illustria.io

illustriaio

Authors: (1) Renato P. dos Santos, CIAGE – Centre for Generative Artificial Intelligence in Cognition and Education. Table of Links Abstract and Introduction Materials And Methods Results and Analyses Prompts and generated texts Conceptualizing chemical reactions Deepening on understanding of chemical reactions Question about combustion Question about a graph of gases turning into water over time Question about the difference between atoms, molecules, and moles Deepening on the concept of mole Question about changing of state Question about an animated representation of water molecules undergoing phase changes Question about plasma, a state of matter Question about chemical bondings Question about illustration of chemical bonds Question about the essence of the type of chemical bonding Further analysis Conclusions Limitations of the study and possible future studies Author Contributions, Conflicts of interest, Acknowledgements, and References Question about illustration of chemical bonds *: can you create an illustration explaining these types of chemical bonds?* P1 *: Certainly! Let's break down the explanation and illustration into three parts:* ChatGPT 1. Covalent Bonding 2. Ionic Bonding 3. Hydrogen Bonding 1. Covalent Bonding: *: Covalent bonds are formed when two atoms share electrons. The shared electrons are attracted to the nuclei of both atoms, holding the atoms together. This type of bond is common between non-metals.* Description *: Imagine two circles (representing atoms) with electrons (smaller circles or dots) around them. When they come close, they share some of these electrons.* Illustration ***2. Ionic Bonding: *** *: Ionic bonds are formed when one atom donates one or more electrons to another atom. This creates a positive ion (cation) and a negative ion (anion). The opposite charges attract each other, holding the atoms together. This type of bond is common between metals and non-metals.* Description *: Imagine two circles (representing atoms). One atom has an extra electron which it donates to the other. This creates a "+" sign on the donating atom (cation) and a "-" sign on the receiving atom (anion). They are attracted due to opposite charges.* Illustration ***3. Hydrogen Bonding: *** *: Hydrogen bonds are a type of weak bond that occurs when a hydrogen atom* Description bonded to a strongly electronegative atom (like oxygen or nitrogen) is attracted to another electronegative atom. They are especially important in water molecules and in the structure of DNA. *: Imagine a water molecule (H2O) with a hydrogen atom (H) and an oxygen atom (O). The hydrogen atom in one water molecule is attracted to the oxygen atom in a neighboring water molecule. This attraction is represented by a dotted line, indicating the hydrogen bond. Now, let's create a visual representation for each of these bond types.* Illustration * Here's the visual representation of the three types of bonds: 1. Covalent Bonding: Atoms A and B are sharing electrons, represented by the blue dots between them. The red line indicates the covalent bond formed by this sharing. 2. Ionic Bonding: Atom A has donated an electron to Atom B. As a result, Atom A becomes positively charged (A+) and Atom B becomes negatively charged (B-). The dashed red line indicates the attraction between these oppositely charged ions. 3. Hydrogen Bonding: The hydrogen atom (H) bonded to the oxygen atom (O) is attracted to another oxygen atom. The solid red line indicates the covalent bond within the molecule, while the dashed green line represents the weaker hydrogen bond. I hope this helps clarify the differences between covalent, ionic, and hydrogen bonds! BingChat: * *: Sure. Here is an illustration explaining the different types of chemical bonds:* Bard Ionic bonds ionic bond between a sodium atom and a chlorine atom * In an ionic bond, one atom transfers one or more electrons to another atom. This results in two oppositely charged ions, which are attracted to each other by electrostatic forces. In the image above, the sodium atom transfers one electron to the chlorine atom. The sodium atom becomes a positively charged ion (Na+), and the chlorine atom becomes a negatively charged ion (Cl-). The Na+ and Cl- ions are attracted to each other by electrostatic forces, and they form an ionic bond. Covalent bonds * covalent bond between two hydrogen atoms In a covalent bond, two atoms share one or more pairs of electrons. Covalent bonds are typically formed between nonmetals. In the image above, the two hydrogen atoms share a pair of electrons. The two hydrogen atoms are covalently bonded together. Hydrogen bonds * hydrogen bond between two water molecules A hydrogen bond is a special type of weak bond that forms between a hydrogen atom and a highly electronegative atom, such as oxygen or nitrogen. Hydrogen bonds are responsible for many of the unique properties of water, such as its high boiling point and surface tension. In the image above, the hydrogen atom in one water molecule is attracted to the oxygen atom in another water molecule. This hydrogen bond is weaker than a covalent bond, but it is strong enough to hold the two water molecules together. I hope this illustration helps you to understand the different types of chemical bonds. Please let me know if you have any other questions. *: Here is an illustrated explanation of the three main types of chemical bonds - ionic,covalent, and hydrogen bonds:* Claude * In this illustration: • Ionic bonding is shown with the full electron transfer between the sodium and chlorine atoms, forming charged ions that attract. • Covalent bonding is shown with the shared pair of electrons between two hydrogen atoms. • Hydrogen bonding is shown with the polarity of the water molecule, leading to electrostatic attraction between molecules. Let me know if a visual explanation like this helps differentiate the types of bonds! I can create more illustrated examples to demonstrate chemical concepts. Images and diagrams can sometimes bring these abstract ideas to life. starts with the usual “Certainly!,” followed by "Let's break down the explanation," which demonstrates . After a detailed description of what the visualisation should comprise, repeatedly asking the student to “Imagine” something, it generated a quite crude “visual representation of the three types of bonds”, finalising with a “I hope this help clarify”. solely reliance on four high-quality, yet generic, pictures of chemical bonds not only did the images fail to employ effective , but they also did not promote genuine , failing to deeply engage users in the topic. The content also lacked in , as it missed drawing on parallels or real-world examples. Furthermore, the images also missed adding a , making the learning experience feel less tailored and intimate. used , by providing a comprehensive explanation, accompanied by three specific images sourced from the internet, each distinctly depicting a type of bond, adding . Concluding its explanation, added a touch of , expressing, “I hope this illustration helps you” and prompting further engagement with “Please let me know...” depiction was even more rudimentary, presenting bond details in textual format instead of offering a lucid visual illustration. Such a choice reflected a deficiency in and diminished , even though the final assertion “Images and diagrams can sometimes bring these abstract ideas to life.” While the closing remarks of "Let me know" and "I can create more illustrated examples" hint at a desire for , the lack of an actual tailored visual representation undermines this effort. ChatGPT Reinforcement & Feedback Instructional strategy Personalization & Personable Touch BingChat’s Instructional Strategies Engagement & Interactivity Use of Analogy & Comparative Illustrations Personalization & Personable Touch Bard Instructional strategy Detail Depth & Content Recommendation Bard Personalization & Personable Touch Claude's Instructional Strategies Engagement & Interactivity Personalization & Personable Touch This paper is under CC BY-SA 4.0 DEED license. available on arxiv

