<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Beautiful Violet on Medium]]></title>
        <description><![CDATA[Stories by Beautiful Violet on Medium]]></description>
        <link>https://medium.com/@guider23?source=rss-367b946a7f49------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*8fE66QOVHOr6U-0mnOFtLQ.jpeg</url>
            <title>Stories by Beautiful Violet on Medium</title>
            <link>https://medium.com/@guider23?source=rss-367b946a7f49------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Wed, 18 Nov 2026 18:51:36 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@guider23/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[Understanding Higher-Dimensional Vectors Opened My Third Eye]]></title>
            <link>https://ai.plainenglish.io/understanding-higher-dimensional-vectors-opened-my-third-eye-b03d9e368b2d?source=rss-367b946a7f49------2</link>
            <guid isPermaLink="false">https://medium.com/p/b03d9e368b2d</guid>
            <category><![CDATA[higher-dimensions]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[neural-networks]]></category>
            <dc:creator><![CDATA[Beautiful Violet]]></dc:creator>
            <pubDate>Sat, 25 Oct 2025 08:02:01 GMT</pubDate>
            <atom:updated>2025-12-09T18:16:08.231Z</atom:updated>
            <content:encoded><![CDATA[<p><strong><em>Let me know if you stuck at somewhere while reading :)</em></strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*GTFQPsx4h4S25u-r2IIpXg.jpeg" /></figure><p>Let me start by saying this: you are not capable of truly perceiving 4D or 5D dimensions. Can you guess why? Yep! Humans are limited to 2D and 3D perception — not because of imagination, but because of biology. Our eyes only capture 2D projections of the world onto the retina, and it’s the brain that reconstructs depth to create a 3D understanding of space. But that reconstruction stops at three dimensions. Even within 3D, our perception is limited by occlusion, sparsity, and perspective. Higher dimensions simply don’t appear in daily life because the physical world we move through is three-dimensional.</p><p>But math and music don’t have this limitation. They can operate in 4D, 5D, or even 1000D, because the brain can process the relationships and structures mathematically, even if we cannot visualize them. However, visualization remains a challenge. No human can draw a 512-dimensional plot on a graph.</p><p>Instead, we store such information as <strong>vectors</strong>. A vector is simply a list of numbers. For example:</p><ul><li><strong>[5, 10]</strong> is a 2D vector.</li><li><strong>[3, 5, 3]</strong> is a 3D vector.</li></ul><p>The numbers in the brackets store information. For example, image data can be converted into a 512-dimensional vector, meaning the image information is stored as 512 numbers. After this, if we want to check similarity with another image, we convert the new image into a vector and use distance formulas like cosine similarity. If the vectors are close, we can say the images are similar.</p><p>Here’s what happens in AI: imagine the first image is a dog. Somehow, the neural network understands it and creates a 512-element vector. Then, when you give it a similar dog (not the same image), the network forms a vector that is close to the first vector. You can then use distance formulas to check similarity or even visualize it using UMAP or Plotly.</p><p>In short, <strong>AI, neural networks, and deep learning revolve around higher-dimensional vectors.</strong> Understanding them gives you fundamental insights.</p><blockquote><em>“Written and verified through my own research”</em></blockquote><h3>Understanding HD Vectors</h3><p>At the simplest level, a <strong>vector</strong> is just a list of numbers that represent something:</p><pre>v = [v1, v2, v3, ..., vn]</pre><ul><li><strong>[2, 3]</strong> → 2D vector, a point on a flat surface (x=2, y=3)</li><li><strong>[2, 3, 5]</strong> → 3D vector, a point in a cube (height, width, depth)</li></ul><p>When we say <strong>dimension</strong>, it means how many numbers are needed to describe something:</p><ul><li>2D → 2 numbers → plane</li><li>3D → 3 numbers → space</li></ul><h3><strong>What Happens After 3D?</strong></h3><p>Now imagine a vector like this:</p><pre>[1.2, 1.3, 1.5, ..., 512 numbers]</pre><p>This could be a 5D vector, or even a 512D vector. You can’t visualize 5D — our brains are limited to 3D — but mathematically, it works the same way. Instead of seeing it, we <strong>compute and reason</strong> with it. All the operations you know from 2D or 3D still apply:</p><p><strong>a) Vector Addition:</strong> Add components one by one.