption',\n", " 'name': 'consumption',\n", " 'type': 'ScalarEncoder',\n", " 'minval': 5.0,\n", " 'maxval': 55.0,\n", " 'clipInput': True,\n", " 'n': 478,\n", " 'w': 31,\n", " },\n", " 'timestamp_timeOfDay': {\n", " 'fieldname': 'timestamp',\n", " 'name': 'timestamp_timeOfDay',\n", " 'timeOfDay': (21, 1),\n", " 'type': 'DateEncoder'\n", " },\n", " 'timestamp_weekend': {\n", " 'fieldname': 'timestamp',\n", " 'name': 'timestamp_weekend',\n", " 'type': 'DateEncoder',\n", " 'weekend': 21\n", " },\n", " })\n", " \n", " return map(lambda row : encoder.encode(row), hotgym)" ] }, { "cell_type": "code", "execution_count": 5, "metadata": { "collapsed": true }, "outputs": [], "source": [ "with open(urlretrieve(\"http://mrcslws.com/stuff/rec-center-hourly.csv\")[0],\n", " 'r') as hotgym:\n", " HOTGYM_ENCODED_ALL = encodeHotgym(parseHotgym(hotgym))\n", " HOTGYM_INDICES_ALL = map(lambda encoding : set(encoding.nonzero()[0]),\n", " HOTGYM_ENCODED_ALL)\n", " \n", " HOTGYM_ENCODED = HOTGYM_ENCODED_ALL[:1000]\n", " HOTGYM_INDICES = HOTGYM_INDICES_ALL[:1000]" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "Here I've tuned the encoders to produce sparse outputs and to use every bit." ] }, { "cell_type": "code", "execution_count": 6, "metadata": { "collapsed": false }, "outputs": [ { "data": { "text/plain": [ "73" ] }, "execution_count": 6, "metadata": {}, "output_type": "execute_result" } ], "source": [ "len(HOTGYM_ENCODED[0].nonzero()[0])" ] }, { "cell_type": "code", "execution_count": 7, "metadata": { "collapsed": false }, "outputs": [ { "data": { "text/plain": [ "0.0712890625" ] }, "execution_count": 7, "metadata": {}, "output_type": "execute_result" } ], "source": [ "# Sparsity\n", "len(HOTGYM_ENCODED[0].nonzero()[0]) / float(len(HOTGYM_ENCODED[0]))" ] }, { "cell_type": "code", "execution_count": 8, "metadata": { "collapsed": false }, "outputs": [ { "data": { "text/plain": [ "1024" ] }, "execution_count": 8, "metadata": {}, "output_type": "execute_result" } ], "source": [ "# Number of columns\n", "len(HOTGYM_ENCODED[0])" ] }, { "cell_type": "code", "execution_count": 9, "metadata": { "collapsed": false }, "outputs": [ { "data": { "text/plain": [ "array([140, 157, 186, 208, 225, 228, 228, 228, 229, 230, 231, 231, 231,\n", " 231, 232, 232, 233, 234, 235, 235, 235, 235, 236, 236, 237, 238,\n", " 239, 239, 239, 240, 242, 102, 86, 57, 35, 18, 15, 16, 20,\n", " 21, 21, 22, 24, 24, 25, 24, 24, 23, 23, 22, 22, 25,\n", " 25, 24, 24, 24, 25, 24, 25, 26, 26, 25, 26, 26, 26,\n", " 26, 26, 26, 26, 23, 22, 22, 21, 19, 20, 19, 22, 23,\n", " 24, 24, 25, 29, 27, 27, 28, 28, 28, 27, 28, 30, 29,\n", " 29, 28, 28, 28, 30, 33, 35, 36, 37, 39, 39, 39, 38,\n", " 38, 39, 41, 39, 40, 39, 39, 38, 35, 37, 39, 40, 40,\n", " 39, 40, 39, 36, 36, 38, 40, 40, 39, 37, 35, 33, 32,\n", " 32, 30, 30, 31, 32, 33, 31, 30, 31, 33, 33, 33, 34,\n", " 34, 32, 33, 32, 36, 39, 41, 41, 42, 47, 46, 53, 56,\n", " 63, 66, 66, 67, 71, 69, 68, 67, 67, 69, 68, 69, 68,\n", " 66, 62, 62, 62, 61, 61, 60, 57, 56, 52, 50, 46, 47,\n", " 47, 42, 40, 31, 27, 20, 17, 17, 17, 13, 13, 13, 13,\n", " 11, 8, 8, 7, 7, 7, 7, 7, 6, 6, 5, 6, 6,\n", " 6, 7, 6, 7, 6, 5, 6, 7, 7, 7, 7, 7, 6,\n", " 5, 6, 6, 6, 8, 9, 9, 9, 11, 11, 11, 11, 11,\n", " 12, 12, 15, 15, 15, 15, 17, 18, 17, 18, 18, 17, 17,\n", " 17, 17, 17, 17, 17, 17, 16, 18, 18, 16, 15, 15, 15,\n", " 13, 13, 13, 14, 14, 14, 14, 12, 11, 11, 12, 10, 9,\n", " 10, 10, 13, 14, 13, 13, 15, 17, 18, 23, 28, 32, 35,\n", " 39, 44, 53, 56, 62, 74, 76, 79, 83, 87, 