Name: Towards AI Legal Name: Towards AI, Inc. Description: Towards AI is the world's leading artificial intelligence (AI) and technology publication. Read by thought-leaders and decision-makers around the world. Phone Number: +1-650-246-9381 Email: pub@towardsai.net
228 Park Avenue South New York, NY 10003 United States
Website: Publisher: https://towardsai.net/#publisher Diversity Policy: https://towardsai.net/about Ethics Policy: https://towardsai.net/about Masthead: https://towardsai.net/about
Name: Towards AI Legal Name: Towards AI, Inc. Description: Towards AI is the world's leading artificial intelligence (AI) and technology publication. Founders: Roberto Iriondo, , Job Title: Co-founder and Advisor Works for: Towards AI, Inc. Follow Roberto: X, LinkedIn, GitHub, Google Scholar, Towards AI Profile, Medium, ML@CMU, FreeCodeCamp, Crunchbase, Bloomberg, Roberto Iriondo, Generative AI Lab, Generative AI Lab Denis Piffaretti, Job Title: Co-founder Works for: Towards AI, Inc. Louie Peters, Job Title: Co-founder Works for: Towards AI, Inc. Louis-François Bouchard, Job Title: Co-founder Works for: Towards AI, Inc. Cover:
Towards AI Cover
Logo:
Towards AI Logo
Areas Served: Worldwide Alternate Name: Towards AI, Inc. Alternate Name: Towards AI Co. Alternate Name: towards ai Alternate Name: towardsai Alternate Name: towards.ai Alternate Name: tai Alternate Name: toward ai Alternate Name: toward.ai Alternate Name: Towards AI, Inc. Alternate Name: towardsai.net Alternate Name: pub.towardsai.net
5 stars – based on 497 reviews

Frequently Used, Contextual References

TODO: Remember to copy unique IDs whenever it needs used. i.e., URL: 304b2e42315e

Resources

Take our 85+ lesson From Beginner to Advanced LLM Developer Certification: From choosing a project to deploying a working product this is the most comprehensive and practical LLM course out there!

Publication

Cosine Similarity for 1 Trillion Pairs of Vectors
Latest   Machine Learning

Cosine Similarity for 1 Trillion Pairs of Vectors

Last Updated on July 25, 2023 by Editorial Team

Author(s): Rodrigo Agundez

Originally published on Towards AI.

Introducing ChunkDot

Photo by Tamas Pap on Unsplash

UPDATE
ChunkDot now supports sparse embeddings, you can read more about it here.

Bulk Similarity Calculations for Sparse Embeddings

ChunkDot support for sparse matrices

pub.towardsai.net

Success! I managed to find vector representations for my 1 million items… I just need to calculate the cosine similarity between all pairs to find the 20 most similar items per item. Then, from sklearn.metrics.pairwise import cosine_similarity, throw my 1 million vectorized items into it, and bam!
My computer explodes with a MemoryError or the kernel of my Jupyter notebook dies. The issue, I just tried to hold 1 Trillion (10¹²) values in memory (~8000 GB¹). Good luck!

This blog post introduces ChunkDot, a Python package I built to solve this problem. ChunkDot can do this calculation in ~2 hours while keeping the memory consumption under 20 GB, as shown below.

top_k = 100 ; embedding_dim = 256 ; n_threads = 16

If you would like to skip the explanation and jump directly into the usage instructions:

chunkdot

Multi-threaded matrix multiplication and cosine similarity calculations. Appropriate for the calculation of the K most…

pypi.org

Motivation

The first time I encountered this problem was ~6 years ago while working for a client on an entity-matching use case. I was trying to match millions of records saved as free text to create golden records with the aggregated information from similar records². Back then, Spark was a new and very shiny tool, so I solved it using Spark, with some broadcasting and some C++ code wrapped in Cython wrapped in a UDF. I remember feeling that there might be a more elegant and simple way to solve this problem. I finally had time to look into it.

Since then, I have seen this issue pop up in one-to-many use cases. The idea of having items as vector representations and then trying to find the K most similar or dissimilar items per item appears in numerous use cases in Machine Learning. Some of these use cases are:

  • Similar product recommendations.
  • Name and Entity matching.
  • Forecasting cold-start problem.
  • Recommendation systems cold-start problem.
  • User segmentation.