<br> <strong>b) Scalar Multiplication:</strong> Multiply each component by a number.<br> <strong>c) Dot Product (Inner Product):</strong> Measures “alignment” between vectors. Used in machine learning for similarity.</p><p>Example:</p><pre>[1, 2, 3, 4] ⋅ [4, 0, 2, 1] = 1*4 + 2*0 + 3*2 + 4*1 = 16</pre><p>Each new dimension adds a feature that helps describe the data better. These numbers don’t always represent simple things like “height” or “color.” Instead, each number represents a <strong>hidden feature</strong> that the model learned — things like:</p><ul><li>How “dog-like” it is</li><li>How “sunny” it feels</li><li>Shape, color, emotion, or other abstract characteristics</li></ul><p>All these features are packed into a 512-length vector.</p><p>Example:</p><pre>dog.jpg → [0.1, 0.3, -0.5, ..., 0.7] (512 values)<br>cat.jpg → [0.2, 0.4, -0.3, ..., 0.6]</pre><p>We can then measure similarity using cosine similarity or distance. If vectors are close, the images are similar; if far apart, they are different.</p><h3><strong><em>Why 512 Dimensions?</em></strong></h3><p>In AI, every piece of data — image, text, sound — is converted into a <strong>vector (embedding)</strong> so computers can understand and compare it.</p><ul><li><strong>CLIP (OpenAI):</strong> Converts images or text into 512-dimensional vectors</li><li><strong>BERT:</strong> 768-dimensional vectors</li><li><strong>OpenAI text-embedding-3-large:</strong> 3072-dimensional vectors</li></ul><p>Why these numbers? They are a balance between:</p><ul><li>Enough capacity to encode complex meaning</li><li>Not too large for storage or computation</li></ul><p>Neural networks learn their own features. One neuron might respond to color, another to shape, another to “dog-ness.” By the end of training, models like CLIP or BERT capture hundreds of abstract features.</p><p>Note : The NN model convert images into number based on pixels, It mean the model doesn’t understand the image, It’s made to a state to reduce error. ( I will cover that soon! )</p><h3>Visualizing High-Dimensional Data</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*XMvYFtDyo4gqzKHZdT0y1g.png" /><figcaption>When to use PCA ? — <a href="https://towardsdatascience.com/pca-clearly-explained-how-when-why-to-use-it-and-feature-importance-a-guide-in-python-7c274582c37e/">Towards Data Science</a></figcaption></figure><p>We can’t visualize 512-dimensional vectors directly, but we can reduce them:</p><ul><li><strong>PCA:</strong> Linear, preserves global structure, fast</li><li><strong>UMAP:</strong> Non-linear, preserves local and some global structure, good for clusters</li><li><strong>t-SNE:</strong> Non-linear, excellent for cluster separation, slower and stochastic</li><li><strong>TensorBoard Embedding Projector:</strong> Upload embeddings and labels for interactive visualization</li></ul><h3>Why High-Dimensional Vectors Matter (Not a bluff)</h3><ul><li><strong>Machine Learning:</strong> Images, text, and audio often have hundreds or thousands of dimensions.</li><li><strong>Physics:</strong> States of a system can be represented in high-dimensional spaces.</li><li><strong>Data Science:</strong> Each column is a dimension; each row is a point in that space.</li></ul><p>Think of a vector as a point in as many dimensions as needed:</p><ul><li>1D → single number (e.g., house price)</li><li>2D → flat graph (price vs. square footage)</li><li>3D → cube (price vs. square footage vs. bedrooms)</li></ul><h3>Weird Properties of High Dimensions (try to understand these properties)</h3><ol><li><strong>Most vectors are nearly orthogonal:</strong> In high dimensions (like 1000D), random vectors are almost 90° apart (Orthogonal to each other). This is due to the sheer amount of space. This phenomenon is part of the <strong>curse of dimensionality. </strong>(Refer Below Video)</li><li><strong>Volume concentrates at the edges:</strong> Imagine a hypersphere. In 100D, 99% of the volume is near the surface. The center is nearly empty (Refer Below Image)</li><li><strong>Distances flatten:</strong> In very high dimensions, the nearest and farthest points are nearly the same distance from a reference point.</li></ol><p>These properties make high-dimensional vectors counterintuitive but extremely <strong>useful</strong> in AI, embeddings, and neural networks.</p><p><em>“These properties often appear visually in mathematical or AI-generated plots, such as the one below.”</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/961/1*6L3SM2FYxuixS4z0K6jgHg.png" /><figcaption>AI Generated Image for formula representation for better understanding.