92, 101, 105,\n", " 117, 125, 133, 138, 145, 155, 162, 162, 165, 174, 182, 188, 198,\n", " 204, 207, 212, 215, 218, 216, 218, 217, 220, 220, 217, 219, 222,\n", " 229, 231, 232, 228, 226, 217, 217, 210, 216, 212, 203, 196, 198,\n", " 198, 195, 191, 185, 183, 179, 177, 174, 173, 169, 173, 169, 169,\n", " 166, 163, 156, 156, 155, 148, 142, 139, 138, 138, 139, 135, 134,\n", " 124, 126, 128, 132, 128, 126, 120, 120, 121, 113, 114, 111, 106,\n", " 102, 99, 94, 94, 88, 85, 82, 82, 78, 74, 71, 72, 69,\n", " 66, 64, 62, 58, 57, 56, 52, 49, 46, 45, 43, 44, 41,\n", " 39, 37, 35, 32, 31, 29, 31, 30, 27, 25, 25, 25, 23,\n", " 23, 22, 20, 19, 19, 18, 17, 15, 15, 15, 14, 13, 12,\n", " 10, 10, 10, 9, 8, 7, 7, 5, 5, 4, 4, 1, 1,\n", " 1, 1, 1, 1, 1, 3, 3, 3, 3, 2, 2, 2, 2,\n", " 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,\n", " 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 42, 42, 42,\n", " 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42,\n", " 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42,\n", " 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42,\n", " 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42,\n", " 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42,\n", " 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42,\n", " 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42,\n", " 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42,\n", " 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42,\n", " 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42,\n", " 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42,\n", " 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42,\n", " 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42,\n", " 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42,\n", " 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42,\n", " 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42,\n", " 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42,\n", " 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42,\n", " 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42,\n", " 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42,\n", " 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42,\n", " 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42,\n", " 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42,\n", " 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42,\n", " 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 41, 41,\n", " 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41,\n", " 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41,\n", " 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41,\n", " 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41,\n", " 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41,\n", " 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41,\n", " 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41,\n", " 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41,\n", " 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41,\n", " 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41,\n", " 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41,\n", " 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41,\n", " 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 42, 42, 42,\n", " 42, 42, 42, 42, 42, 42, 42, 682, 682, 682, 682, 682, 682,\n", " 682, 682, 682, 682, 682, 682, 682, 682, 682, 682, 682, 682, 682,\n", " 682, 682, 318, 318, 318, 318, 318, 318, 