How GenAIbots Repond to Questions About Illustration of Chemical Bonds

কিভাবে GenAIbots রাসায়নিক বন্ডের ইলাস্ট্রেশন সম্পর্কে প্রশ্নের উত্তর দেয়

So reagieren GenAIbots auf Fragen zur Darstellung chemischer Bindungen

Cómo responden los GenAIbots a las preguntas sobre la ilustración de enlaces químicos

Comment les GenAIbots répondent aux questions sur l'illustration des liaisons chimiques

GenAIbots रासायनिक बंधों के चित्रण के बारे में प्रश्नों का उत्तर कैसे देते हैं

GenAIbotsが化学結合の図解に関する質問にどのように応答するか

GenAIbot이 화학 결합 예시에 대한 질문에 응답하는 방법

Como GenAIbots responde a perguntas sobre ilustração de ligações químicas

Как GenAIbots отвечают на вопросы об иллюстрации химических связей

GenAIbot'lar Kimyasal Bağların İllüstrasyonu Hakkında Sorulara Nasıl Yanıt Veriyor?

Cách GenAIbot trả lời câu hỏi về minh họa liên kết hóa học

GenAIbots 如何回答有关化学键说明的问题

For the creatives in the world interested in creating your tool to help manage your workflows on the Adobe Creative Cloud ecosystem, I understand your . pain Some of the common pain points: : ExtendScript is based on an older version of JavaScript (ECMAScript 3), which lacks many modern features and conveniences found in the latest ECMAScript versions. This can make the development process feel outdated and less efficient. Outdated JavaScript Version : Common development tools like package managers (npm, yarn), bundlers (webpack, rollup), or modern frameworks (React, Vue) are not directly compatible with the ExtendScript environment, which can limit development practices. Lack of Modern Development Tools : Unlike web development, where developers manipulate the Document Object Model (DOM) for dynamic changes, ExtendScript for Illustrator does not allow for direct DOM manipulation, leading to different approaches for UI development. No Direct DOM Manipulation : ExtendScript tools for debugging are not as advanced as those available for modern JavaScript development. There is no built-in console for logging, and the ExtendScript Toolkit debugger can be less intuitive to use compared to browsers' developer tools or integrated development environment (IDE) debuggers. Limited Debugging Tools : The older version of JavaScript used in Illustrator's scripting environment does not support promises or async/await, which are now common patterns for handling asynchronous operations in modern JavaScript. Asynchronous Code Handling : For creating user interfaces, developers must use Adobe's ScriptUI, which has a different paradigm from HTML/CSS and lacks the styling flexibility and dynamic capabilities developers are accustomed to on the web. ScriptUI Framework : Scripts running in Illustrator have access to the file system, which is different from the sandboxed environment of a web browser. This requires a different approach to file handling and can pose additional security considerations. File System Access Whereas it can take only a to get a React app up and running, creating even a basic extension within Adobe is filled with issues. few minutes So let’s get started. To start, here are 3 resources to refer to for those who have a preference for different tech stacks: Article Name Link Tech stack CEP-React-Webpack https://reactjsexample.com/an-awesome-cep-extension-boilerplate-with-excellent-development-experience/ React, Webpack, TypeScript Vue CEP Sample Extension https://github.com/Inventsable/Simple-resolution-panel Vue, Extension Manifest v 9.0 Adobe Extension Testing https://github.com/NTProductions/adobe-extension-testing/tree/master JS, Adobe After Effects You can stop reading > and take a stab at these if you’d like. If you want a more theoretical path, look forward to the next part of this series. Adobe Illustrator® is a registered trademark of Adobe Inc., used here for illustrative purposes only.

Getting Started With Software Development for Adobe Illustrator

I believe I first heard about this format a long time ago, maybe five years ago, but many junior designers I teach don't know about the possibilities of the SVG format. Let’s get into it. What is SVG? Scalable Vector Graphics (SVG) consists of circles, rectangles, and paths represented in XML and combined into images on web pages. You can apply solid fills, gradients, and some SVG filters - not all browsers support all types of filters. You can include text and images and you can copy your SVGs as much as you want. SVG is most often used in graphics programs, to create diagrams, illustrations, and animations. The pros of SVG from a website user's point of view: clear image lightweight simplicity and time-saving in small projects. Cons for development: If we use SVG code, the code itself becomes quite large as the icon itself can be very complex in terms of colors and the number of vectors inside The second problem is if we connect the icon in this way, then it is problematic for us to change its color through CSS or even JS. You will have to export a duplicate of the icon but with a different color. In the case of the object tag, the link will not work properly at all if we wrap it in the <a> tag. There is also a small disadvantage - it is related to the fact that we cannot set the icon size through the font-size property (that is, we do not inherit the font size). You have to "play" with width and height all the time . Anyway working as a designer in products and studios, I have learned that this is the most common format for transferring icons to the layout. I often use graphics format for icons, and vector images for a wide range of projects - from services to e-commerce. You can get this format in both Adobe Illustrator and . Figma Today we're gonna talk about how to use it in both cases. Export SVG in Adobe Illustrator Once the design is approved, we assemble it all into one Adobe Illustrator file. Open the Assets Export panel (Window-Asset Export). Drag and drop each item there and give it a name. Customize export parameters by clicking the button at the very bottom of the Asset Exports panel. Next, go to the settings and check All Assets (if we want everything to be highlighted) select the place where the icons will be saved, and customize SVG settings (click on the gear). Next, select SVG and customize the parameters as shown in the image below Click Save Settings, check again the path where the file is saved, and click Export Asset. This is a final step. It could be useful, especially if you're used to Adobe Illustrator or looking for an alternative to Figma for exporting SVG icons