</figcaption></figure><ul><li><strong>Recommendation Engines (Netflix, YouTube):</strong> Netflix might rate movies on 50,000 different features (genres, actors, directors, mood, etc.). Your “taste” is a single vector (a point) in that 50,000-dimensional space. To find a movie you’ll like, Netflix looks for “nearby” movie vectors in that space.</li><li><strong>Natural Language Processing (AI Chatbots):</strong> To understand language, AIs map every word to a vector, often in 300 to 1,000 dimensions. The “meaning” of a word is its position. The famous example is that the vector for “King” — “Man” + “Woman” ends up very close to the vector for “Queen”.</li></ul><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.youtube.com%2Fembed%2Fk4CxJLXc3-0%3Ffeature%3Doembed&amp;display_name=YouTube&amp;url=https%3A%2F%2Fwww.youtube.com%2Fwatch%3Fv%3Dk4CxJLXc3-0&amp;image=https%3A%2F%2Fi.ytimg.com%2Fvi%2Fk4CxJLXc3-0%2Fhqdefault.jpg&amp;type=text%2Fhtml&amp;schema=youtube" width="854" height="480" frameborder="0" scrolling="no"><a href="https://medium.com/media/26da720f882446fc767d828602f4a54c/href">https://medium.com/media/26da720f882446fc767d828602f4a54c/href</a></iframe><h3>Takeaways</h3><p>Higher-dimensional vectors are fundamental to AI, neural networks, deep learning, physics, and data science. They allow computers to represent complex data efficiently. Understanding these concepts opens a “third eye” into how AI perceives and reasons about information.</p><p>Next time, we’ll explore <strong>how databases like Pinecone handle 512-dimensional vectors</strong> and whether the curse of dimensionality applies. Stay tuned!</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=b03d9e368b2d" width="1" height="1" alt=""><hr><p><a href="https://ai.plainenglish.io/understanding-higher-dimensional-vectors-opened-my-third-eye-b03d9e368b2d">Understanding Higher-Dimensional Vectors Opened My Third Eye</a> was originally published in <a href="https://ai.plainenglish.io/">Artificial Intelligence in Plain English</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Building a Blockchain is easy as a fuk !]]></title>
            <link>https://blog.blockmagnates.com/building-a-decentralized-image-sharing-system-with-ipfs-and-blockchain-5a6bc3ef4493?source=rss-367b946a7f49------2</link>
            <guid isPermaLink="false">https://medium.com/p/5a6bc3ef4493</guid>
            <category><![CDATA[ipfs]]></category>
            <category><![CDATA[smart-contracts]]></category>
            <category><![CDATA[blockchain]]></category>
            <dc:creator><![CDATA[Beautiful Violet]]></dc:creator>
            <pubDate>Sun, 23 Feb 2025 15:28:15 GMT</pubDate>
            <atom:updated>2025-10-25T08:16:59.234Z</atom:updated>
            <content:encoded><![CDATA[<h3>Building a Decentralized Image Sharing System with IPFS and Blockchain</h3><h3>Introduction</h3><p>Decentralization is the future of data storage and sharing. In this blog, we will explore how to build a decentralized image-sharing system using <strong>IPFS (InterPlanetary File System) for storage</strong> and <strong>Ethereum smart contracts for ownership tracking</strong>. We will also go through the mistakes made during the process and how they were fixed <a href="https://github.com/guider23/blockChain-imageSharing">[Get Files]</a></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*k4NFnCTb20Rf02c9_SEJBQ.png" /></figure><h3>What I Built</h3><p>We created a smart contract that:</p><ul><li>Allows users to <strong>upload an image to IPFS</strong> and store its hash on the blockchain.</li><li>Tracks <strong>image ownership</strong> by linking an Ethereum address to each image hash.</li><li>Provides a function to <strong>retrieve all stored images</strong> along with their owners.</li></ul><p>By the end of this guide, you’ll understand how to integrate IPFS with Ethereum smart contracts and interact with them using Hardhat and Ethers.js.</p><h3>1. Setting Up the Project</h3><h3>Install Dependencies</h3><p>First, install Hardhat and the required packages:</p><pre>mkdir blockchain-image-sharing<br>cd blockchain-image-sharing<br>npm init -y<br>npm install --save-dev hardhat @nomicfoundation/hardhat-toolbox ethers dotenv</pre><h3>Initialize Hardhat</h3><p>Run Hardhat setup:</p><pre>npx hardhat</pre><p>Select <strong>“Create a basic sample project”</strong> and follow the prompts.