The avatar for @jeyabbalas
Jeya Balaji Balasubramanian
AI Researcher in Medicine. All opinions are my own.
Public
Edited
Nov 8, 2022
5 forks
11 stars
Artificial Intelligence
Interactively Assessing Disentanglement in GANsAnalyzing the Design Space for Visualizing Neural Attention in Text Classification
Tensors
Image tensorsVisualization in Deep LearningInteractive Convolutional Neural NetworkMachine Learning in JS with TensorFlow.js (Part I)How to visualize timeline of a Wiki article?Simulation ResultsMachine learning in the browserRecommender systemsPython in Observable through PyodideArtificial intelligence systems Timelinebackground-position ScrubberDrawings to HumansahajBERT - Bubbles chartPretrained models (image classification)Advanced pretrained models (object detection)Robotics Perturbation: Broken Joint (Rotation)Multi-Agent Evaluation: Object Counting
Also listed in…
TensorFlow.js
Insert cell
Insert cell
Insert cell
`The TF.js version used in this notebook is: ${tf.version.tfjs}`
Insert cell
Insert cell
myTensor = tf.tensor([1, 2, 3, 10, 20, 30], [2, 3], 'int32')
Insert cell
Insert cell
Insert cell
myTensor.toString()
Insert cell
Insert cell
`
This tensor's dimension is: ${myTensor.rank}.
The shape of this tensor is: [${myTensor.shape}].
The number of elements in this tensor is: ${myTensor.size}.
The data type of this tensor is: ${myTensor.dtype}.
`
Insert cell
Insert cell
myTensor.array()
Insert cell
myTensor.data()
Insert cell
Insert cell
tf.memory()
Insert cell
Insert cell
{
let numTensorsStart = tf.memory().numTensors;
let numTensorsInsideTidy, numTensorsOutsideTidy, numTensorsClean, numTensorsOrphaned;
let toKeep, toReturn, toCollect, toBatman;

toBatman = tf.tensor([1, 2, 3]);
tf.tidy(() => {
toKeep = tf.tensor([4, 5, 6]);
toReturn = tf.tensor([7, 8, 9]);
toCollect = tf.tensor([10, 11, 12]);

numTensorsInsideTidy = tf.memory().numTensors - numTensorsStart;
tf.keep(toKeep);
return toReturn;
});
numTensorsOutsideTidy = tf.memory().numTensors - numTensorsStart;

toKeep.dispose();
toReturn.dispose();
numTensorsClean = tf.memory().numTensors - numTensorsStart;
toBatman = toBatman.asType('int32'); // original Float32 is orphaned; source of memory leak
toBatman.dispose();
numTensorsOrphaned = tf.memory().numTensors - numTensorsStart;
return `
After all four tensors were created, the number of new tensors inside tidy() was: ${numTensorsInsideTidy}.
After tidy() garbage collected the toCollect tensor, the number of tensors just outside tidy() was: ${numTensorsOutsideTidy}.
After disposing off toKeep and toReturn, the number of tensors was: ${numTensorsClean}.
After orphaning toBatman, the number of tensors was: ${numTensorsOrphaned}.
`
}
Insert cell
Insert cell
myTensorBuffer = tf.tensor(myTensor.arraySync()).buffer()
Insert cell
Insert cell
myTensorBuffer.get(0, 2)
Insert cell
Insert cell
myTensorBuffer.set(999, 0, 0)
Insert cell
Insert cell
myModifiedTensor = myTensorBuffer.toTensor()
Insert cell
myModifiedTensor.toString()
Insert cell
Insert cell
numHoursStudied = [0.50, 0.75, 1.00, 1.25, 1.50, 1.75, 1.75, 2.00, 2.25, 2.50,
2.75, 3.00, 3.25, 3.50, 4.00, 4.25, 4.50, 4.75, 5.00, 5.50]
Insert cell
Insert cell
studentPassedExam = [0, 0, 0, 0, 0, 0, 1, 0, 1, 0,
1, 0, 1, 0, 1, 1, 1, 1, 1, 1]
Insert cell
Insert cell
predictor = tf.tensor(numHoursStudied, [numHoursStudied.length, 1])
Insert cell
predictor.shape
Insert cell
predictor.toString()
Insert cell
Insert cell
ones = predictor.onesLike()
Insert cell
ones.toString()
Insert cell
ones.shape
Insert cell
Insert cell
x = ones.concat([predictor], 1)
Insert cell
x.toString()
Insert cell
x.shape
Insert cell
Insert cell
betas = [-4.1, 1.5]
Insert cell
Insert cell
b = tf.tensor(betas, [1, betas.length])
Insert cell
b.toString()
Insert cell
b.shape
Insert cell
Insert cell
logits = x.matMul(b.transpose())
Insert cell
logits.toString()
Insert cell
Insert cell
yPred = logits.sigmoid()
Insert cell
Insert cell
yPred.toString()
Insert cell
Insert cell
predictedOutcomes = yPred.data()
Insert cell
Insert cell
Insert cell
Insert cell
Insert cell
xRange = tf.linspace(0, 6, 100).reshape([100, 1])
Insert cell
Insert cell
xExtra = tf.concat([tf.onesLike(xRange), xRange], 1)
Insert cell
Insert cell
yPredExtra = xExtra.matMul(b.transpose()).sigmoid()
Insert cell
Insert cell
xRangeArray = xRange.data()
Insert cell
yPredExtraArray = yPredExtra.data()
Insert cell
Insert cell
Insert cell
Insert cell
tf = require('@tensorflow/tfjs@4.0.0/dist/tf.min.js')
Insert cell
Plotly = require("https://cdn.plot.ly/plotly-2.14.0.min.js")
Insert cell

One platform to build and deploy the best data apps

Experiment and prototype by building visualizations in live JavaScript notebooks. Collaborate with your team and decide which concepts to build out.
Use Observable Framework to build data apps locally. Use data loaders to build in any language or library, including Python, SQL, and R.
Seamlessly deploy to Observable. Test before you ship, use automatic deploy-on-commit, and ensure your projects are always up-to-date.
Learn more