How To Export SVG in Adobe Illustrator

Authors: (1) Davide Viviano, Department of Economics, Harvard University; (2) Lihua Lei, Graduate School of Business, Stanford University; (3) Guido Imbens, Graduate School of Business and Department of Economics, Stanford University; (4) Brian Karrer, FAIR, Meta; (5) Okke Schrijvers, Meta Central Applied Science; (6) Liang Shi, Meta Central Applied Science. Table of Links Abstract & Introduction Setup (When) should you cluster? Choosing the cluster design Empirical illustration and numerical studies Recommendations for practice References A) Notation B) Endogenous peer effects C) Proofs 5 Empirical illustration and numerical studies In this section we illustrate the properties of the procedure in two empirical applications, one using unique data from the Facebook friendship and messaging graphs, and one using data from an experiment conducted in rural China by Cai et al. [2015]. We provide further evidence supporting our theoretical analysis using simulated networks in Section 5.3. 5.1 Clustering on Facebook graphs We first study the procedure’s properties for two clustering algorithms implemented at scale on social networks owned by Meta: Louvain algorithm [Blondel et al., 2008], Balanced Partitioning [Kabiljo et al., 2017]. We consider two graphs owned by Meta. In each graph, edges are continuous variables. In the first graph, edges capture the strength of the friendship in the Facebook graph, and in the second graph they capture connections based on messaging on Facebook, in both cases the data was aggregated and de-identified. We compute our statistics by setting the edges to be zero if below the 5th, 10th, and 50th percentile. We will refer to these graphs obtained as dense, moderate (mod), and sparse after thresholding (graphs all have a bounded maximum degree). We report the bias and variance, with the variance weighted by a parameter ξn, defined respectively as Louvain and Balanced Partitioning clustering produce a hierarchical clustering structure with a growing number of clusters. We consider three “types” of such algorithms, corresponding to three levels in the clustering structure hierarchy (i.e., different numbers of clusters), defined as Type 1, 2, and 3, respectively. For Louvain clustering, higher-order types denote more clusters, whereas it is the opposite for Balanced Partitioning. For each clustering, we report log(Kn/n), log-ratio of the number of clusters over the population size. Table 2 collects each of these algorithms’ worst-case bias and variance and the number of clusters per individual. Louvain algorithm dominates Balanced Partitioning both in worstcase variance and bias. A larger number of clusters increases the bias but decreases the variance, which we observe throughout all graphs. The bias increases for sparser networks because the denominator |Ni | decreases. Figure 3 reports the objective function for different graphs, Louvain algorithms, and degrees of sparsity. For ξn ≈ 1, the Louvain algorithm with the largest number of clusters (Type 3) dominates the other two algorithms in all but one case. This result provides some interpretable comparisons for the number of clusters. We also observe a trade-off between the number of clusters and the degree of sparsity of the adjacency matrix as ξn decreases. Louvain clustering with the largest number of clusters (Type 1) is optimal for a dense graph with ξn ≫ 1. Such clustering, however, is sub-optimal as the graph becomes more sparse. This result is intuitive: for denser networks, a larger number of clusters may best control the estimator’s bias relative to the bias induced by other clustering algorithms. These comparisons motivate trade-offs in the choice of the graph (and its density) and the clustering algorithm. We recommend practitioners compare clustering by averaging objective functions over different sparsity thresholding. For example, it is possible to compare clustering based on the average objective across different graphs using some priors on the degree of sparsity without affecting our theoretical results. Finally, note that using Theorem 3.6 it is possible to compute the smallest values of the spillover effects that would motivate using a cluster instead of Bernoulli design. From Theorem 3.6, researchers should prefer a cluster over a Bernoulli design if In summary, our results shed light on using clustering algorithms for large-scale implementation on online platforms. These results suggest that Louvain clustering with possibly many clusters performs best in practice. It confirms intuitive arguments in Karrer et al. [2021] who also recommend Louvain clustering based on AA-tests. 5.2 Clustering in the field We consider as an application the problem of informing individuals to increase insurance takeup studied in Cai et al. [2015]. The authors collected network information in approximately 184 villages in rural China from 48 larger regions. We use network data collected by Cai et al. [2015] to study the properties of the proposed method, where we assume that two individuals are connected if at least one of the two indicates the other as a friend. Because we do not require information from end-line outcomes, we construct a network using information from surveyed individuals as well as their friends. The network has in total 7649 nodes, once we also include individuals who are friends of surveyed individuals (but do not necessarily live in their same village or were surveyed in the experiment). Individuals between villages present on average 50% of their connections outside their village. On the other hand, individuals in different regions (with approximately 50 regions) have 99% of their connections within the same region. We use an adjacency matrix a “weak ties” adjacency matrix, where two individuals are connected if either indicates the other has a friend. We study clustering within each region and report results for different clustering algorithms. Clustering within each region is performed as in Algorithm 1, where we first estimate Xˆ (K) via semi-definite programming, use K-means algorithm to retrieve the clusters from Xˆ (K) and iterate to estimate the optimal number of clusters K. We report the frontier with respect to the absolute bias consistently with Algorithm 1. 5.2.1 How many clusters should we choose? Figure 4 presents the main results. The left-hand side panel reports the average number of estimated clusters, estimated by the proposed method divided by the population size in a given cluster. Different choices of ξn justify a different number of clusters. For ξn = 1, the number of clusters is 7%, the total population size, whereas for ξn = 15, the number of clusters is roughly half the population size. To gain further intuition of its practical implication, take ψ¯ = 0.24 approximately equal to the outcomes’ variance in Cai et al. [2015] and ϕ¯ n = 0.27 as in Table 2 in Cai et al. [2015]. A recommended choice of ξn is approximately ψ/¯ ϕ¯2 n = 3.29. This choice puts slightly more weight on the bias than the variance when λ = 1, and therefore can be considered a conservative choice for the number of clusters.[13] Even with such a conservative choice, the suggested number of clusters is large relative to the population size. In this case, the optimal number of clusters is around 15% of the overall population size, approximately 600 clusters with a surveyed sample size of 3600 individuals. This number is consistent (in scale) with the choice in the number of clusters in other applications in development, such as Egger et al. [2022], Alatas et al. [2012], but larger than the number of villages that we have in the application in Cai et al. [2015]. 5.2.2 Comparison with other clustering methods Next, we compare the procedure (denoted as “Causal Clustering”) with the following alternative clustering algorithms: ε-net clustering as in Eckles et al. [2017] with ε = 3 as suggested in Eckles et al. [2017]; spectral clustering with a fixed number of clusters equal to n/3, where n is the population size in a given region (note that spectral clustering do not optimize over the number of clusters); Louvain clustering with default parameters selected by the R-package igraph; clustering based on village identity of each individual (creating one additional cluster for those individuals whose village is missing in the data). We contrast the objective function with alternative algorithms in the right-hand side of Figure 4. The proposed method achieves the smallest objective function across all competitors, providing suggestive evidence that Algorithm 1 (with a semi-definite relaxation) works particularly well in practice. For a smaller number of clusters the method is closer to the Louvain clustering algorithm, while for a larger number of clusters the objective of the algorithm is close to the objective of the spectral algorithm. Interestingly, clustering based on village identity leads to the largest loss function. This is because individuals are connected within and between villages in this application. Finally, in Figure 5 (first three panels), we report the mean-squared error obtained by simulating the model in Cai et al. [2015] (Table 2, Column 4) and vary the variance of the residuals σ 2 ∈ {1/4, 1/2, 1} to emulate settings with high, medium and low signal to noise ratio. The model in Cai et al. [2015] assumes that individual outcomes depend on neighbors’ treatments, consistently with our Assumption 1. Our method uniformly outperforms competitors in terms of mean-squared error, except for the Louvain algorithm when the signal-to-noise ratio is particularly high, but not in the remaining settings. In the last panel, we consider settings where the local interference assumption is violated, and we calibrate our model to the specification in Cai et al. [2015] (Column 4, Table 5), where the individual outcome depends on the neighbors’ outcomes, violating Assumption 1 and introducing global dependence. We fix the residuals’ variance to be σ 2 = 1/4, whereas the results are robust as we increase σ 2 . In this setup, we observe a larger mean-squared error of all methods, with the proposed method achieving the lowest mean-squared error. 5.3 A numerical study Next, we illustrate the properties of the method in numerical studies. We consider three different data generating processes for the network formation: a geometric network formation, Albert Barabasi network and Erdos-Renyi graph. The geometric network takes the form Ai,j = 1{|Xi,1 − Xj,1|/2 +|Xi,2 − Xj,2|/2 ≤ rn} where rn = p 4/2.75n similarly to simulations in Leung [2020]. Here, Xi,1, Xi,2 are drawn independently from a uniform distribution between [−1, 1]. For the Albert-Barabasi network we first draw n/5 edges uniformly according to Erdos-Renyi graph with probabilities p = 10/n, and second, we draw sequentially connections of the new nodes to the existing ones with probability equal to the number of connection of each pre-existing node divided by the overall number of connections. Finally, the Erdos-Renyi graph is generated with probability of connection p = 2/n. We study the properties of the proposed method and of the same competitors discussed in Section 5.2. Figure 6 showcases that the proposed procedure leads to the lowest objective function. This result is consistent with our theoretical findings. Figure 7 showcases that the number of clusters increases in the sample size, varies substantially across different data generating processes, and increases in ξn. [13] This follows from the fact that ξn is proportional to ψ/¯ ϕ¯2 n for clusters whose bias is of order O(ϕ¯ n), and researcher’s goal is to minimize the worst-case mean squared error (λ = 1). This paper is under CC 1.0 license. available on arxiv

