I am trying to create a simple neural network viewer like the diagram below. I can get the trained weights, but where are node values stored in a tensorflow js layer when prediction has run? In other words, I can get the line values, but not the circled values. In a simple network these are as simple as the x and y passed into the fit method.
getWeigths allows to retrieve the weights of a layer
Using tf.model, one can output the prediction of each layer
const input = tf.input({shape: [5]});
const denseLayer1 = tf.layers.dense({units: 10, activation: 'relu'});
const denseLayer2 = tf.layers.dense({units: 2, activation: 'softmax'});
const output1 = denseLayer1.apply(input);
const output2 = denseLayer2.apply(output1);
const model = tf.model({inputs: input, outputs: [output1, output2]});
const [firstLayer, secondLayer] = model.predict(tf.ones([2, 5]));
console.log(denseLayer1.getWeights().length) // 2 W and B for a dense layer
denseLayer1.getWeights()[1].print()
console.log(denseLayer2.getWeights().length) // also 2
// output of each layer WX + B
firstLayer.print();
secondLayer.print()
<html>
<head>
<!-- Load TensorFlow.js -->
<script src="https://cdn.jsdelivr.net/npm/#tensorflow/tfjs#0.12.0"> </script>
</head>
<body>
</body>
</html>
One can also do the same thing using tf.sequential()
const model = tf.sequential();
// first layer
model.add(tf.layers.dense({units: 10, inputShape: [4]}));
// second layer
model.add(tf.layers.dense({units: 1}));
// get all the layers of the model
const layers = model.layers
layers[0].getWeights()[0].print()
<html>
<head>
<!-- Load TensorFlow.js -->
<script src="https://cdn.jsdelivr.net/npm/#tensorflow/tfjs#0.12.0"> </script>
</head>
<body>
</body>
</html>
However with tf.sequential, one cannot get the prediction of each layer as the way one can with tf.model using the output passed as parameter in the config of the model
Related
Trying to retrieve part of a district, however for some reason cannot see the whole area, even if zoom level is at 0, where (supposedly) we should see the whole world.
I am using L.CRS.Simple because this uses the EPSG:3763 and cannot see that one on the CRS list. I am retrieving the data in JSON cause when tying with geoJSON, was not able to transform the 3D coordinates data into 2D planes ones.
const queryRegionText = "where=OBJECTID > 0"
const geoJsonURL2 = "https://sig.cm-figfoz.pt/arcgis/rest/services/Internet/MunisigWeb_DadosContexto/MapServer/2/query?f=json&returnGeometry=true&geometryType=esriGeometryPolyline&spatialRel=esriSpatialRelIntersects&outFields=*&outSR=3763&" + queryRegionText
var map = L.map('mapid', {
crs: L.CRS.Simple
}).setView([-58216.458338, 42768.347232], 0);
L.control.scale({ metric: true }).addTo(map);
fetch(geoJsonURL2).then(function (response) {
response.json().then(function (data) {
data.features.forEach(element => {
if (element.geometry.rings) {
element.geometry.rings.forEach(point => {
L.polyline(point, { color: 'red' }).addTo(map);
})
}
});
});
});
var popup = L.popup();
function onMapClick(e) {
popup
.setLatLng(e.latlng)
.setContent("You clicked the map at " + e.latlng.toString())
.openOn(map);
}
map.on('click', onMapClick);
<html>
<head>
<title>Leaflet - testing</title>
<meta charset="utf-8" />
<link rel="stylesheet" href="https://unpkg.com/leaflet#1.7.1/dist/leaflet.css" />
<script src="https://unpkg.com/leaflet#1.7.1/dist/leaflet.js"></script>
</head>
<body>
<div id="mapid" style="width: 600px; height: 400px;"></div>
</body>
</html>
TL;DR: When creating the map, set the minimum zoom below zero. This should work:
var map = L.map('mapid', {
crs: L.CRS.Simple, minZoom: -6
}).setView([-57728, 55296], -6);
Explanation
Normally, Leaflet translates from a latitude/longitude coordinate system to screen pixels using an assumption that the world is 256 pixels high at Zoom level 0. At each higher Zoom Level, the number of pixels doubles (explained nicely in the Zoom levels tutorial). With this assumption, the options for the map default to {minZoom: 0, maxZoom: Infinity} (as you are not adding any Layer that sets these values to anything different).