ChunkDot

The core of the problem is that the number of calculations scales with the square of the number of items. If we have N items, then the number of item pairs scales as N².
There is a silver lining, though, we only need the K most similar items per item. To determine which are the most similar K, we need to calculate all similarity pairs, but what if we do the calculation in batches (or chunks)? Keep the K most similar items for each batch, discard the rest, and then continue with the next batch until we have processed all items. This is the idea behind ChunkDot.

Let’s go over an example, suppose I have 10⁶ items, and I only require the top 100 most similar items per item. Then the number of similarity values I need is not 10¹² but 10⁶ x 100 = 10⁸, a reduction of 99.99%.³ This does not mean that the memory consumption will be reduced by 99.99%. It means the memory is no longer set by the output matrix but by the memory consumption of the algorithm itself.

Cosine Similarity Top K

Let’s go over how ChunkDot works. Even though the core of the implementation is in chunkdot._chunkdot, I’ll focus on the chunkdot.cosine_similarity_top_k function as shown below.

Diagram describing ChunkDot’s Cosine Similarity Top K algorithm

The algorithm consists of 3 parts, splitting the embeddings matrix into the optimal number of chunks, performing the parallelized operations over the chunks, and collecting the data from each parallel thread to form the similarity matrix.

Splitting the input embeddings into chunks

First part of ChunkDot’s Cosine Similarity Top K algorithm

The first part of the algorithm receives as input a matrix of embeddings where each row is the vector representation of an item. Then it calculates the optimal chunk size to use during the parallelization. The optimal chunk size is the maximum number of rows to process per thread without exceeding the given maximum memory to use. If no maximum memory to use is given, then the system’s available memory is used. In parallel, an optional L2 normalization of the rows can be done. Finally, the normalized embeddings matrix gets split into pieces where each piece contains the number of rows equal to the chunk size. Now the data is ready to be parallelized.

Divide and conquer

Second part of ChunkDot’s Cosine Similarity Top K algorithm

The second part of the algorithm performs all the parallel computations; each parallel thread calculates the cosine similarity between the chunk of items vs. all the items. As shown in the figure above, matrix multiplication is sufficient since rows are L2 normalized. Then, only the indices and values of the K most similar items are retrieved so that the density matrix is garbage collected to free up memory.

There is no free lunch; the price to pay for the parallelization and memory footprint reduction is that instead of having one big matrix multiplication, we have thousands of smaller ones plus the overhead of parallelization. This added complexity can raise the execution time to a non-usable level in practice. To manage the parallelization overhead efficiently and to perform these calculations as fast as possible ChunkDot uses Numba.

“Numba generates optimized machine code from pure Python code using the LLVM compiler infrastructure. With a few simple annotations, array-oriented and math-heavy Python code can be just-in-time optimized to performance similar as C, C++ and Fortran, without having to switch languages or Python interpreters.”

To take the most out of Numba, the code to optimize needs to be supported in no-python mode. At the moment of this writing, the numpy.argpartition function is not supported, which is why there is a numba_argpartition.py file in ChunkDot’s source code. I submitted this implementation to Numba via the PR below, which has been merged. The support should become native on Numba’s next release.

Add numpy argpartition function support by rragundez · Pull Request #8732 · numba/numba

Add numpy argpartition function support. I used some of the implementation of partition. Something to note it that I…

github.com

After the cosine similarity calculations and the retrieval of the K most similar items, we can collect all the values and indices to construct the final sparse similarity matrix.

Similarities as a CSR sparse matrix

Third part of ChunkDot’s Cosine Similarity Top K algorithm

The third part of the algorithm consists of creating the sparse similarity matrix containing the K most similar items per item. This will be a sparse matrix of dimensions n_items x n_items containing only n_items x top_k non-zero values. The final output will the returned as a SciPy CSR sparse matrix.

Chunk size calculation

The idea behind calculating the biggest possible chunk size is that the bigger the chunk, the fewer threads are needed, therefore reducing the execution time of the algorithm.
To reduce memory utilization, all threads outputs are collected in arrays that are passed by reference to each thread. ChunkDot’s algorithm memory consumption is then described by⁴:

Where the number of threads is given by Numba’s get_num_threads function. The optimal chunk size is the solution for chunk_size in the equation above.

Memory and speed

I made a benchmark comparison with SciKit-Learn’s cosine similarity on memory consumption and execution time.