318, 318, 318, 318, 318,\n", " 318, 318, 318, 318, 318, 318, 318, 318, 318, 318], dtype=uint32)" ] }, "execution_count": 9, "metadata": {}, "output_type": "execute_result" } ], "source": [ "# Show how often each bit is used\n", "np.set_printoptions(threshold=1024)\n", "reduce(lambda accum, encoding: accum + encoding,\n", " HOTGYM_ENCODED,\n", " np.zeros(1024, dtype='uint32'))" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "# Experiment: TP.py" ] }, { "cell_type": "code", "execution_count": 10, "metadata": { "collapsed": true }, "outputs": [], "source": [ "class ColumnStatesPatcher(object):\n", " def __init__(self):\n", " self.output = BytesIO()\n", "\n", " def patchTP(self, tp):\n", " csvOutput = csv.writer(self.output)\n", "\n", " headerRow = [\n", " 'n-unpredicted-active-columns',\n", " 'n-predicted-inactive-columns',\n", " 'n-predicted-active-columns',\n", " ]\n", " csvOutput.writerow(headerRow)\n", "\n", " computeMethod = tp.compute\n", " def myCompute(bottomUpInput, **kwargs):\n", " activeColumns = set(bottomUpInput.nonzero()[0])\n", "\n", " npPredictedCells = tp.getPredictedState().reshape(-1).nonzero()[0]\n", " predictedColumns = set([cell / tp.cellsPerColumn for cell in npPredictedCells.tolist()])\n", "\n", " computeResult = computeMethod(bottomUpInput, **kwargs)\n", "\n", " row = (\n", " len(activeColumns - predictedColumns),\n", " len(predictedColumns - activeColumns),\n", " len(activeColumns & predictedColumns),\n", " )\n", " csvOutput.writerow(row)\n", "\n", " return computeResult\n", "\n", " tp.compute = myCompute" ] }, { "cell_type": "code", "execution_count": 11, "metadata": { "collapsed": true }, "outputs": [], "source": [ "class SegmentLifetimeChartBuilderTP(object):\n", " def __init__(self):\n", " # segID => {}\n", " self.segments = {}\n", " # (segID, {}) because a segment might get destroyed multiple times\n", " self.destroyedSegments = []\n", " self.timestep = 0\n", " \n", " self.prevActiveSegments = []\n", " self.activeSegments = []\n", " \n", " def getOutput(self):\n", " outputSegments = []\n", " for k, data in self.destroyedSegments:\n", " out = {\n", " 'birthstep': data['birthstep'],\n", " 'deathstep': data['deathstep'],\n", " }\n", " \n", " if len(data['correctMatches']) > 0:\n", " out['correctMatches'] = data['correctMatches']\n", " \n", " if len(data['correctActivations']) > 0:\n", " out['correctActivations'] = data['correctActivations']\n", " \n", " if len(data['incorrectActivations']) > 0:\n", " out['incorrectActivations'] = data['incorrectActivations']\n", " \n", " outputSegments.append(out)\n", " for k, data in self.segments.items():\n", " out = {\n", " 'birthstep': data['birthstep'],\n", " }\n", " \n", " if len(data['correctMatches']) > 0:\n", " out['correctMatches'] = data['correctMatches']\n", " \n", " if len(data['correctActivations']) > 0:\n", " out['correctActivations'] = data['correctActivations']\n", " \n", " if len(data['incorrectActivations']) > 0:\n", " out['incorrectActivations'] = data['incorrectActivations']\n", " \n", " outputSegments.append(out)\n", " outputSegments = sorted(outputSegments, key = lambda x : x['birthstep'])\n", " output = {\n", " 'nTimesteps': self.timestep,\n", " 'segments': outputSegments\n", " }\n", " return json.dumps(output)\n", " \n", " def patchTP(self, tp):\n", " \n", " def onCorrectMatchingSegment(segment):\n", " k = segment.segID\n", " segmentData = None\n", " if k in self.segments:\n", " segmentData = self.segments[k]\n", " else:\n", " for k2, data