When you use L.CRS.Simple, at Zoom level 0 it maps 1 coordinate unit to 1 screen pixel. Your data looks like it is about 18000 coordinate units tall, so it doesn't fit in your 400 pixel high map. To make it fit, we need each screen pixel to map to about 45 coordinate units. 2^5 is 32, and 2^6 is 64, so we need to zoom out between 5 and 6 times. Luckily, Leaflet accepts negative Zoom Levels, so setting zoom to -6 does the trick. But to make it work properly, you need to set {minZoom: -6}, so the map doesn't get stuck at zoom level 0. There's a good worked example in the Non-geographical Maps tutorial.
Using L.CRS.Simple should work for you, so long as the approximation holds that each latitude unit is the same length as each longitude unit (a square world). Since this isn't generally true in the real world, using the Simple projection will cause some distortion. If the distortion is significant for the features you are interested in, then you will need to look up how to use EPSG:3763 properly, using L.CRS and Proj4Leaflet, as suggested by #IvanSanchez.
So, after some reading on the proj4leaflet, come up with this code. Thanks in advance for the comments and the reply above.
const queryRegionText = "where=OBJECTID > 0"
const geoJsonURL2 = "https://sig.cm-figfoz.pt/arcgis/rest/services/Internet/MunisigWeb_DadosContexto/MapServer/2/query?f=geojson&returnGeometry=true&geometryType=esriGeometryPolyline&spatialRel=esriSpatialRelIntersects&outFields=*&outSR=3763&" + queryRegionText
const map = L.map('map', {
center: [40.14791, -8.87009],
zoom: 13
});
proj4.defs("EPSG:3763", "+proj=tmerc +lat_0=39.66825833333333 +lon_0=-8.133108333333334 +k=1 +x_0=0 +y_0=0 +ellps=GRS80 +units=m +no_defs");
fetch(geoJsonURL2).then(function (response) {
response.json().then(function (data) {
L.Proj.geoJson(data).addTo(map);
});
});
var popup = L.popup();
function onMapClick(e) {
popup
.setLatLng(e.latlng)
.setContent("You clicked the map at " + e.latlng.toString())
.openOn(map);
}
map.on('click', onMapClick);
<head>
<link rel="stylesheet" href="https://unpkg.com/leaflet#1.7.1/dist/leaflet.css" />
<link rel="stylesheet" href="main.css">
<script src="https://cdnjs.cloudflare.com/ajax/libs/leaflet/1.7.1/leaflet.js"></script>
<script src="https://cdnjs.cloudflare.com/ajax/libs/proj4js/2.7.4/proj4.js"></script>
<script src="https://cdnjs.cloudflare.com/ajax/libs/proj4leaflet/1.0.2/proj4leaflet.js"></script>
</head>
<body class="">
<div id="map" style="width: 600px; height: 400px;"></div>
</div>
<script src="main.js"></script>
</body>
I'm trying to find my way into Svelte combined with leaflet. Where I'm stuck is how to correctly split the leaflet components into files. For learning, I'm trying to build the official official leaflet quickstart with svelte.
This is how my app.svelte looks like:
<script>
import L from 'leaflet';
import { onMount } from "svelte";
import { Circle } from "./components/Circle.svelte";
let map;
onMount(async () => {
map = L.map("map");
L.tileLayer("https://a.tile.openstreetmap.org/{z}/{x}/{y}.png ", {
attribution:
'Map data © OpenStreetMap contributors, CC-BY-SA',
maxZoom: 18,
tileSize: 512,
zoomOffset: -1
}).addTo(map);
map.setView([51.505, -0.09], 13);
Circle.addTo(map);
});
</script>
<style>
html,body {
padding: 0;
margin: 0;
}
html, body, #map {
height: 100%;
width: 100vw;
}
</style>
<svelte:head>
<link
rel="stylesheet"
href="https://unpkg.com/leaflet#1.6.0/dist/leaflet.css"
integrity="sha512-xwE/Az9zrjBIphAcBb3F6JVqxf46+CDLwfLMHloNu6KEQCAWi6HcDUbeOfBIptF7tcCzusKFjFw2yuvEpDL9wQ=="
crossorigin="" />
</svelte:head>
<div id="map" />
and my circle component:
<script context="module">
import L from 'leaflet';
export let map_obj;
export let Circle = L.circle([51.508, -0.11], {
color: "red",
fillColor: '#f03',
fillOpacity: 0.5,
radius: 500
});
</script>
While this is working I do not think it's effective to consider every component and add it to the map with Circle.addTo(map);. How could I pass in the map object to the circle component or is there some better pattern to build the map with several components?
Note: I do know of svelte/leaflet but like to start from scratch for learning.
