- Project
- About Data
- TodoList
- Todo
- machine learning model
- read all triples from neo4j
- get most connected subgraph
- write all data to database? text file
- [[#cancel-large-graph-train][[CANCEL] large graph train]]
- Export to rust
- machine learning model
- Add word embedding
- Train data and label
- Serving
- Usage
cargo run --release --example inject_dataTrunk supports an optional Trunk.toml config file. An example config file is included in the Trunk repo, and shows all available config options along with their default values. By default, Trunk will look for a Trunk.toml config file in the current working directory. Trunk supports the global –config option to specify an alternative location for the file.
Note that any relative paths declared in a Trunk.toml file will be treated as being relative to the Trunk.toml file itself.
# An example Trunk.toml with all possible fields along with their defaults.
[build]
# The index HTML file to drive the bundling process.
target = "index.html"
# Build in release mode.
release = false
# The output dir for all final assets.
dist = "dist"
# The public URL from which assets are to be served.
public_url = "/"
[watch]
# Paths to watch. The `build.target`'s parent folder is watched by default.
watch = []
# Paths to ignore.
ignore = []
[serve]
# The port to serve on.
port = 8080
# Open a browser tab once the initial build is complete.
open = false
# Disable auto-reload of the web app.
no_autoreload = false
[clean]
# The output dir for all final assets.
dist = "dist"
# Optionally perform a cargo clean.
cargo = false
[tools]
# Default wasm-bindgen version to download.
wasm_bindgen = "0.2.74"
# Default wasm-opt version to download.
wasm_opt = "version_101"
## proxy
# Proxies are optional, and default to `None`.
# Proxies are only run as part of the `trunk serve` command.
[[proxy]]
# This WebSocket proxy example has a backend and ws field. This example will listen for
# WebSocket connections at `/api/ws` and proxy them to `ws://localhost:9000/api/ws`.
backend = "ws://localhost:9000/api/ws"
ws = true
[[proxy]]
# This proxy example has a backend and a rewrite field. Requests received on `rewrite` will be
# proxied to the backend after rewriting the `rewrite` prefix to the `backend`'s URI prefix.
# E.G., `/api/v1/resource/x/y/z` -> `/resource/x/y/z`
rewrite = "/api/v1/"
backend = "http://localhost:9000/"
[[proxy]]
# This proxy specifies only the backend, which is the only required field. In this example,
# request URIs are not modified when proxied.
backend = "http://localhost:9000/api/v2/"
all build data go into dist directory
include script: copy-file then foo.js
- network := https://www.d3-graph-gallery.com/network
- network graph slides := http://elijahmeeks.com/networkviz/
- D3 in Depth
var nodeEnter = d3.select("svg").selectAll("g.node")
.data(nodes)
.enter()
.append("g")
.attr("class", "node")
.call(force.drag());
nodeEnter.append("circle")
.attr("r", 5)
.style("fill", "#CC9999");
nodeEnter.append("text")
.style("text-anchor", "middle")
.attr("y", 15)
.text(function (d) {return d.id})
function updateNetwork() {
d3.select("svg").selectAll("g.node")
.attr("transform", function (d) {return "translate(" + d.x + "," + d.y + ")"});var nodeEnter = d3.select("svg").selectAll("g.node")
.data(nodes)
.enter()
.append("g")
.attr("class", "node")
.on("click", nodeClick)
.on("dblclick", nodeDoubleClick)
.on("mouseover", nodeOver)
.on("mouseout", nodeOut)
.call(force.drag());
function nodeClick(d) {
d.fixed = true;
}
function nodeDoubleClick(d) {
d.fixed = false;
force.start();
}
function nodeOver() {
force.stop();
}
function nodeOut() {
force.start();
}全部数据 隐藏展开模式?