Design of Cluster Experiments: Empirical Illustration and Numerical Studies

A Conversation with Picatrix Picori — The Manga-Style AI Illustrator

পিকাট্রিক্স পিকোরির সাথে একটি কথোপকথন — দ্য মাঙ্গা-স্টাইল এআই ইলাস্ট্রেটর

Ein Gespräch mit Picatrix Picori – dem KI-Illustrator im Manga-Stil

Una conversación con Picatrix Picori: el ilustrador de IA estilo manga

Une conversation avec Picatrix Picori — L'illustrateur IA de style manga

पिकाट्रिक्स पिकोरी के साथ बातचीत - मंगा-स्टाइल एआई इलस्ट्रेटर

マンガ風のAIイラストレーター、Picatrix Picori氏との対談

Picatrix Picori와의 대화 — 만화 스타일의 AI 일러스트레이터

Uma conversa com Picatrix Picori – o ilustrador de IA estilo mangá

Разговор с Пикатриксом Пикори — иллюстратором искусственного интеллекта в стиле манга

Manga Stili Yapay Zeka İllüstratörü Picatrix Picori ile Söyleşi

Cuộc trò chuyện với Picatrix Picori - Họa sĩ minh họa AI theo phong cách Manga

对话 Picatrix Picori — 漫画风格的 AI 插画师

The Drawings of Leonardo da Vinci by C. Lewis Hind, is part of the HackerNoon Books Series. You can jump to any chapter in this book here . Illustrations Illustrations PROFILE OF A WARRIOR (BRITISH MUSEUM) PHOTO, Autotype Company PORTRAIT OF ISABELLA D'ESTE (LOUVRE) PHOTO, BRAUN, CLÉMENT STUDY OF AN OLD MAN (MILAN) PHOTO, BRAUN, CLÉMENT STUDY OF DRAPERIES FOR KNEELING FIGURES (BRITISH MUSEUM) PHOTO, Autotype Company STUDY OF A BACCHUS (ACADEMY, VENICE) PHOTO, BRAUN, CLÉMENT HEAD OF A MAN (LOUVRE) PHOTO, BRAUN, CLÉMENT BATTLE BETWEEN HORSEMEN AND MONSTERS (MILAN) PHOTO, BRAUN, CLÉMENT WOMAN SEATED ON GROUND AND A CHILD KNEELING (MILAN) PHOTO, BRAUN, CLÉMENT STUDIES OF HEADS (UFFIZI, FLORENCE) PHOTO, BRAUN, CLÉMENT YOUTH ON HORSEBACK (WINDSOR) PHOTO, BRAUN, CLÉMENT STUDIES FOR THE EQUESTRIAN STATUE OF FRANCESCO SFORZA (WINDSOR) PHOTO, BRAUN, CLÉMENT THE VIRGIN, ST. ANNE AND INFANT (LOUVRE) PHOTO, BRAUN, CLÉMENT STUDIES OF CHILDREN (CHANTILLY) PHOTO, BRAUN, CLÉMENT THE COMBAT (LOUVRE) PHOTO, BRAUN, CLÉMENT STUDY FOR A MADONNA (UFFIZI, FLORENCE) PHOTO, BRAUN, CLÉMENT STUDIES FOR "THE HOLY FAMILY" (WINDSOR) PHOTO, BRAUN, CLÉMENT STUDY FOR "THE LAST SUPPER" (ACADEMY, VENICE) PHOTO, BRAUN, CLÉMENT COURTYARD OF A CANNON-FOUNDRY (WINDSOR) PHOTO, BRAUN, CLÉMENT STUDY OF THE HEAD OF AN APOSTLE (WINDSOR) PHOTO, BRAUN, CLÉMENT STUDY FOR BACKGROUND OF "THE ADORATION OF THE MAGI" (UFFIZI, FLORENCE) PHOTO, BRAUN, CLÉMENT STUDY OF LANDSCAPE DATED 1473. (UFFIZI, FLORENCE) PHOTO, BRAUN, CLÉMENT STUDY OE A TREE (WINDSOR) PHOTO, BRAUN, CLÉMENT TWO HEADS. CARICATURES (WINDSOR) PHOTO, BRAUN, CLÉMENT ST. JOHN THE BAPTIST (WINDSOR) PHOTO, BRAUN, CLÉMENT THE HEAD OF CHRIST (BRERA, MILAN) PHOTO, BRAUN, CLÉMENT CARICATURES (WINDSOR) PHOTO, BRAUN, CLÉMENT HEAD OF AN ANGEL (TURIN) PHOTO, BRAUN, CLÉMENT STUDY OF A MAN'S HEAD (BRITISH MUSEUM) PHOTO, BRAUN, CLÉMENT STUDIES OF HANDS (WINDSOR) PHOTO, BRAUN, CLÉMENT DRAGON FIGHTING WITH A LION (UFFIZI, FLORENCE) PHOTO, BRAUN, CLÉMENT MAN KNEELING (WINDSOR) PHOTO, BRAUN, CLÉMENT PORTRAIT STUDY (MILAN) PHOTO, BRAUN, CLÉMENT STUDIES OF ANIMALS (WINDSOR) PHOTO, BRAUN, CLÉMENT PORTRAIT OF LEONARDO, BY HIMSELF (TURIN) PHOTO, ANDERSON SIX HEADS OF MEN AND A BUST OF A WOMAN CARICATURES (VENICE) PHOTO, BRAUN, CLÉMENT STUDY OF A HEAD (UFFIZI, FLORENCE) PHOTO, BRAUN, CLÉMENT THE SAINT ANNE CARTOON (BURLINGTON HOUSE) PHOTO, HOLLYER HEADS OF A WOMAN AND A CHILD (CHATSWORTH) PHOTO, BRAUN, CLÉMENT STUDY OF DRAPERY FOR A KNEELING FIGURE (WINDSOR) PHOTO, BRAUN, CLÉMENT KNIGHT IN ARMOUR (MILAN) PHOTO, BRAUN, CLÉMENT STUDY OF A YOUTHFUL HEAD (BRITISH MUSEUM) PHOTO, Autotype Company STUDY FOR "LEDA" (CHATSWORTH) PHOTO, BRAUN, CLÉMENT HEAD OF AN OLD MAN (WINDSOR) PHOTO, BRAUN, CLÉMENT STUDY OF A HEAD (WINDSOR) PHOTO, BRAUN, CLÉMENT STUDY OF THE HEAD OF ST. PHILIP FOR THE LAST SUPPER (WINDSOR) PHOTO, BRAUN, CLÉMENT STUDY OF DRAPERY (LOUVRE) PHOTO, BRAUN, CLÉMENT GIRL'S HEAD (MILAN) PHOTO, BRAUN, CLÉMENT About HackerNoon Book Series: We bring you the most important technical, scientific, and insightful public domain books. This book is part of the public domain. C. Lewis Hind (2018). The Drawings of Leonardo da Vinci. Urbana, Illinois: Project Gutenberg. Retrieved https://www.gutenberg.org/cache/epub/56844/pg56844-images.html This eBook is for the use of anyone anywhere at no cost and with almost no restrictions whatsoever. You may copy it, give it away or re-use it under the terms of the Project Gutenberg License included with this eBook or online at www.gutenberg.org , located at https://www.gutenberg.org/policy/license.html .