This is not entirely a fair comparison since the purpose of SciKit-Learn’s cosine similarity function is to retrieve the similarities of all pairs of items and not just the K most similar. Having said that, I do not know of any other similar and simple approach that can be executed on a common laptop without a rigorous setup. I also think most of us will be looking first at the SciKit-Learn function to solve this problem. Therefore it seemed appropriate to benchmark against it.

The figure below shows the results. For the SciKit-Learn implementation, I did not go above 50K items as it was already taking ~20 GB of memory. I did include a projection of the expected memory consumption.

top_k = 100 ; embedding_dim = 256 ; n_threads = 16

The results show that the algorithm works, it can do the calculation for a large number of items while keeping the memory consumption capped. In the above-left figure, SciKit-Learn’s implementation is compared with ChunkDot using 3 different maximum memories: 5 GB, 10 GB, and 20 GB.

Nowadays is common to have 20 GB of RAM; you can easily compute 100K different items and retrieve their K most similar items in around 1min.
Quite unexpectedly, the ChunkDot algorithm is also faster, thanks, Numba! and it seems that the maximum memory to use does not matter that much for execution time… weird.
I did a thorough review of the results and methodology and reran them several times, but I wouldn’t claim that ChunkDot is univocally faster. What I do know for certain is that I was able to do the calculation for 100K and 1 million items without a problem.
For 1 million items, which are 1 trillion comparisons, I only had to wait ~2 hours using 20 GB of memory, as shown in the figure below.

top_k = 100 ; embedding_dim = 256 ; n_threads = 16

Usage

pip install -U chunkdot

Calculate the 50 most similar and dissimilar items for 100K items.

import numpy as np
from chunkdot import cosine_similarity_top_k

embeddings = np.random.randn(100000, 256)
# using all you system's memory
cosine_similarity_top_k(embeddings, top_k=50)
# most dissimilar items using 20 GB
cosine_similarity_top_k(embeddings, top_k=-50, max_memory=20E9)

<100000x100000 sparse matrix of type '<class 'numpy.float64'>'
with 5000000 stored elements in Compressed Sparse Row format>

Conclusion

I hope you find this blog post and ChunkDot useful! I enjoyed a lot working on this problem and getting satisfactory results. Some potential improvements:

  • Add support for input embeddings as a sparse matrix. This might already be supported, but I just haven’t thoroughly unit tested for this. I decided to return a TypeError exception to avoid silent errors. This will especially enable NLP use cases where the vectorization of items is done using n-grams, bag-of-words, or similar.

UPDATE. ChunkDot now supports sparse embeddings, you can read more about it here.

Bulk Similarity Calculations for Sparse Embeddings

ChunkDot support for sparse matrices

pub.towardsai.net

  • The cosine similarity between item i and j, is equal to the similarity between j and i. It is enough to perform the calculations that render the upper-triangular matrix. This can be done via the Einstein summation function in Numpy, unfortunately, is not supported by Numba at the moment.
  • Instead of returning the K most similar items, return items whose similarity is above/below a certain threshold.
  • Add GPU support since Numba supports it.

Numba for CUDA GPUs — Numba 0.56.4+0.g288a38bbd.dirty-py3.7-linux-x86_64.egg documentation

Callback into the Python Interpreter from within JIT’ed code

numba.readthedocs.io

[1] These calculations are normally returned as float64. Not that it matters, though, with float32 I would still need ~4000 GB of memory.

[2] The situation was a bit more complex, but this is the part related to this blog post.

[3] Here I have ignored the memory contribution from storing the indexes needed to assign each similarity value to an item pair. Their memory contribution, though, is irrelevant to this conclusion.

[4] Supposing that all threads could be holding the maximum amount of memory at the same time.

Join thousands of data leaders on the AI newsletter. Join over 80,000 subscribers and keep up to date with the latest developments in AI. From research to projects and ideas. If you are building an AI startup, an AI-related product, or a service, we invite you to consider becoming a sponsor.