function nodeOver(d) {
force.stop();
highlightEgoNetwork(d);
}
function nodeOut() {
force.start();
d3.selectAll("g.node > circle")
.style("fill", "#CC9999");
d3.selectAll("line")
.style("stroke", "#996666")
.style("stroke-width", "1px");
}
function highlightEgoNetwork(d) {
var egoIDs = [];
var filteredEdges = edges.filter(function (p) {return p.source == d || p.target == d});
filteredEdges
.forEach(function (p) {
if (p.source == d) {
egoIDs.push(p.target.id)
}
else {
egoIDs.push(p.source.id)
}
});
d3.selectAll("line")
.filter(function (p) {return filteredEdges.indexOf(p) > -1})
.style("stroke", "#66CCCC")
.style("stroke-width", "2px");
d3.selectAll("circle")
.filter(function (p) {return egoIDs.indexOf(p.id) > -1})
.style("fill", "#66CCCC");
}
function updateNetwork() {
d3.select("svg").selectAll("g.node > circle")
.attr("r", function (d) {return d.weight});edgeEnter
.append("line")
.attr("class", "highlight")
.style("stroke-width", "8px")
.style("stroke", "#66CCCC")
.style("opacity", 0)
.on("dblclick", deleteEdge)
.on("mouseover", edgeOver)
.on("mouseout", edgeOut);
var nodeEnter = d3.select("svg").selectAll("g.node")
.data(nodes, function (d) {return d.id})
.enter()
.append("g")
.attr("class", "node")
.on("dblclick", deleteNode)
.on("mouseover", nodeOver)
.on("mouseout", nodeOut)
.call(force.drag());
function deleteNode(d) {
var currentNodes = d3.selectAll("g.node").data();
var currentEdges = d3.selectAll("g.edge").data();
var filteredNodes = currentNodes.filter(function (p) {return p !== d});
var filteredEdges = currentEdges.filter(function (p) {return p.source !== d && p.target !== d});
force.stop();
force.nodes(filteredNodes);
force.links(filteredEdges);
d3.selectAll("g.node").data(filteredNodes, function (d) {return d.id})
.exit()
.transition()
.duration(500)
.style("opacity", 0)
.remove();
d3.selectAll("g.edge").data(filteredEdges, function (d) {return d.id})
.exit()
.transition()
.duration(500)
.style("opacity", 0)
.remove();
force.start();
}
function deleteEdge(d) {
var currentEdges = d3.selectAll("g.edge").data();
var filteredEdges = currentEdges.filter(function (p) {return p !== d});
force.stop();
force.links(filteredEdges);
d3.selectAll("g.edge").data(filteredEdges, function (d) {return d.id})
.exit()
.transition()
.duration(500)
.style("opacity", 0)
.remove();
force.start();
}
function nodeOver() {
d3.select(this).select("circle")
.style("fill", "#66CCCC")
.style("stroke", "#66CCCC")
.style("stroke-width", "3px");
}
function nodeOut() {
d3.selectAll("circle")
.style("fill", "#CC9999")
.style("stroke", "black")
.style("stroke-width", "1px");
}
function edgeOver(d) {
d3.select(this).style("opacity", 0.75);
}
function edgeOut() {
d3.selectAll("line.highlight").style("opacity", 0);
}var modules = nodes.map(function (d) {return d.module})
.filter(onlyUnique)
.map(function (d) {return {id: d, members: []}});
var moduleEdges = [];
var singletons = {id: "singletons", members: []};
var moduleNodeHash = {};
modules.forEach(function (module) {
module.members = nodes.filter(function (d) {return d.module === module.id});
moduleNodeHash[module.id] = module;
if (module.members.length === 1) {
singletons.members.push(module.members[0]);