The Illustrations of Leonardo Da Vinci's drawing

Oftentimes, the marketing team handles the illustrations found on a company’s website. These images are typically created by graphic or brand designers, and they focus solely on marketing objectives. However, product designers often view these pictures as decorative add-ons. They may choose not to use them at all or leave areas within the design labeled as "picture to be inserted here." Nevertheless, incorporating graphics can greatly enhance the user experience while also solving various interface design challenges. In this article, I will share examples of how I, as a product designer, have utilized illustrations in my work. Additionally, I will provide tips on how to quickly learn to draw suitable images, even if you lack prior experience. How illustrations improve the user experience 1. Visualization of site / article / page content The title picture of the page can reveal the whole essence of the content below. Not only to show the user what this page is about in general, but also to describe in detail what the company does or what the application does. 2. An illustration of “how it works” or a step-by-step process Adding pictures to the "how it works" blocks can improve readability and speed up comprehension. Visualizing textual information can assist users in understanding complex step-by-step processes more quickly, particularly when dealing with intricate schemes. 3. Product visualization It often happens that there are no photos of the product, or they are too poor quality. In this case, you can draw illustrations. Even schematic contour pictures can help to show the product and remove a lot of questions from the user. If we are talking about a lot of similar products, illustrations will help to distinguish them from each other in the list and show their similarities/differences. 4. Visualization of the scheme of a complex process Instead of a complex flowchart of a process (for example, production), you can draw an illustration where all the stages and the relationship between them are schematically shown. This makes the scheme clearer for a user who does not even have specialized knowledge. 5. Error screen improvement You can add illustrations on the error screens "nothing was found", "something went wrong" etc. They not only fill the empty space, showing that the page has loaded completely but also mitigate the frustration of the errors themselves. 6. Also the usual use of illustrations Naturally, do not forget about the usual functions of images. They are needed for landing pages to illustrate text. Also for onboarding mobile applications to describe the basic functions, as well as for instructions on the use of physical objects. Techniques for quickly creating illustrations It would seem that when vector illustrations are needed for a website /application /presentation, you can go and buy them on the stock. But in real life, in real cases, this usually does not work out. You have to spend hours surfing on ready-made pictures, and eventually compromise: take not what you need to accurately convey the meaning, but what is there. Therefore, it is much more reliable to draw pictures yourself, even if they are not perfect. Here I will tell you how to do it quickly. 1. Outline illustrations based on photos The simplest method that is available to any confident user of the vector editor. You just need to take a suitable photo and outline the objects. You can assemble a picture from several photos and get a complete composition. You can fill individual shapes with color, as well as use variable line thickness to highlight individual objects. Something can be completed manually. Some necessary elements that are not in the photos. The main advantage of such a technique is that it is easy to master and scale. You do not need to be able to draw, understand the anatomy and rules of composition, and spend a lot of time mastering it. At the same time, you can easily convey all the necessary information in the illustrations. In addition, you can make a long series of images in a single style, which can be difficult when using stock illustrations from different authors. The disadvantage of the approach is "boring" unemotional pictures that convey meaning, but do not cause a lot of emotions and are poorly remembered. The technique is ideal as an entry point into the field of creating vector illustrations for those who have never done it. And also a way out of the situation when there is no illustrator in the team, and there are no suitable pictures on the stock 2. Flat vector illustrations As you master vector graphics, you can gradually move to the next level. Start simplifying the contours obtained from the photos, give them a unique shape, use color, and, most importantly, volume. Gradually, you can start creating your own objects without any references. Such images make it possible to convey the meaning much better - to highlight the necessary objects, and remove unnecessary ones, create unique scenes that cannot be found in photos. In addition, the pictures are much more vivid and emotional. When mastering the technique, a unique style inevitably arises, even if initially you will imitate other artists' forms and approaches. Therefore, the images will also become unique, which is always useful for prominence a website or application from competitors. Drawing such pictures will already take a lot of time. Both for mastering the technique and for creating the pictures themselves. It will take a long time to pull out all the vector shapes, and colors and add volume using shadows and highlights. However, if you fill your hand on outline pictures before that, it will be much easier to master this technique. You can create contours in an illustrator, and add volume in Photoshop. With the help of noise in shadows and glare, you can take pictures even more attractive, close to real drawing techniques on paper or canvas. 2. Isometric illustrations The next step is the transition to isometric, "pseudo-3D" images. This technique is the most difficult to master, but it gives the best result - 3D illustrations that allow you to convey dynamics and create more complex compositions. You can put many objects in one illustration and show various relationships between them. In theory, everything is very simple - you need to build guides in three dimensions and start drawing with geometric primitives, gradually complicating. The classic angle between the horizontal guides is 120 degrees, but you can also change it a little, depending on the scene. You can start with the pseudo "low-poly" technique, where all objects will be quite primitive, and consist of cubes and balls. It's both faster and sometimes looks more interesting. For the first sketches, you can use photos or pictures from stock - to pick up the composition and understand the shape of different objects (cars, people, furniture, etc.). Creating detailed rough scenes greatly speeds up the work, where the objects will already be correctly positioned and all the details are thought out. These sketches can be approved by the customer - and then start drawing the final illustration. With further development, it is already possible to move away from classical isometry and add perspective - to make the scenes even more dramatic and voluminous. For example, as in this illustration: Thanks for your attention. In the following articles, I will talk about how I apply my drawing skills in product design. For example, to draw tables that will not only be convenient, but also beautiful.