Published via Towards AI

Feedback ↓

Sign Up for the Course
`; } else { console.error('Element with id="subscribe" not found within the page with class "home".'); } } }); // Remove duplicate text from articles /* Backup: 09/11/24 function removeDuplicateText() { const elements = document.querySelectorAll('h1, h2, h3, h4, h5, strong'); // Select the desired elements const seenTexts = new Set(); // A set to keep track of seen texts const tagCounters = {}; // Object to track instances of each tag elements.forEach(el => { const tagName = el.tagName.toLowerCase(); // Get the tag name (e.g., 'h1', 'h2', etc.) // Initialize a counter for each tag if not already done if (!tagCounters[tagName]) { tagCounters[tagName] = 0; } // Only process the first 10 elements of each tag type if (tagCounters[tagName] >= 2) { return; // Skip if the number of elements exceeds 10 } const text = el.textContent.trim(); // Get the text content const words = text.split(/\s+/); // Split the text into words if (words.length >= 4) { // Ensure at least 4 words const significantPart = words.slice(0, 5).join(' '); // Get first 5 words for matching // Check if the text (not the tag) has been seen before if (seenTexts.has(significantPart)) { // console.log('Duplicate found, removing:', el); // Log duplicate el.remove(); // Remove duplicate element } else { seenTexts.add(significantPart); // Add the text to the set } } tagCounters[tagName]++; // Increment the counter for this tag }); } removeDuplicateText(); */ // Remove duplicate text from articles function removeDuplicateText() { const elements = document.querySelectorAll('h1, h2, h3, h4, h5, strong'); // Select the desired elements const seenTexts = new Set(); // A set to keep track of seen texts const tagCounters = {}; // Object to track instances of each tag // List of classes to be excluded const excludedClasses = ['medium-author', 'post-widget-title']; elements.forEach(el => { // Skip elements with any of the excluded classes if (excludedClasses.some(cls => el.classList.contains(cls))) { return; // Skip this element if it has any of the excluded classes } const tagName = el.tagName.toLowerCase(); // Get the tag name (e.g., 'h1', 'h2', etc.) // Initialize a counter for each tag if not already done if (!tagCounters[tagName]) { tagCounters[tagName] = 0; } // Only process the first 10 elements of each tag type if (tagCounters[tagName] >= 10) { return; // Skip if the number of elements exceeds 10 } const text = el.textContent.trim(); // Get the text content const words = text.split(/\s+/); // Split the text into words if (words.length >= 4) { // Ensure at least 4 words const significantPart = words.slice(0, 5).join(' '); // Get first 5 words for matching // Check if the text (not the tag) has been seen before if (seenTexts.has(significantPart)) { // console.log('Duplicate found, removing:', el); // Log duplicate el.remove(); // Remove duplicate element } else { seenTexts.add(significantPart); // Add the text to the set } } tagCounters[tagName]++; // Increment the counter for this tag }); } removeDuplicateText(); //Remove unnecessary text in blog excerpts document.querySelectorAll('.blog p').forEach(function(paragraph) { // Replace the unwanted text pattern for each paragraph paragraph.innerHTML = paragraph.innerHTML .replace(/Author\(s\): [\w\s]+ Originally published on Towards AI\.?/g, '') // Removes 'Author(s): XYZ Originally published on Towards AI' .replace(/This member-only story is on us\. Upgrade to access all of Medium\./g, ''); // Removes 'This member-only story...' }); //Load ionic icons and cache them if ('localStorage' in window && window['localStorage'] !== null) { const cssLink = 'https://code.ionicframework.com/ionicons/2.0.1/css/ionicons.min.css'; const storedCss = localStorage.getItem('ionicons'); if (storedCss) { loadCSS(storedCss); } else { fetch(cssLink).then(response => response.text()).then(css => { localStorage.setItem('ionicons', css); loadCSS(css); }); } } function loadCSS(css) { const style = document.createElement('style'); style.innerHTML = css; document.head.appendChild(style); } //Remove elements from imported content automatically function removeStrongFromHeadings() { const elements = document.querySelectorAll('h1, h2, h3, h4, h5, h6, span'); elements.forEach(el => { const strongTags = el.querySelectorAll('strong'); strongTags.forEach(strongTag => { while (strongTag.firstChild) { strongTag.parentNode.insertBefore(strongTag.firstChild, strongTag); } strongTag.remove(); }); }); } removeStrongFromHeadings(); "use strict"; window.onload = () => { /* //This is an object for each category of subjects and in that there are kewords and link to the keywods let keywordsAndLinks = { //you can add more categories and define their keywords and add a link ds: { keywords: [ //you can add more keywords here they are detected and replaced with achor tag automatically 'data science', 'Data science', 'Data Science', 'data Science', 'DATA SCIENCE', ], //we will replace the linktext with the keyword later on in the code //you can easily change links for each category here //(include class="ml-link" and linktext) link: 'linktext', }, ml: { keywords: [ //Add more keywords 'machine learning', 'Machine learning', 'Machine Learning', 'machine Learning', 'MACHINE LEARNING', ], //Change your article link (include class="ml-link" and linktext) link: 'linktext', }, ai: { keywords: [ 'artificial intelligence', 'Artificial intelligence', 'Artificial Intelligence', 'artificial Intelligence', 'ARTIFICIAL INTELLIGENCE', ], //Change