This seemingly easy task is complicated due to the not-really-straightforward lifecycle of frameworks like Svelte, and the really-straightforward let-me-do-DOM-stuff architecture of Leaflet.
There are several approaches to this. I'll describe one, based on nesting Svelte components for Leaflet layers inside a Svelte component for a Leaflet map, and using setContext and getContext to handle the Leaflet L.Map instance around. (I'm borrowing this technique from https://github.com/beyonk-adventures/svelte-mapbox )
So a Svelte component for a L.Marker would look like:
<script>
import L from 'leaflet';
import { getContext } from "svelte";
export let lat = 0;
export let lng = 0;
let map = getContext('leafletMapInstance');
L.marker([lat, lng]).addTo(map);
</script>
Easy enough - get the L.Map instance from the Svelte context via getContext, instantiate the L.Marker, add it. This means that there must be a Svelte component for the map setting the context, which will need the components for the markers slotted in, i.e.
<script>
import LeafletMap from './LeafletMap.svelte'
import LeafletMarker from './LeafletMarker.svelte'
</script>
<LeafletMap>
<LeafletMarker lat=40 lng=-3></LeafletMarker>
<LeafletMarker lat=60 lng=10></LeafletMarker>
</LeafletMap>
...and then the Svelte component for the Leaflet map will create the L.Map instance, set it as the context, and be done, right? Not so fast. This is where things get weird.
Because of how Svelte lifecycle works, children components will get "rendered" before parent components, but the parent component needs a DOM element to create the L.Map instance (i.e. the map container). So this could get delayed until the onRender Svelte lifecycle callback, but that would happen after the slotted children get instantiated and their onRender lifecycle callbacks are called. So waiting for Svelte to instantiate a DOM element to contain the map and then instantiate the L.Map and then pass that instance to the context and then getting the context in the marker elements can be quite a nightmare.
So instead, an approach to this is to create a detached DOM element, instantiate a L.Map there, i.e. ...
let map = L.map(L.DomUtil.create('div')
...set it in the context, i.e. ...
import { setContext } from "svelte";
setContext('leafletMapInstance', map);
...this will allow Leaflet layers instantiated by the slotted components to be added to a detached (and thus invisible) map. And once all the lifecycle stuff lets the Svelte component for the L.Map have an actual DOM element attached to the DOM, attach the map container to it, i.e. have this in the HTML section of the Svelte component...
<div class='map' bind:this={mapContainer}>
...and once it's actually attached to the DOM, attach the map container to it and set its size, i.e. ...
let mapContainer;
onMount(function() {
mapContainer.appendChild(map.getContainer());
map.getContainer().style.width = '100%';
map.getContainer().style.height = '100%';
map.invalidateSize();
});
So the entire Svelte component for this Leaflet L.Map would look more or less like...
<script>
import L from "leaflet";
import { setContext, onMount } from "svelte";
let mapContainer;
let map = L.map(L.DomUtil.create("div"), {
center: [0, 0],
zoom: 0,
});
setContext("leafletMapInstance", map);
console.log("map", map);
L.tileLayer("https://a.tile.openstreetmap.org/{z}/{x}/{y}.png ", {
attribution:
'Map data © OpenStreetMap contributors, CC-BY-SA',
}).addTo(map);
onMount(() => {
mapContainer.appendChild(map.getContainer());
map.getContainer().style.width = "100%";
map.getContainer().style.height = "100%";
map.invalidateSize();
});
</script>
<svelte:head>
<link
rel="stylesheet"
href="https://unpkg.com/leaflet#1.6.0/dist/leaflet.css"
integrity="sha512-xwE/Az9zrjBIphAcBb3F6JVqxf46+CDLwfLMHloNu6KEQCAWi6HcDUbeOfBIptF7tcCzusKFjFw2yuvEpDL9wQ=="
crossorigin=""
/>
</svelte:head>
<style>
.map {
height: 100vh;
width: 100vw;
}
</style>
<div class="map" bind:this="{mapContainer}">
<slot></slot>
</div>
See a working example here.
As a side note, I'll say that one should think it twice before sandwiching Leaflet in another JS framework, and think twice about the architecture for this (slotted components seem the cleanest and most extensible, but maybe a big data structure and some imperative programming for the Leaflet bits would be simpler). Sometimes, making sense of the lifecycle implications of more than one framework working at once can be very confusing, and very time-consuming when bugs appear.
I am using several L.markerClusterGroup({})'s so that I can switch them in a Layers Control.
But the Clusters hide behind each other.