}
});
modules.push(singletons);
var moduleEdgeHash = {};
edges.forEach(function (edge) {
if (!moduleEdgeHash[moduleNodeHash[edge.source.module].id + "-" + moduleNodeHash[edge.target.module].id]) {
var moduleEdge = {source: moduleNodeHash[edge.source.module], target: moduleNodeHash[edge.target.module], weight: 1};
moduleEdgeHash[moduleNodeHash[edge.source.module].id + "-" + moduleNodeHash[edge.target.module].id] = moduleEdge;
moduleEdges.push(moduleEdge);
}
else {
moduleEdgeHash[moduleNodeHash[edge.source.module].id + "-" + moduleNodeHash[edge.target.module].id].weight += 1;
}
})
return {nodes: modules, edges: moduleEdges}
var modularityGraph = modularityCensus(nodes, edges, result);
var modularityForce = d3.layout.force()
.nodes(modularityGraph.nodes)
.links(modularityGraph.edges)
.size([125,125])
.charge(function (d) {return d.members.length * -25})
.gravity(0.75)
.on("tick", updateModularityNetwork);
var mSVG = d3.select("#viz").append("svg")
.attr("class", "modularity")
.attr("height", 100)
.attr("width", 100)
.style("height", 100)
.style("width", 100)
.style("position", "absolute")
.style("left", 375)
.style("top", 375)
.style("background", "white");
mSVG.selectAll("line")
.data(modularityGraph.edges)
.enter()
.append("line")
.attr("class", "modularity")
.style("stroke-width", function (d) {return d.weight * 2})
.style("stroke", "black");
mSVG.selectAll("circle")
.data(modularityGraph.nodes.filter(function(d) {return d.members.length > 1}))
.enter()
.append("circle")
.attr("class", "modularity")
.attr("r", function (d) {return d.members.length})
.style("stroke", "black")
.style("stroke-width", "1px")
.style("fill", function (d) {return d.id == "singletons" ? "lightgray" : colors(d.id)})
modularityForce.start();
function updateModularityNetwork() {
d3.select("svg.modularity").selectAll("line")
.attr("x1", function (d) {return d.source.x})
.attr("y1", function (d) {return d.source.y})
.attr("x2", function (d) {return d.target.x})
.attr("y2", function (d) {return d.target.y});
d3.select("svg.modularity").selectAll("circle")
.attr("transform", function (d) {return "translate(" + d.x + "," + d.y + ")"});
}function moduleOver(d) {
d3.select(this)
.style("stroke-width", "4px")
d3.select("svg.main").selectAll("circle")
.style("stroke-width", function (p) {return p.module == d.id ? "4px" : "1px"})
}
function moduleOut(d) {
d3.select(this)
.style("stroke-width", "1px")
d3.select("svg.main").selectAll("circle")
.style("stroke-width", "1px")
}
function updateNetwork() {
d3.select("svg.modularity").selectAll("circle")
.each(function (d) {
var theseNodes = d.members;
var avgX = d3.mean(theseNodes, function (p) {return p.x});
var avgY = d3.mean(theseNodes, function (p) {return p.y});
d.x = avgX / 5;
d.y = avgY / 5;
})
.attr("transform", function (d) {return "translate(" + d.x + "," + d.y + ")"});
d3.select("svg.modularity").selectAll("line")
.attr("x1", function (d) {return d.source.x})
.attr("y1", function (d) {return d.source.y})
.attr("x2", function (d) {return d.target.x})
.attr("y2", function (d) {return d.target.y});var drag = force.drag()
.on("drag", drag)
.on("dragend", dragend)
.on("dragstart", dragstart);