your article link (include class="ml-link" and linktext) link: 'linktext', }, nl: { keywords: [ 'NLP', 'nlp', 'natural language processing', 'Natural Language Processing', 'NATURAL LANGUAGE PROCESSING', ], //Change your article link (include class="ml-link" and linktext) link: 'linktext', }, des: { keywords: [ 'data engineering services', 'Data Engineering Services', 'DATA ENGINEERING SERVICES', ], //Change your article link (include class="ml-link" and linktext) link: 'linktext', }, td: { keywords: [ 'training data', 'Training Data', 'training Data', 'TRAINING DATA', ], //Change your article link (include class="ml-link" and linktext) link: 'linktext', }, ias: { keywords: [ 'image annotation services', 'Image annotation services', 'image Annotation services', 'image annotation Services', 'Image Annotation Services', 'IMAGE ANNOTATION SERVICES', ], //Change your article link (include class="ml-link" and linktext) link: 'linktext', }, l: { keywords: [ 'labeling', 'labelling', ], //Change your article link (include class="ml-link" and linktext) link: 'linktext', }, pbp: { keywords: [ 'previous blog posts', 'previous blog post', 'latest', ], //Change your article link (include class="ml-link" and linktext) link: 'linktext', }, mlc: { keywords: [ 'machine learning course', 'machine learning class', ], //Change your article link (include class="ml-link" and linktext) link: 'linktext', }, }; //Articles to skip let articleIdsToSkip = ['post-2651', 'post-3414', 'post-3540']; //keyword with its related achortag is recieved here along with article id function searchAndReplace(keyword, anchorTag, articleId) { //selects the h3 h4 and p tags that are inside of the article let content = document.querySelector(`#${articleId} .entry-content`); //replaces the "linktext" in achor tag with the keyword that will be searched and replaced let newLink = anchorTag.replace('linktext', keyword); //regular expression to search keyword var re = new RegExp('(' + keyword + ')', 'g'); //this replaces the keywords in h3 h4 and p tags content with achor tag content.innerHTML = content.innerHTML.replace(re, newLink); } function articleFilter(keyword, anchorTag) { //gets all the articles var articles = document.querySelectorAll('article'); //if its zero or less then there are no articles if (articles.length > 0) { for (let x = 0; x < articles.length; x++) { //articles to skip is an array in which there are ids of articles which should not get effected //if the current article's id is also in that array then do not call search and replace with its data if (!articleIdsToSkip.includes(articles[x].id)) { //search and replace is called on articles which should get effected searchAndReplace(keyword, anchorTag, articles[x].id, key); } else { console.log( `Cannot replace the keywords in article with id ${articles[x].id}` ); } } } else { console.log('No articles found.'); } } let key; //not part of script, added for (key in keywordsAndLinks) { //key is the object in keywords and links object i.e ds, ml, ai for (let i = 0; i < keywordsAndLinks[key].keywords.length; i++) { //keywordsAndLinks[key].keywords is the array of keywords for key (ds, ml, ai) //keywordsAndLinks[key].keywords[i] is the keyword and keywordsAndLinks[key].link is the link //keyword and link is sent to searchreplace where it is then replaced using regular expression and replace function articleFilter( keywordsAndLinks[key].keywords[i], keywordsAndLinks[key].link ); } } function cleanLinks() { // (making smal functions is for DRY) this function gets the links and only keeps the first 2 and from the rest removes the anchor tag and replaces it with its text function removeLinks(links) { if (links.length > 1) { for (let i = 2; i < links.length; i++) { links[i].outerHTML = links[i].textContent; } } } //arrays which will contain all the achor tags found with the class (ds-link, ml-link, ailink) in each article inserted using search and replace let dslinks; let mllinks; let ailinks; let nllinks; let deslinks; let tdlinks; let iaslinks; let llinks; let pbplinks; let mlclinks; const content = document.querySelectorAll('article'); //all articles content.forEach((c) => { //to skip the articles with specific ids if (!articleIdsToSkip.includes(c.id)) { //getting all the anchor tags in each article one by one dslinks = document.querySelectorAll(`#${c.id} .entry-content a.ds-link`); mllinks = document.querySelectorAll(`#${c.id} .entry-content a.ml-link`); ailinks = document.querySelectorAll(`#${c.id} .entry-content a.ai-link`); nllinks = document.querySelectorAll(`#${c.id} .entry-content a.ntrl-link`); deslinks = document.querySelectorAll(`#${c.id} .entry-content a.des-link`); tdlinks = document.querySelectorAll(`#${c.id} .entry-content a.td-link`); iaslinks = document.querySelectorAll(`#${c.id} .entry-content a.ias-link`); mlclinks = document.querySelectorAll(`#${c.id} .entry-content a.mlc-link`); llinks = document.querySelectorAll(`#${c.id} .entry-content a.l-link`); pbplinks = document.querySelectorAll(`#${c.id} .entry-content a.pbp-link`); //sending the anchor tags list of each article one by one to remove extra anchor tags removeLinks(dslinks); removeLinks(mllinks); removeLinks(ailinks); removeLinks(nllinks); removeLinks(deslinks); removeLinks(tdlinks); removeLinks(iaslinks); removeLinks(mlclinks); removeLinks(llinks); removeLinks(pbplinks); } }); } //To remove extra achor tags of each category (ds, ml, ai) and only have 2 of each category per article cleanLinks(); */ //Recommended Articles var ctaLinks = [ /* ' ' + '