</p><h3>Install IPFS</h3><p>We will use <strong>local IPFS</strong> for testing:</p><pre>npm install ipfs-http-client</pre><p>Start a local IPFS node by running:</p><pre>ipfs daemon</pre><h3>2. Writing the Smart Contract</h3><p>Create a new file contracts/ImageStorage.sol and add the following code:</p><pre>// SPDX-License-Identifier: MIT<br>pragma solidity ^0.8.19;</pre><pre>contract ImageStorage {<br>    mapping(string =&gt; address) public imageOwners;<br>    string[] public imageHashes;</pre><pre>    event ImageUploaded(string ipfsHash, address owner);</pre><pre>    function uploadFile(string memory _hash) public {<br>        require(imageOwners[_hash] == address(0), &quot;Image already uploaded!&quot;);<br>        imageOwners[_hash] = msg.sender;<br>        imageHashes.push(_hash);<br>        emit ImageUploaded(_hash, msg.sender);<br>    }</pre><pre>    function getImageOwner(string memory _hash) public view returns (address) {<br>        return imageOwners[_hash];<br>    }</pre><pre>    function getImages() public view returns (string[] memory) {<br>        return imageHashes;<br>    }<br>}</pre><p>This contract does the following:</p><ol><li><strong>Stores IPFS hashes</strong> mapped to Ethereum addresses.</li><li><strong>Prevents duplicate uploads.</strong></li><li><strong>Emits an event when an image is uploaded.</strong></li><li><strong>Retrieves all stored images and owners.</strong></li></ol><h3>3. Deploying the Smart Contract</h3><p>Create a scripts/deploy.js file:</p><pre>const hre = require(&quot;hardhat&quot;);<br>async function main() {<br>  const ImageStorage = await hre.ethers.getContractFactory(&quot;ImageStorage&quot;);<br>  const contract = await ImageStorage.deploy();</pre><pre>  await contract.waitForDeployment();<br>  console.log(&quot;Contract deployed to:&quot;, await contract.getAddress());<br>}</pre><pre>main().catch((error) =&gt; {<br>  console.error(error);<br>  process.exit(1);<br>});</pre><p>Run the Hardhat local blockchain:</p><pre>npx hardhat node</pre><p>Deploy the contract:</p><pre>npx hardhat run scripts/deploy.js --network localhost</pre><p>You should see an output like:</p><pre>Contract deployed to: 0x5FbDB2315678afecb367f032d93F642f64180aa3</pre><h3>4. Interacting with the Smart Contract</h3><h3>Open Hardhat Console</h3><pre>npx hardhat console --network localhost</pre><h3>Connect to the Contract</h3><pre>const contract = await ethers.getContractAt(&quot;ImageStorage&quot;, &quot;0x5FbDB2315678afecb367f032d93F642f64180aa3&quot;);</pre><h3>Upload an Image Hash</h3><pre>await contract.uploadFile(&quot;QmfJwqRTWPp4Ag4j2tzcUWBcYZZ3JHoEdzpmfFyRx1aZ4a&quot;);</pre><h3>Retrieve Uploaded Images</h3><pre>const images = await contract.getImages();<br>console.log(images);</pre><p>Expected Output:</p><pre>[&#39;QmfJwqRTWPp4Ag4j2tzcUWBcYZZ3JHoEdzpmfFyRx1aZ4a&#39;]</pre><h3>Get Image Owner</h3><pre>const owner = await contract.getImageOwner(&quot;QmfJwqRTWPp4Ag4j2tzcUWBcYZZ3JHoEdzpmfFyRx1aZ4a&quot;);<br>console.log(&quot;Owner Address:&quot;, owner);</pre><p>Expected Output:</p><pre><br>Owner Address: 0xf39Fd6e51aad88F6F4ce6aB8827279cffFb92266</pre><pre>const imageList = await contract.getImages();<br>console.log(imageList);<br><br>//Get a array of stored hashes</pre><h3>5. Common Mistakes and Fixes</h3><h3>1. Function Not Found Error</h3><p><strong>Issue:</strong></p><pre>Uncaught TypeError: contract.getFiles is not a function</pre><p><strong>Fix:</strong> Ensure the function name matches exactly as defined in Solidity (getImages instead of getFiles).</p><h3>2. “Contract Not Defined” Error</h3><p><strong>Issue:</strong></p><pre>Uncaught ReferenceError: contract is not defined</pre><p><strong>Fix:</strong> Ensure you have correctly initialized the contract instance before calling functions.</p><h3>3. Duplicate Upload Prevention</h3><p><strong>Issue:</strong> Uploading the same image hash again gives an error. <strong>Fix:</strong> This is intentional to prevent duplicate storage on the blockchain.</p><h3>6. Next Steps</h3><p>Now that we have a working decentralized image-sharing system, we can:</p><ul><li><strong>Build a Frontend</strong> using React.js to interact with the contract.</li><li><strong>Implement User Authentication</strong> using MetaMask.</li><li><strong>Deploy the Contract to a Public Testnet</strong> like Goerli or Sepolia.</li></ul><p>Stay tuned for the next part where we integrate this with a frontend!</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=5a6bc3ef4493" width="1" height="1" alt=""><hr><p><a href="https://blog.blockmagnates.com/building-a-decentralized-image-sharing-system-with-ipfs-and-blockchain-5a6bc3ef4493">Building a Blockchain is easy as a fuk !</a> was originally published in <a href="https://blog.blockmagnates.com/">Block Magnates</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
    </channel>
</rss>