I would like to be able to get the total number of both Clusters.
What am I missing?
The issue is that each Leaflet Marker Cluster Group (i.e. L.markerClusterGroup) will perform its own clustering and render Clusters irrespective of what other Cluster Groups may display. Therefore if you have some individual Markers (or any point Features) that are in different Cluster Groups but close to each other, these Groups will display Clusters also close to each other, which may end up overlapping, especially at low zoom level, exactly like in your screenshot.
If you want your Markers to cluster all together (mixing "oranges with apples"), you should use a single Marker Cluster Group.
Now if I understand correctly, your "difficulty" is that you would like to be able to add and remove the Markers dynamically, i.e. in your case the user can use a Layers Control to switch on/off some Features on map.
In that case, you would probably be interested in Leaflet.FeatureGroup.SubGroup plugin (see demo). Simply create 1 subgroup per "switchable" Features Group and set their parent as your single Marker Cluster Group:
var map = L.map('map', {
maxZoom: 18,
}).setView([48.86, 2.35], 11);
var parentGroup = L.markerClusterGroup().addTo(map);
var overlays = {};
for (var i = 1; i <= 5; i += 1) {
overlays['Group ' + i] = L.featureGroup.subGroup(
parentGroup,
getArrayOfMarkers()
).addTo(map);
}
L.control.layers(null, overlays, {
collapsed: false,
}).addTo(map);
function getArrayOfMarkers() {
var result = [];
for (var i = 0; i < 10; i += 1) {
result.push(L.marker(getRandomLatLng()));
}
return result;
}
function getRandomLatLng() {
return [
48.8 + 0.1 * Math.random(),
2.25 + 0.2 * Math.random()
];
}
L.tileLayer('https://{s}.tile.openstreetmap.org/{z}/{x}/{y}.png', {
attribution: '© OpenStreetMap contributors'
}).addTo(map);
html,
body,
#map {
height: 100%;
margin: 0;
}
<!-- Leaflet assets -->
<link rel="stylesheet" href="https://unpkg.com/leaflet#1.3.3/dist/leaflet.css" integrity="sha512-Rksm5RenBEKSKFjgI3a41vrjkw4EVPlJ3+OiI65vTjIdo9brlAacEuKOiQ5OFh7cOI1bkDwLqdLw3Zg0cRJAAQ==" crossorigin="" />
<script src="https://unpkg.com/leaflet#1.3.3/dist/leaflet.js" integrity="sha512-tAGcCfR4Sc5ZP5ZoVz0quoZDYX5aCtEm/eu1KhSLj2c9eFrylXZknQYmxUssFaVJKvvc0dJQixhGjG2yXWiV9Q==" crossorigin=""></script>
<!-- Leaflet.markercluster assets -->
<link rel="stylesheet" href="https://unpkg.com/leaflet.markercluster#1.3.0/dist/MarkerCluster.css">
<link rel="stylesheet" href="https://unpkg.com/leaflet.markercluster#1.3.0/dist/MarkerCluster.Default.css">
<script src="https://unpkg.com/leaflet.markercluster#1.3.0/dist/leaflet.markercluster-src.js"></script>
<!-- Leaflet.FeatureGroup.SubGroup assets -->
<script src="https://unpkg.com/leaflet.featuregroup.subgroup#1.0.2/dist/leaflet.featuregroup.subgroup.js"></script>
<div id="map"></div>
See also Cluster multiple Layers with markercluster
Note: for more complex use cases, there is also another plugin Leaflet.MarkerCluster.LayerSupport.
See also How to apply leaflet marker cluster using layers
Disclosure: I am the author of these plugins.
I have the following data in a Google Spreadsheet:
ColumnA ColumnB
Departament Quantity
01/01/2016 23
02/01/2016 43
04/01/2016 5
06/01/2016 65
10/01/2016 12
11/01/2016 32
13/01/2016 22
15/02/2016 2
And want to draw a linechart using HTML Templates: This is my code so far now
<!DOCTYPE html>
<html>
<head>
<script src="https://www.google.com/jsapi"></script>
</head>
<body>
<div id="main"></div>
<script>
google.load('visualization', '1', {packages: ['corechart', 'bar']});
google.setOnLoadCallback(getSpreadsheetData);
function getSpreadsheetData() {
google.script.run.withSuccessHandler(drawChart).getSpreadsheetData();
}
function drawChart(rows) {
var options = {'title':'Example','width':400,'height':300};
var data = google.visualization.arrayToDataTable(rows, false);
var chart = newgoogle.visualization.LineChart(document.getElementById("main"));
chart.draw(data, options)
}
</script>
</body>
</html>
My script to read the sheet:
function getSpreadsheetData() {
var ssID = "12GvIStMKqmRFNBM-C67NCDeb89-c55K7KQtcuEYmJWQ",
sheet = SpreadsheetApp.openById(ssID).getSheets()[0],
data = sheet.getDataRange().getValues();
return data;
}
However i cant draw the plot, this issue dissapear when i change the data to a numeric (i.e. 42370), but thats not what i want!