Subscribe to our AI newsletter!

' + */ '

Take our 85+ lesson From Beginner to Advanced LLM Developer Certification: From choosing a project to deploying a working product this is the most comprehensive and practical LLM course out there!

'+ '

Towards AI has published Building LLMs for Production—our 470+ page guide to mastering LLMs with practical projects and expert insights!

' + '
' + '' + '' + '

Note: Content contains the views of the contributing authors and not Towards AI.
Disclosure: This website may contain sponsored content and affiliate links.

' + 'Discover Your Dream AI Career at Towards AI Jobs' + '

Towards AI has built a jobs board tailored specifically to Machine Learning and Data Science Jobs and Skills. Our software searches for live AI jobs each hour, labels and categorises them and makes them easily searchable. Explore over 10,000 live jobs today with Towards AI Jobs!

' + '
' + '

🔥 Recommended Articles 🔥

' + 'Why Become an LLM Developer? Launching Towards AI’s New One-Stop Conversion Course'+ 'Testing Launchpad.sh: A Container-based GPU Cloud for Inference and Fine-tuning'+ 'The Top 13 AI-Powered CRM Platforms
' + 'Top 11 AI Call Center Software for 2024
' + 'Learn Prompting 101—Prompt Engineering Course
' + 'Explore Leading Cloud Providers for GPU-Powered LLM Training
' + 'Best AI Communities for Artificial Intelligence Enthusiasts
' + 'Best Workstations for Deep Learning
' + 'Best Laptops for Deep Learning
' + 'Best Machine Learning Books
' + 'Machine Learning Algorithms
' + 'Neural Networks Tutorial
' + 'Best Public Datasets for Machine Learning
' + 'Neural Network Types
' + 'NLP Tutorial
' + 'Best Data Science Books
' + 'Monte Carlo Simulation Tutorial
' + 'Recommender System Tutorial
' + 'Linear Algebra for Deep Learning Tutorial
' + 'Google Colab Introduction
' + 'Decision Trees in Machine Learning
' + 'Principal Component Analysis (PCA) Tutorial
' + 'Linear Regression from Zero to Hero
'+ '

', /* + '

Join thousands of data leaders on the AI newsletter. It’s free, we don’t spam, and we never share your email address. Keep up to date with the latest work in AI. From research to projects and ideas. If you are building an AI startup, an AI-related product, or a service, we invite you to consider becoming a sponsor.