var nodeEnter = d3.select("svg.main").selectAll("g.node")
.data(nodes, function (d) {return d.id})
.enter()
.append("g")
.attr("class", "node")
.call(drag);
function dragstart() {
nodes.forEach(function (node) {
node.fixed = true;
})
}
function drag(d) {
var nodeDom = this;
var foundOverlap = false
nodes.forEach(function (otherNode) {
var distance = Math.sqrt(Math.pow(otherNode.x - d.x, 2) + Math.pow(otherNode.y - d.y, 2));
if (otherNode != d && distance < 16) {
foundOverlap = true;
}
})
if (foundOverlap == true) {
d3.select(nodeDom).select("circle.background")
.style("opacity", 0.5)
}
else {
d3.select(nodeDom).select("circle.background")
.style("opacity", 0)
}
}
function dragend(d) {
force.stop();
d3.selectAll("circle.background")
.style("opacity", 0);
nodes.forEach(function (otherNode) {
otherNode.fixed = false;
var distance = Math.sqrt(Math.pow(otherNode.x - d.x, 2) + Math.pow(otherNode.y - d.y, 2));
if (otherNode != d && distance < 16) {
var newEdge = {id: d.id + "-" + otherNode.id, source: d, target: otherNode};
if (edges.map(function (d) {return d.id}).indexOf(newEdge.id) == -1) {
edges.push(newEdge);
}
}
});
force.links(edges);
var edgeEnter = d3.select("svg.main").selectAll("g.edge")
.data(edges, function (d) {return d.id})
.enter()
.insert("g", "g.node")
.attr("class", "edge");
edgeEnter
.append("line")
.style("stroke-width", function (d) {return d.border ? "3px" : "1px"})
.style("stroke", "black")
.style("pointer-events", "none");
var community = jLouvain().nodes(node_data).edges(edge_data);
var result = community();
nodes.forEach(function (node) {
node.module = result[node.id]
});
node_data = nodes.map(function (d) {return d.id});
edge_data = edges.map(function (d) {return {source: d.source.id, target: d.target.id, weight: 1}; });
modularityCensus(nodes, edges, result);
d3.selectAll("circle.foreground")
.style("fill", function (d) {return colors(d.module)})
.style("stroke-width", function (d) {return d.border ? "3px" : "1px"});
force.start();
}side effect!!! change data.nodes and data.links
"nodes":{
"id": 2,
"index": 1,
"name": "B",
"vx": 0,
"vy": 0,
"x": -9.03088751750192,
"y": 8.273032735715967
},
"links":{
"source": {
"id": 3,
"index": 2,
"name": "C",
"vx": 0,
"vy": 0,
"x": 1.3823220809823638,
"y": -15.750847141167634,
},
"target": {
"id": 4,
"index": 3,
"name": "D",
"vx": 0,
"vy": 0,
"x": 11.382848792909423,
"y": 14.846910566099618
}
}
https://www.d3-graph-gallery.com/graph/custom_legend.html
****
npm install --save @types/d3https://github.com/Lemoncode/d3js-typescript-examples
selection.data default bases on array index, use key fn to customize data key
swc src --out-dir build/src -w --sync & tsc -w --pretty --skipLibCheck --noEmitgroup other SVG elements svg tut
The <g>-element doesn’t have x and y attributes. To move the contents of a <g>-element you can only do so using the transform attribute, using the “translate” function, like this: transform=”translate(x,y)”.
sudo sysctl fs.inotify.max_user_watches=32768./node_modules/.bin/esbuild src/index.ts --bundle --outfile=out.js- [x] include bundle data
- [x] import all export js snippet
- [x] navbar
- [ ] mobile responsive
https://github.com/ecomfe/echarts-wordcloud/