in reversed(self.destroyedSegments):\n", " if k2 == k:\n", " segmentData = data\n", " break\n", "\n", " assert segmentData is not None\n", " \n", " segmentData['correctMatches'].append(self.timestep)\n", " tp.onCorrectMatchingSegment = onCorrectMatchingSegment\n", "\n", " def onActiveSegment(segment):\n", " self.activeSegments.append(segment.segID)\n", " tp.onActiveSegment = onActiveSegment\n", "\n", " def onDestroySegment(segment):\n", " k = segment.segID\n", "\n", " data = self.segments[k]\n", " del self.segments[k]\n", "\n", " data['deathstep'] = self.timestep\n", "\n", " v = (k, data)\n", " self.destroyedSegments.append(v)\n", " tp.onDestroySegment = onDestroySegment\n", " \n", " def onCreateSegment(c, i, segment):\n", " k = segment.segID\n", " assert k not in self.segments\n", " \n", " cell = c*tp.cellsPerColumn + i\n", "\n", " self.segments[k] = {\n", " 'cell': cell,\n", " 'birthstep': self.timestep,\n", " 'correctMatches': [],\n", " 'correctActivations': [],\n", " 'incorrectActivations': []\n", " }\n", " tp.onCreateSegment = onCreateSegment\n", " \n", " compute = tp.compute\n", " def myCompute(bottomUpInput, **kwargs):\n", " self.beforeCompute(tp)\n", " compute(bottomUpInput, **kwargs)\n", " self.afterCompute(tp)\n", " tp.compute = myCompute\n", "\n", " \n", " def beforeCompute(self, tp):\n", " self.prevActiveSegments = self.activeSegments\n", " self.activeSegments = []\n", " \n", " def afterCompute(self, tp):\n", " for k in self.prevActiveSegments:\n", " segmentData = None\n", " if k in self.segments:\n", " segmentData = self.segments[k]\n", " else:\n", " for k2, data in reversed(self.destroyedSegments):\n", " if k2 == k:\n", " segmentData = data\n", " break\n", "\n", " assert segmentData is not None\n", "\n", " activeCells = set(tp.getActiveState().nonzero()[0].tolist())\n", " if segmentData['cell'] in activeCells:\n", " segmentData['correctActivations'].append(self.timestep)\n", " else:\n", " segmentData['incorrectActivations'].append(self.timestep)\n", "\n", " self.timestep += 1" ] }, { "cell_type": "code", "execution_count": 12, "metadata": { "collapsed": false }, "outputs": [], "source": [ "experiment1_column_states = ColumnStatesPatcher()\n", "experiment1_segment_lifetimes = SegmentLifetimeChartBuilderTP()\n", "\n", "def experiment1(tp=None):\n", " if tp is None:\n", " tp = nupic.research.TP.TP(\n", " numberOfCols=1024,\n", " cellsPerColumn=4,\n", " initialPerm=0.21,\n", " connectedPerm=0.50,\n", " minThreshold=10,\n", " newSynapseCount=20,\n", " permanenceInc=0.10,\n", " permanenceDec=0.10,\n", " globalDecay=0.0,\n", " maxAge=0,\n", " activationThreshold=13,\n", " maxSegmentsPerCell=4,\n", " maxSynapsesPerSegment=255,\n", " maxInfBacktrack=0,\n", " maxLrnBacktrack=0,\n", " maxSeqLength=0,\n", " seed=42)\n", "\n", " experiment1_column_states.patchTP(tp)\n", " experiment1_segment_lifetimes.patchTP(tp)\n", "\n", " i = 0\n", " for encoding in HOTGYM_ENCODED:\n", " i += 1\n", " if i % 100 == 0:\n", " print \"timestep %d\" % i\n", " tp.compute(encoding, enableLearn=True, computeInfOutput=True)\n", " \n", " return tp" ] }, { "cell_type": "code", "execution_count": 13, "metadata": { "collapsed": false }, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "timestep 100\n", "timestep 200\n", "timestep 300\n", "timestep 400\n", "timestep 500\n", "timestep 600\n", "timestep 700\n", "timestep 800\n", "timestep 900\n", "timestep 1000\n" ] } ], "source": [ "e1_tp = experiment1()" ] }, { "cell_type": "code", "execution_count": 14, "metadata": { "collapsed": false }, "outputs": [ { "data": { "text/html": [ "\n", "