',*/ ]; var replaceText = { '': '', '': '', '
': '
' + ctaLinks + '
', }; Object.keys(replaceText).forEach((txtorig) => { //txtorig is the key in replacetext object const txtnew = replaceText[txtorig]; //txtnew is the value of the key in replacetext object let entryFooter = document.querySelector('article .entry-footer'); if (document.querySelectorAll('.single-post').length > 0) { //console.log('Article found.'); const text = entryFooter.innerHTML; entryFooter.innerHTML = text.replace(txtorig, txtnew); } else { // console.log('Article not found.'); //removing comment 09/04/24 } }); var css = document.createElement('style'); css.type = 'text/css'; css.innerHTML = '.post-tags { display:none !important } .article-cta a { font-size: 18px; }'; document.body.appendChild(css); //Extra //This function adds some accessibility needs to the site. function addAlly() { // In this function JQuery is replaced with vanilla javascript functions const imgCont = document.querySelector('.uw-imgcont'); imgCont.setAttribute('aria-label', 'AI news, latest developments'); imgCont.title = 'AI news, latest developments'; imgCont.rel = 'noopener'; document.querySelector('.page-mobile-menu-logo a').title = 'Towards AI Home'; document.querySelector('a.social-link').rel = 'noopener'; document.querySelector('a.uw-text').rel = 'noopener'; document.querySelector('a.uw-w-branding').rel = 'noopener'; document.querySelector('.blog h2.heading').innerHTML = 'Publication'; const popupSearch = document.querySelector$('a.btn-open-popup-search'); popupSearch.setAttribute('role', 'button'); popupSearch.title = 'Search'; const searchClose = document.querySelector('a.popup-search-close'); searchClose.setAttribute('role', 'button'); searchClose.title = 'Close search page'; // document // .querySelector('a.btn-open-popup-search') // .setAttribute( // 'href', // 'https://medium.com/towards-artificial-intelligence/search' // ); } // Add external attributes to 302 sticky and editorial links function extLink() { // Sticky 302 links, this fuction opens the link we send to Medium on a new tab and adds a "noopener" rel to them var stickyLinks = document.querySelectorAll('.grid-item.sticky a'); for (var i = 0; i < stickyLinks.length; i++) { /* stickyLinks[i].setAttribute('target', '_blank'); stickyLinks[i].setAttribute('rel', 'noopener'); */ } // Editorial 302 links, same here var editLinks = document.querySelectorAll( '.grid-item.category-editorial a' ); for (var i = 0; i < editLinks.length; i++) { editLinks[i].setAttribute('target', '_blank'); editLinks[i].setAttribute('rel', 'noopener'); } } // Add current year to copyright notices document.getElementById( 'js-current-year' ).textContent = new Date().getFullYear(); // Call functions after page load extLink(); //addAlly(); setTimeout(function() { //addAlly(); //ideally we should only need to run it once ↑ }, 5000); }; function closeCookieDialog (){ document.getElementById("cookie-consent").style.display = "none"; return false; } setTimeout ( function () { closeCookieDialog(); }, 15000); console.log(`%c 🚀🚀🚀 ███ █████ ███████ █████████ ███████████ █████████████ ███████████████ ███████ ███████ ███████ ┌───────────────────────────────────────────────────────────────────┐ │ │ │ Towards AI is looking for contributors! │ │ Join us in creating awesome AI content. │ │ Let's build the future of AI together → │ │ https://towardsai.net/contribute │ │ │ └───────────────────────────────────────────────────────────────────┘ `, `background: ; color: #00adff; font-size: large`); //Remove latest category across site document.querySelectorAll('a[rel="category tag"]').forEach(function(el) { if (el.textContent.trim() === 'Latest') { // Remove the two consecutive spaces (  ) if (el.nextSibling && el.nextSibling.nodeValue.includes('\u00A0\u00A0')) { el.nextSibling.nodeValue = ''; // Remove the spaces } el.style.display = 'none'; // Hide the element } }); // Add cross-domain measurement, anonymize IPs 'use strict'; //var ga = gtag; ga('config', 'G-9D3HKKFV1Q', 'auto', { /*'allowLinker': true,*/ 'anonymize_ip': true/*, 'linker': { 'domains': [ 'medium.com/towards-artificial-intelligence', 'datasets.towardsai.net', 'rss.towardsai.net', 'feed.towardsai.net', 'contribute.towardsai.net', 'members.towardsai.net', 'pub.towardsai.net', 'news.towardsai.net' ] } */ }); ga('send', 'pageview'); -->