2:#疾病 rdf-schema#subClassOf owl#Thing
4:#症状 rdf-schema#subClassOf owl#Thing
6:#科室 rdf-schema#subClassOf owl#Thing
8:#检查 rdf-schema#subClassOf owl#Thing
10:#药品 rdf-schema#subClassOf owl#Thing
12:#中药 rdf-schema#subClassOf #药品
14:#西药 rdf-schema#subClassOf #药品
16:#中医症状 rdf-schema#subClassOf #症状
18:#西医症状 rdf-schema#subClassOf #症状
245236:急性发热性嗜中性白细胞增多性皮肤病或隆起性红斑 owl#sameAs 斯威特氏综合征
245237:酞咪脉啶酮,沙利窦迈,撒利多胺 owl#sameAs 撒利多胺
245238:新达罗颗粒剂,头孢克洛颗粒剂 owl#sameAs 头孢克洛颗粒剂
245239:先天性红细胞生成不良性贫血,遗传性良性网织红细胞增生性贫血 owl#sameAs 先天性红细胞生成不良性贫血
245240:上气 owl#sameAs 气喘,喘逆,喘促
245241:野蜀葵 owl#sameAs 三叶
245242:妊娠试验 owl#sameAs 尿HCG
245243:老年人慢性肾衰竭 owl#sameAs 老年人慢性肾功能衰竭
245244:小儿主动脉弓综合征 owl#sameAs 小儿高安氏动脉炎
245245:晶 owl#sameAs 晶状体诱导的葡萄膜炎
245246:小儿房中隔缺损 owl#sameAs 小儿房缺
245247:斯欧蜜 owl#sameAs 火府丸
245248:疔 owl#sameAs 外阴疖
245249:海冻菜,红丝 owl#sameAs 红丝
245250:绿萼梅,绿梅花 owl#sameAs 白梅花
200746:雌二醇地屈孕酮片复合包装 rdf-schema#label 雌二醇地屈孕酮片复合包装"@Z
204680:肾微粒体抗体测定 rdf-schema#label 肾微粒体抗体测定"@Z
208670:雌二醇地屈孕酮片复合包装) rdf-schema#label 雌二醇地屈孕酮片复合包装)"@Z
209788:1.2型 rdf-schema#label 1.2型"@Z
210669:中毒性表皮坏死松解型药疹 rdf-schema#label 中毒性表皮坏死松解型药疹"@Z
213265:地塞米松眼膏) rdf-schema#label 地塞米松眼膏)"@Z
213735:S)型 rdf-schema#label S)型"@Z
214565:粒*100粒*5瓶 rdf-schema#label 粒*100粒*5瓶"@Z
214825:长链脂肪乳注射液(C8～24Ve) rdf-schema#label 长链脂肪乳注射液(C8～24Ve)"@Z
216010:氢氯噻嗪片 rdf-schema#label 氢氯噻嗪片"@Z
217294:克拉维酸钾片 rdf-schema#label 克拉维酸钾片"@Z
218617:盒包TYS rdf-schema#label 盒包TYS"@Z
221118:30(精蛋白锌重组人胰岛素混合注射液) rdf-schema#label 30(精蛋白锌重组人胰岛素混合注射液)"@Z
224908:粒)) rdf-schema#label 粒))"@Z
225809:三氯生乳膏 rdf-schema#label 三氯生乳膏"@Z
228247:地塞米松眼膏 rdf-schema#label 地塞米松眼膏"@Z
228794:甲氧苄啶 rdf-schema#label 甲氧苄啶"@Z
20:#症状相关症状 rdf-schema#domain #症状
23:#症状相关疾病 rdf-schema#domain #症状
26:#症状相关药品 rdf-schema#domain #症状
29:#症状相关科室 rdf-schema#domain #症状
32:#症状相关检查 rdf-schema#domain #症状
35:#症状相关部位 rdf-schema#domain #症状
38:#疾病相关症状 rdf-schema#domain #疾病
41:#疾病相关疾病 rdf-schema#domain #疾病
44:#疾病相关药品 rdf-schema#domain #疾病
47:#疾病相关科室 rdf-schema#domain #疾病
50:#疾病相关检查 rdf-schema#domain #疾病
53:#疾病相关部位 rdf-schema#domain #疾病
56:#检查相关症状 rdf-schema#domain #检查
59:#检查相关疾病 rdf-schema#domain #检查
62:#检查相关科室 rdf-schema#domain #检查
65:#检查相关检查 rdf-schema#domain #检查
68:#检查相关部位 rdf-schema#domain #检查
21:#症状相关症状 rdf-schema#range #症状
24:#症状相关疾病 rdf-schema#range #疾病
27:#症状相关药品 rdf-schema#range #药品
30:#症状相关科室 rdf-schema#range #科室
33:#症状相关检查 rdf-schema#range #检查
36:#症状相关部位 rdf-schema#range rdf-schema#Literal
39:#疾病相关症状 rdf-schema#range #症状
42:#疾病相关疾病 rdf-schema#range #疾病
45:#疾病相关药品 rdf-schema#range #药品
48:#疾病相关科室 rdf-schema#range #科室
51:#疾病相关检查 rdf-schema#range #检查
54:#疾病相关部位 rdf-schema#range rdf-schema#Literal
57:#检查相关症状 rdf-schema#range #症状
60:#检查相关疾病 rdf-schema#range #疾病
63:#检查相关科室 rdf-schema#range #科室
66:#检查相关检查 rdf-schema#range #检查
69:#检查相关部位 rdf-schema#range rdf-schema#Literal
27915 17 52146 { “检查相关症状”, “疾病相关疾病”, “疾病相关症状”, “检查相关部位”, “症状相关部位”, “症状相关症状”, “症状相关科室”, “症状相关疾病”, “症状相关检查”, “疾病相关检查”, “疾病相关部位”, “疾病相关药品”, “检查相关检查”, “检查相关疾病”, “疾病相关科室”, “症状相关药品”, “检查相关科室”, }
togs_head: 687 togs_tail: 1385
“检查相关症状”,=> 检查后遗症 “症状相关检查”, => 检查
区分性不大 后者质量更差 “疾病相关检查”, “检查相关疾病”,
症状 - 推测 -> 疾病 -> 科室
症状相关部位 症状相关症状 症状相关科室 症状相关疾病 症状相关检查 症状相关药品 all: 200043 unique: 10894
*