Why a Product Designer Needs Illustrations, and How to Draw Them Quickly

The illustrious mathematician

Optimizing workflows with AI has become a game-changer for me, especially when it comes to searching for information. For a while, I relied on the free version of Perplexity, but recently, I upgraded to Perplexity Pro. Meanwhile, as a longtime user of ChatGPT, I’ve also been exploring its new ChatGPT Search feature. In this article, I’ll share my experiences with both tools, compare their strengths and weaknesses, and help you decide which one might be the better fit for your needs. Perplexity I’d heard about Perplexity.ai many times, promised myself I’d try it, and eventually did, but it didn’t quite click. It’s hard to fully appreciate a tool until you use it regularly. I forgot about it until I got a test assignment for a job interview in a new field, requiring tons of research. Honestly, I’m not great at googling. But since I frequently use ChatGPT and have gotten pretty good at writing prompts, I thought of giving Perplexity another shot. This was before the release of ChatGPT Search, and at that point, I’d already maxed out what ChatGPT could offer. Perplexity turned out to be a lifesaver. It made long, tedious searches much quicker and was my first truly game-changing experience with AI search. At the time, I used the free version. Compared to browser searches, here’s what I found more convenient in AI search: Fewer distractions from ads: Most browsers put ads at the top, so you have to scroll past them just to get to the first actual search result. Even without the ads, the top links are often not very relevant, and there are still too many to sift through. This makes it easy to lose motivation to click beyond the first couple of pages. AI does all that for you, providing a list of sources you can explore further on your own.


Threaded searches with context: I loved how searches are organized into threads, preserving context. For my test assignment, since I was learning a new field from scratch, one question would lead to many others. It was convenient to continue and expand the search without starting over each time. If I remembered something later, I could easily return to a specific thread and pick up where I left off.


Helpful question suggestions: Each thread in Perplexity includes prompts for related questions, which makes navigating a new topic much easier.


Incognito mode: Like most browsers, Perplexity offers an incognito mode. I haven’t used it yet, but it’s a nice feature.


Language flexibility: If you write in a certain language, you’ll get sources in that language. This is a standard approach for browsers but not always convenient. For a while, I continued using the free version, but recently I switched to Pro. The free version uses the GPT-3.5 model, whereas the Pro plan grants access to the latest models: GPT-4 Omni, Claude-3 Sonnet/Opus, and Sonar Large 32k, so all the best features are integrated into your search. Of course, the Pro version offers better, smarter, and faster search. In addition to the features mentioned above, here’s what I found most useful: Broaden your search with images: Pro provides access to Playground V2.5, DALLE 3, and Stable Diffusion XL models that can expand your search. For example, I was recently researching fund flow modeling, and having visual examples was incredibly helpful.