Question is: what do i have to add in order to change to the right format in x axis?
You can use the hAxis.format configuration option
var options = {'title':'Example','width':400,'height':300,//-->'format':'MM/dd/yyyy'};
The valid format options can be found here...
And if you need to load with a specific locale, see this...
I am trying to add min and max limits to a Google chart, which I generate using a Perl script from CSV data - by using the interval role for these 2 values.
Unfortunately the I-lines are not displayed at my line chart, even though I've set the min and max limits to the -100 and 100 for the sake of testing.
Only the main data is being displayed:
Can anybody please spot the error, what is wrong with my very simple test case?
Please just save the code below as an HTML-file and open in a browser:
<!DOCTYPE HTML>
<html>
<head>
<script type="text/javascript" src="https://ajax.googleapis.com/ajax/libs/jquery/1/jquery.min.js"></script>
<script type="text/javascript" src="https://www.google.com/jsapi?autoload={'modules':[{'name':'visualization','version':'1','packages':['corechart']}]}"></script>
<script type="text/javascript">
var data = {"L_B8_ACLR_50_0_QPSK_1_H":{"rows":[
{"c":[{"v":"UTRA_1_DOWN"},{"v":-100},{"v":100},{"v":"-42.46912"}]},
{"c":[{"v":"E-UTRA_1_DOWN"},{"v":-100},{"v":100},{"v":"-39.9545"}]},
{"c":[{"v":"E-UTRA_1_UP"},{"v":-100},{"v":100},{"v":"-48.68408"}]},
{"c":[{"v":"UTRA_1_UP"},{"v":-100},{"v":100},{"v":"-49.45148"}]},
{"c":[{"v":"UTRA_2_UP"},{"v":-100},{"v":100},{"v":"-58.96674"}]}],
"cols":[
{"p":{"role":"domain"},"label":"MEASUREMENT","type":"string"},
{"p":{"role":"interval"},"label":"LSL","type":"number"},
{"p":{"role":"interval"},"label":"USL","type":"number"},
{"p":{"role":"data"},"label":"1142926087","type":"number"}]}};
function drawCharts() {
for (var csv in data) {
var x = new google.visualization.DataTable(data[csv]);
var options = {
title: csv,
width: 800,
height: 600
};
var chart = new google.visualization.LineChart(document.getElementById(csv));
chart.draw(x, options);
}
}
$(function() {
google.setOnLoadCallback(drawCharts);
});
</script>
</head>
<body>
<div id="L_B8_ACLR_50_0_QPSK_1_H"></div>
</body>
</html>
(I don't want to use methods like addColumn or addRows. Instead I prepare my data as data structure in my Perl script and then JSON-encode and pass it to DataTable ctr).
You must specify the interval-role column after the data-column. As written in the API :
"All columns except domain columns apply to the nearest left neighbor to which it can be applied"
So if you change the order (and here with some smaller intervals)
var data = {"L_B8_ACLR_50_0_QPSK_1_H":{"rows":[
{"c":[{"v":"UTRA_1_DOWN"},{"v":"-42.46912"},{"v":-50},{"v":-45}]},
{"c":[{"v":"E-UTRA_1_DOWN"},{"v":"-39.9545"},{"v":-50},{"v":-45}]},
{"c":[{"v":"E-UTRA_1_UP"},{"v":"-48.68408"},{"v":-50},{"v":-45}]},
{"c":[{"v":"UTRA_1_UP"},{"v":"-49.45148"},{"v":-50},{"v":-45}]},
{"c":[{"v":"UTRA_2_UP"},{"v":"-58.96674"},{"v":-50},{"v":-45}]}],
"cols":[
{"p":{"role":"domain"},"label":"MEASUREMENT","type":"string"},
{"p":{"role":"data"},"label":"1142926087","type":"number"},
{"p":{"role":"interval"},"label":"LSL","type":"number"},
{"p":{"role":"interval"},"label":"USL","type":"number"}
]}};
..You end up with :