\n", " \n", " " ], "text/plain": [ "" ] }, "execution_count": 14, "metadata": {}, "output_type": "execute_result" } ], "source": [ "drawEverything(experiment1_column_states, experiment1_segment_lifetimes)" ] }, { "cell_type": "code", "execution_count": 15, "metadata": { "collapsed": false }, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "timestep 0\n", "timestep 100\n", "timestep 200\n", "timestep 300\n", "timestep 400\n", "timestep 500\n", "timestep 600\n", "timestep 700\n", "timestep 800\n", "timestep 900\n" ] } ], "source": [ "for i in xrange(1000):\n", " if i % 100 == 0:\n", " print \"timestep %d\" % i\n", "\n", " activeColumns = random.sample(xrange(1024), 73)\n", " encoding = np.zeros(1024, dtype='uint32')\n", " encoding[activeColumns] = 1\n", " e1_tp.compute(encoding, enableLearn=True, computeInfOutput=True)" ] }, { "cell_type": "code", "execution_count": 16, "metadata": { "collapsed": false }, "outputs": [ { "data": { "text/html": [ "\n", "
\n", " \n", " " ], "text/plain": [ "" ] }, "execution_count": 16, "metadata": {}, "output_type": "execute_result" } ], "source": [ "drawEverything(experiment1_column_states, experiment1_segment_lifetimes)" ] }, { "cell_type": "code", "execution_count": 17, "metadata": { "collapsed": false }, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "timestep 100\n", "timestep 200\n", "timestep 300\n", "timestep 400\n", "timestep 500\n", "timestep 600\n", "timestep 700\n", "timestep 800\n", "timestep 900\n", "timestep 1000\n" ] }, { "data": { "text/plain": [ "" ] }, "execution_count": 17, "metadata": {}, "output_type": "execute_result" } ], "source": [ "experiment1(e1_tp)" ] }, { "cell_type": "code", "execution_count": 18, "metadata": { "collapsed": false }, "outputs": [ { "data": { "text/html": [ "\n", "
\n", " \n", " " ], "text/plain": [ "" ] }, "execution_count": 18, "metadata": {}, "output_type": "execute_result" } ], "source": [ "drawEverything(experiment1_column_states, experiment1_segment_lifetimes)" ] }, { "cell_type": "code", "execution_count": 19, "metadata": { "collapsed": false }, "outputs": [], "source": [ "with open('2016.05.03.experiment1_column_states.csv', 'w') as f:\n", " f.write(experiment1_column_states.output.getvalue())" ] }, { "cell_type": "code", "execution_count": 20, "metadata": { "collapsed": true }, "outputs": [], "source": [ "with open('2016.05.03.experiment1_segment_lifetimes.json', 'w') as f:\n", " f.write(experiment1_segment_lifetimes.getOutput())" ] }, { "cell_type": "code", "execution_count": 6, "metadata": { "collapsed": true }, "outputs": [], "source": [ "tp = nupic.research.TP.TP(\n", " maxInfBacktrack=0,\n", " maxLrnBacktrack=0,\n", " maxSeqLength=0,\n", " numberOfCols=1024,\n", " cellsPerColumn=4,\n", " initialPerm=0.21,\n", " connectedPerm=0.50,\n", " minThreshold=10,\n", " newSynapseCount=20,\n", " permanenceInc=0.10,\n", " permanenceDec=0.10,\n", " globalDecay=0.0,\n", " maxAge=0,\n", " activationThreshold=13,\n", " maxSegmentsPerCell=4,\n", " maxSynapsesPerSegment=255,\n", " seed=42)" ] } ], "metadata": { "kernelspec": { "display_name": "Python 2", "language": "python", "name": "python2" }, "language_info": { "codemirror_mode": { "name": "ipython", "version": 2 }, "file_extension": ".py", "mimetype": "text/x-python", "name": "python", "nbconvert_exporter": "python", "pygments_lexer": "ipython2", "version": "2.7.10" } }, "nbformat": 4, "nbformat_minor": 0 }