Files, code, and more input options: You can upload unlimited files, code, pictures, etc. It’s really fast and much more convenient than doing this through a standard browser because it’s all in one space with seamless AI interaction. ChatGPT Search This feature was launched in October 2024 and is based on GPT-4o, specifically trained for search. It’s integrated into ChatGPT, allowing you to enable it with a Search button or simply ask ChatGPT to search the web. As I mentioned, I’ve been a long-time user of ChatGPT for several years, so I seamlessly integrated the search feature into my routine. Since free Perplexity uses GPT-3.5, all its features like AI search with no ads, no endless scrolling, and results organized into threads with helpful tips are available here as well. My favorite feature here is its accessibility, offering text-to-speech, and the ability to create queries via audio, and audio dialogs. Choose your fighter Search Quality Perplexity excels at finding information, while ChatGPT Search is better at explaining it. From a user perspective, I wouldn’t say I’ve encountered any critical issues with not finding something. It’s just a matter of different approaches. Perplexity provides more an overview of sources, whereas ChatGPT tends to give more explanation than summary. Generated Answers Without Sources Both tools provide sources for their answers, but sometimes the information might be generated without a reliable link. Even when a link is provided, the relevant information might not actually be there. It’s always better to double-check and read the sources yourself. Based on my experience, ChatGPT is more prone to generating unsupported content than Perplexity, but I’ve encountered this issue in both tools. Search Speed I didn’t notice a significant difference between the two in terms of speed. However, both are slower than a standard browser search. Chrome extensions Both tools offer Chrome extensions, and I tested both. The ChatGPT extension lets you use ChatGPT as your primary search engine, replacing, for example, Google. When you enter something in the search bar, it redirects you to ChatGPT, where your query is processed. Honestly, I wasn’t a fan. It’s slower than Google, and you can feel it. I still prefer a traditional search bar for quick lookups. Perplexity’s Chrome extension works differently. Instead of replacing your search engine, it opens a small window over the current page, allowing you to googling and query Perplexity simultaneously. I really loved this one, because it’s more useful for me. The drawback is that you can only enter one query at a time, and if you want to continue the conversation, it redirects you to a separate Perplexity tab. Unique Features Since both tools use GPT models, they share many great features. Perplexity Pro stands out with its use of multiple models and built-in illustrations. On the other hand, I really appreciate ChatGPT’s accessibility and seamless integration into its apps, which I use for more than just search. Bonus: Hangman game test I saw a funny tweet claiming ChatGPT can’t play Hangman. Turns out, neither can Perplexity Pro, even with all its models. Final Take: Which Works Best for What To start with, I’m a fan of AI search. Usually, it takes longer than quick googling, so here’s how I see it: if I need a super quick answer from the first link, I’ll use a standard browser search. But if I need to find more information or conduct deeper research, AI search is the way to go. It’s the best method to uncover valuable insights more efficiently. As for ChatGPT Search and Perplexity, in my experience, they solve different problems. ChatGPT Search perfectly extends ChatGPT’s capabilities, given that ChatGPT’s knowledge base only goes up to the year 2021. It’s perfect for studying and gaining insights using up-to-date web information. Perplexity, on the other hand, is better for finding something more thoroughly. It takes more from browser search, like incognito pages, and its approach to work with your prompts. It is about searching for information first and explaining second, in contrast to ChatGPT Search. For me: Need to find something? Perplexity.
Need to understand something? ChatGPT + Search. They complement each other perfectly in work. Optimizing workflows with AI has become a game-changer for me, especially when it comes to searching for information. For a while, I relied on the free version of Perplexity , but recently, I upgraded to Perplexity Pro . Meanwhile, as a longtime user of ChatGPT , I’ve also been exploring its new ChatGPT Search feature. In this article, I’ll share my experiences with both tools, compare their strengths and weaknesses, and help you decide which one might be the better fit for your needs. Perplexity Perplexity Pro ChatGPT ChatGPT Search Perplexity I’d heard about Perplexity.ai many times, promised myself I’d try it, and eventually did, but it didn’t quite click. It’s hard to fully appreciate a tool until you use it regularly. I forgot about it until I got a test assignment for a job interview in a new field, requiring tons of research. Honestly, I’m not great at googling. But since I frequently use ChatGPT and have gotten pretty good at writing prompts, I thought of giving Perplexity another shot. This was before the release of ChatGPT Search, and at that point, I’d already maxed out what ChatGPT could offer. Perplexity turned out to be a lifesaver. It made long, tedious searches much quicker and was my first truly game-changing experience with AI search. Perplexity.ai At the time, I used the free version. Compared to browser searches, here’s what I found more convenient in AI search: Fewer distractions from ads: Most browsers put ads at the top, so you have to scroll past them just to get to the first actual search result. Even without the ads, the top links are often not very relevant, and there are still too many to sift through. This makes it easy to lose motivation to click beyond the first couple of pages. AI does all that for you, providing a list of sources you can explore further on your own. Threaded searches with context: I loved how searches are organized into threads, preserving context. For my test assignment, since I was learning a new field from scratch, one question would lead to many others. It was convenient to continue and expand the search without starting over each time. If I remembered something later, I could easily return to a specific thread and pick up where I left off. Helpful question suggestions: Each thread in Perplexity includes prompts for related questions, which makes navigating a new topic much easier. Incognito mode: Like most browsers, Perplexity offers an incognito mode. I haven’t used it yet, but it’s a nice feature. Language flexibility: If you write in a certain language, you’ll get sources in that language. This is a standard approach for browsers but not always convenient. Fewer distractions from ads: Most browsers put ads at the top, so you have to scroll past them just to get to the first actual search result. Even without the ads, the top links are often not very relevant, and there are still too many to sift through. This makes it easy to lose motivation to click beyond the first couple of pages. AI does all that for you, providing a list of sources you can explore further on your own. Fewer distractions from ads: Most browsers put ads at the top, so you have to scroll past them just to get to the first actual search result. Even without the ads, the top links are often not very relevant, and there are still too many to sift through. This makes it easy to lose motivation to click beyond the first couple of pages. AI does all that for you, providing a list of sources you can explore further on your own. Fewer distractions from ads: Threaded searches with context: I loved how searches are organized into threads, preserving context. For my test assignment, since I was learning a new field from scratch, one question would lead to many others. It was convenient to continue and expand the search without starting over each time. If I remembered something later, I could easily return to a specific thread and pick up where I left off. Threaded searches with context: I loved how searches are organized into threads, preserving context. For my test assignment, since I was learning a new field from scratch, one question would lead to many others. It was convenient to continue and expand the search without starting over each time. If I remembered something later, I could easily return to a specific thread and pick up where I left off. Threaded searches with context: Helpful question suggestions: Each thread in Perplexity includes prompts for related questions, which makes navigating a new topic much easier. Helpful question suggestions: Each thread in Perplexity includes prompts for related questions, which makes navigating a new topic much easier. Helpful question suggestions: Incognito mode: Like most browsers, Perplexity offers an incognito mode. I haven’t used it yet, but it’s a nice feature. Incognito mode: Like most browsers, Perplexity offers an incognito mode. I haven’t used it yet, but it’s a nice feature. Incognito mode: Language flexibility: If you write in a certain language, you’ll get sources in that language. This is a standard approach for browsers but not always convenient. Language flexibility: If you write in a certain language, you’ll get sources in that language. This is a standard approach for browsers but not always convenient. Language flexibility: For a while, I continued using the free version, but recently I switched to Pro . The free version uses the GPT-3.5 model, whereas the Pro plan grants access to the latest models: GPT-4 Omni, Claude-3 Sonnet/Opus, and Sonar Large 32k, so all the best features are integrated into your search. Of course, the Pro version offers better, smarter, and faster search. Pro In addition to the features mentioned above, here’s what I found most useful: In addition to the features mentioned above, here’s what I found most useful: Broaden your search with images: Pro provides access to Playground V2.5, DALLE 3, and Stable Diffusion XL models that can expand your search. For example, I was recently researching fund flow modeling, and having visual examples was incredibly helpful. Files, code, and more input options: You can upload unlimited files, code, pictures, etc. It’s really fast and much more convenient than doing this through a standard browser because it’s all in one space with seamless AI interaction. Broaden your search with images: Pro provides access to Playground V2.5, DALLE 3, and Stable Diffusion XL models that can expand your search. For example, I was recently researching fund flow modeling, and having visual examples was incredibly helpful. Broaden your search with images: Pro provides access to Playground V2.5, DALLE 3, and Stable Diffusion XL models that can expand your search. For example, I was recently researching fund flow modeling, and having visual examples was incredibly helpful. Broaden your search with images: Files, code, and more input options: You can upload unlimited files, code, pictures, etc. It’s really fast and much more convenient than doing this through a standard browser because it’s all in one space with seamless AI interaction. Files, code, and more input options: You can upload unlimited files, code, pictures, etc. It’s really fast and much more convenient than doing this through a standard browser because it’s all in one space with seamless AI interaction. Files, code, and more input options: ChatGPT Search ChatGPT Search This feature was launched in October 2024 and is based on GPT-4o, specifically trained for search. It’s integrated into ChatGPT , allowing you to enable it with a Search button or simply ask ChatGPT to search the web. ChatGPT As I mentioned, I’ve been a long-time user of ChatGPT for several years, so I seamlessly integrated the search feature into my routine. Since free Perplexity uses GPT-3.5, all its features like AI search with no ads, no endless scrolling, and results organized into threads with helpful tips are available here as well. My favorite feature here is its accessibility, offering text-to-speech, and the ability to create queries via audio, and audio dialogs. Choose your fighter Search Quality Search Quality Search Quality Search Quality Perplexity excels at finding information, while ChatGPT Search is better at explaining it. From a user perspective, I wouldn’t say I’ve encountered any critical issues with not finding something. It’s just a matter of different approaches. Perplexity provides more an overview of sources, whereas ChatGPT tends to give more explanation than summary. overview explanation Generated Answers Without Sources Generated Answers Without Sources Generated Answers Without Sources Generated Answers Without Sources Both tools provide sources for their answers, but sometimes the information might be generated without a reliable link. Even when a link is provided, the relevant information might not actually be there. It’s always better to double-check and read the sources yourself. Based on my experience, ChatGPT is more prone to generating unsupported content than Perplexity, but I’ve encountered this issue in both tools. Search Speed Search Speed Search Speed Search Speed I didn’t notice a significant difference between the two in terms of speed. However, both are slower than a standard browser search. Chrome extensions Chrome extensions Chrome extensions Chrome extensions Both tools offer Chrome extensions, and I tested both. The ChatGPT extension lets you use ChatGPT as your primary search engine, replacing, for example, Google. When you enter something in the search bar, it redirects you to ChatGPT, where your query is processed. Honestly, I wasn’t a fan. It’s slower than Google, and you can feel it. I still prefer a traditional search bar for quick lookups. Perplexity’s Chrome extension works differently. Instead of replacing your search engine, it opens a small window over the current page, allowing you to googling and query Perplexity simultaneously. I really loved this one, because it’s more useful for me. The drawback is that you can only enter one query at a time, and if you want to continue the conversation, it redirects you to a separate Perplexity tab. Unique Features Unique Features Unique Features Unique Features Since both tools use GPT models, they share many great features. Perplexity Pro stands out with its use of multiple models and built-in illustrations. On the other hand, I really appreciate ChatGPT’s accessibility and seamless integration into its apps, which I use for more than just search. Bonus: Hangman game test Bonus: Hangman game test Bonus: Hangman game test Bonus: Hangman game test I saw a funny tweet claiming ChatGPT can’t play Hangman. Turns out, neither can Perplexity Pro, even with all its models. Final Take: Which Works Best for What To start with, I’m a fan of AI search. Usually, it takes longer than quick googling, so here’s how I see it: if I need a super quick answer from the first link, I’ll use a standard browser search. But if I need to find more information or conduct deeper research, AI search is the way to go. It’s the best method to uncover valuable insights more efficiently. As for ChatGPT Search and Perplexity, in my experience, they solve different problems. ChatGPT Search perfectly extends ChatGPT’s capabilities, given that ChatGPT’s knowledge base only goes up to the year 2021. It’s perfect for studying and gaining insights using up-to-date web information. Perplexity, on the other hand, is better for finding something more thoroughly. It takes more from browser search, like incognito pages, and its approach to work with your prompts. It is about searching for information first and explaining second, in contrast to ChatGPT Search. For me: For me: Need to find something? Perplexity. Need to understand something? ChatGPT + Search. Need to find something? Perplexity. Need to find something? Need to understand something? ChatGPT + Search. Need to understand something? They complement each other perfectly in work.