疾病相关疾病 疾病相关症状 疾病相关检查 疾病相关部位 疾病相关药品 疾病相关科室 all: 359438 unique: 14664
检查相关症状 检查相关部位 检查相关检查 检查相关疾病 检查相关科室 all: 45217 unique: 2361
检查相关部位 症状相关部位 疾病相关部位 all: 17274 unique: 111
疾病相关药品 症状相关药品 all: 109246 unique: 14277
症状相关科室 疾病相关科室 检查相关科室 all: 70040 unique: 291
CREATE SPACE [IF NOT EXISTS] <graph_space_name> [(partition_num = <partition_number>, replica_factor = <replica_number>, vid_type = {FIXED_STRING(<N>)) | INT64}];
CREATE {TAG | EDGE} {<tag_name> | <edge_type>}(<property_name> <data_type> [, <property_name> <data_type> …]);
INSERT VERTEX <tag_name> (<property_name>[, <property_name>…]) [, <tag_name> (<property_name>[, <property_name>…]), …] {VALUES | VALUE} <vid>: (<property_value>[, <property_value>…]) [, <vid>: (<property_value>[, <property_value>…];
INSERT EDGE <edge_type> (<property_name>[, <property_name>…]) {VALUES | VALUE} <src_vid> -> <dst_vid>[@<rank>] : (<property_value>[, <property_value>…]) [, <src_vid> -> <dst_vid>[@<rank> : (<property_name>[, <property_name>…]), …]
GO [[<M> TO] <N> STEPS ] FROM <vertex_list> OVER <edge_type_list> [REVERSELY] [BIDIRECT] [WHERE <expression> [AND | OR expression …])] YIELD [DISTINCT] <return_list>
FETCH PROP ON {<tag_name> | <tag_name_list> | *} <vid_list> [YIELD [DISTINCT] <return_list>]
FETCH PROP ON <edge_type> <src_vid> -> <dst_vid>[@<rank>] [, <src_vid> -> <dst_vid> …] [YIELD [DISTINCT] <return_list>]
LOOKUP ON {<tag_name> | <edge_type>} WHERE <expression> [AND expression …])] [YIELD <return_list>]
MATCH <pattern> [<WHERE clause>] RETURN <output>
match (s:Symptom{name: "血压过低"}) delete s;
MATCH (a:Person),(b:Movie)
WHERE a.name = 'Tom Hanks' AND b.title = 'Forrest Gump'
CREATE (a)-[r:ACTED_IN { roles:['Forrest'] }]->(b)
RETURN r;
MATCH(s:Symptom), (a:Area) WHERE s.name=$sname AND a.name=$aname
CREATE (s) -[r:symptom_relate_area]->(a)
RETURN r
CREATE CONSTRAINT ON (n:Person) ASSERT n.name IS UNIQUE;
“match (ps:{}{{name:$name}}) -[r]-> (pt) return ps,r,pt”,
MATCH (n:Disease) RETURN n LIMIT 25
match (ps:Disease{name:'风热疮'}) -[r]-> (pt) return ps,r,pt
match (pt) -[r]-> (ps:Disease{name:'风热疮'}) return ps,r,pt
match (pt) -[r]-> (ps:Symptom{name:'肩背痛'}) return ps,r,pt
match (ps:Symptom{name:'肩背痛'}) -[r]-> (pt) return ps,r,pt
- [X] display single main node one hop neighbors
- [ ] select data and train
- [ ] get stats
- [x] 启动旧的工程？
- [ ] 再造旧的界面
- [x] split
server / client - [ ] web debug css
- [x] learn d3.js
- [x] typescript
- [ ] Echarts 文档通读
- [ ] split the demo.html
- [ ] modulize the demo.html
- [x] 词云
- [ ] 整理代码片段
- [ ] split the network and stats
match (x)-[r]->(y) return distinct head(labels(x)) as head, type(r), head(labels(y)) as tail limit 20
match (x)-[r]->(y) return distinct x as head, type(r), r as rname, y as tail limit 20match (n) where not (n)-->() return distinct n
match () -->(t) with t, count(t) as ts where ts = 1 and not (t) --> () return t,tshead return the first element in a list
CREATE CONSTRAINT constraint_Symptom_name IF NOT EXISTS ON (s:Symptom) ASSERT s.name IS UNIQUE;
CREATE CONSTRAINT constraint_Disease_name IF NOT EXISTS ON (s:Disease) ASSERT s.name IS UNIQUE;
CREATE CONSTRAINT constraint_Drug_name IF NOT EXISTS ON (s:Drug) ASSERT s.name IS UNIQUE;
CREATE CONSTRAINT constraint_Department_name IF NOT EXISTS ON (s:Department) ASSERT s.name IS UNIQUE;
CREATE CONSTRAINT constraint_Check_name IF NOT EXISTS ON (s:Check) ASSERT s.name IS UNIQUE;
CREATE CONSTRAINT constraint_Area_name IF NOT EXISTS ON (a:Area) ASSERT a.name IS UNIQUE;MATCH (n:Symptom)
WITH n.name as name, collect(n) AS nodes
WHERE size(nodes) > 1
RETURN nodesMATCH (n:Symptom) WITH n.name as name, collect(n) AS nodes WHERE size(nodes) > 1 with head(nodes) as nn return nn;
MATCH (n:Symptom) WITH collect(n) AS nodes WHERE size(nodes) > 1 WITH head(nodes) as n return n;
MATCH (n:Symptom) WITH n.name as name, collect(n) AS nodes WHERE size(nodes) > 1 UNWIND tail(nodes) as n return n;
MATCH (n:Symptom) WITH n.name as name, collect(n) AS nodes WHERE size(nodes) > 1 return nodes
MATCH (n:Area) WITH n.name as name, collect(n) AS nodes WHERE size(nodes) > 1 return nodes
MATCH (g:Symptom) WITH g.id as id, collect(g) AS nodes WHERE size(nodes) > 1 FOREACH (g in tail(nodes) | DETACH DELETE g)
MATCH (g:Area) WITH g.name as name, collect(g) AS nodes WHERE size(nodes) > 1 FOREACH (g in tail(nodes) | DETACH DELETE g)
match () –>(t) with t, count(t) as ts where ts = 1 and not (t) –> () return t,ts match () –>(t) with t, count(t) as ts where not (t) –> () return t,ts
match (h)-->(t) return t UNION match () -->(t) with t, count(t) as ts where ts = 1 and not (t) --> () return t;- nodes number: 28754
- triples number: 148686
- leaf nodes: 7735
- non-leaf triples: 140951
neo4j@neo4j> match (h) –> (n:Disease{name:”藏毛窦,藏毛病”}) return count(n);
----------
| count(n) |
----------
| 3046 |
----------
match (x) -[r]-> (y) return id(x), id(y), type(r) limit 30;https://github.com/alibaba/euler/wiki/ScalableGCN ScalableGCN是一种由阿里妈妈提出的在大规模图上加速Mini-batch GCN训练速度方法。在Mini-batch GCN中，需要投入的计算力与GCN的层数成指数关系。 并且，下层模型的训练频率显著的高于上层。ScalableGCN利用前向计算和反向计算的Cache，在 *mini-batch之间共享中间层表示的计算结果*，同时维护每个顶点上的异步梯度更新的通路。达到在与GCN层数成线性关系的时间内训练GCN模型的目的。
usetf.scatter_nd_update
https://github.com/WENGSYX/Chinese-Word2vec-Medicine
# gensim.models.KeyedVectors
model = KeyedVectors.load_word2vec_format('Medical.txt', binary=False)https://github.com/lancopku/pkuseg-python
NE: 28754 Missing: 6474 Missing ratio: 22.5%
(h,r): (t1,t2,t3,t4,t5) => train: (h,r): (t1,t2,t3) val: (h,r): (t4) also need (t1,t2,t3) for masked the prediction test: (h,r): (t5) also need (t1,t2,t3) for masked the prediction
record <= hr_map_trn; hr_map_val; hr_map_test <- hr_map
MetaGraphDef with tag-set: 'serve' contains the following SignatureDefs:
signature_def['serving_default']:
The given SavedModel SignatureDef contains the following input(s):
inputs['e1'] tensor_info:
dtype: DT_INT64
shape: (-1)
name: input/e1:0
inputs['rel'] tensor_info:
dtype: DT_INT64
shape: (-1)
name: input/rel:0
The given SavedModel SignatureDef contains the following output(s):
outputs['prediciton'] tensor_info:
dtype: DT_FLOAT
shape: (-1, 28754)
name: prediction:0
Method name is: tensorflow/serving/predict
- repo https://github.com/kykosic/actix-tensorflow-example/blob/master/server/src/main.rs
- blog Face Detection with Actix Web
frontend -> rel::name -> map id -> AdaEInput::from -> predict
cd server && mk r
cd frontend && mk fr