Perplexity Pro vs. ChatGPT Search: Which AI Tool is Best Fits Your Workflow Needs?

Perplexity супраць пошуку ChatGPT: які інструмент AI лепш за ўсё адпавядае вашым патрэбам працоўнага працэсу?

Perplexity срещу ChatGPT Search: Кой AI инструмент отговаря най-добре на вашите нужди от работен процес?

Perplexity vs. Búsqueda en ChatGPT: ¿Qué herramienta de IA se adapta mejor a sus necesidades de flujo de trabajo?

Շփոթմունքն ընդդեմ ChatGPT որոնման. Ո՞ր AI գործիքն է լավագույնս համապատասխանում ձեր աշխատանքային հոսքի կարիքներին:

Perplexity と ChatGPT Search: どちらの AI ツールがワークフローのニーズに最適ですか?

ភាពច្របូកច្របល់ធៀបនឹងការស្វែងរក ChatGPT៖ តើឧបករណ៍ AI មួយណាដែលសាកសមបំផុតនឹងតម្រូវការលំហូរការងាររបស់អ្នក?

Perplexity vs. ChatGPT paieška: kuris AI įrankis geriausiai atitinka jūsų darbo eigos poreikius?

Fikarohana ChatGPT: Inona no fitaovana AI mifanaraka amin'ny filanao?

Збунетост наспроти Пребарување ChatGPT: која алатка за вештачка интелигенција најдобро одговара на вашите потреби на работниот тек?

Urujijo na ChatGPT Gushakisha: Niki Gikoresho cya AI gihuye neza nakazi kawe gakenewe?

குழப்பம் மற்றும் ChatGPT தேடல்: எந்த AI கருவி உங்கள் பணிப்பாய்வு தேவைகளுக்கு மிகவும் பொருத்தமானது?

Perplexity กับ ChatGPT Search: เครื่องมือ AI ตัวใดที่เหมาะกับความต้องการเวิร์กโฟลว์ของคุณที่สุด

Authors:
(1) Nora Schneider, Computer Science Department, ETH Zurich, Zurich, Switzerland (nschneide@student.ethz.ch);
(2) Shirin Goshtasbpour, Computer Science Department, ETH Zurich, Zurich, Switzerland and Swiss Data Science Center, Zurich, Switzerland (shirin.goshtasbpour@inf.ethz.ch);
(3) Fernando Perez-Cruz, Computer Science Department, ETH Zurich, Zurich, Switzerland and Swiss Data Science Center, Zurich, Switzerland (fernando.perezcruz@sdsc.ethz.ch). Table of Links Abstract and 1 Introduction 2 Background 2.1 Data Augmentation 2.2 Anchor Regression 3 Anchor Data Augmentation 3.1 Comparison to C-Mixup and 3.2 Preserving nonlinear data structure 3.3 Algorithm 4 Experiments and 4.1 Linear synthetic data 4.2 Housing nonlinear regression 4.3 In-distribution Generalization 4.4 Out-of-distribution Robustness 5 Conclusion, Broader Impact, and References A Additional information for Anchor Data Augmentation B Experiments A Additional information for Anchor Data Augmentation A.1 Derivation of ADA for nonlinear data A.2 Additional information on hyperparameters of ADA In this section, we illustrate in a simple 1D example (i.e. cosine data used in Figure 1) how changes in the hyperparameter values modify the data and affect the achieved estimation. Additionally, we show in Appendix B.4 how ADA performance on real-world data is impacted by changes in the hyperparameter values. Anchor Matrices and Locality: Anchor variable A is assumed to be the exogenous variable that generates heterogeneity in the target and has an approximately linear relation with (X, y) (see AR loss in Equation 3). It is recommended to choose the variable relying on expert knowledge about the features that the target has a higher dependence on or is possibly misrepresented in the dataset so that we encourage the robustness of the trained model against this type of discrepancy. After deciding the features, one way to construct the anchor matrix A is to partition the dataset according to the similarity of the features, using for example binning or clustering algorithms. Then we can fill the rows of A with a one-hot encoding of the partition index that each sample belongs to. We use the following nonlinear Cosine data model as a running example to demonstrate more clearly how A is constructed and affects the augmentation procedure. For illustration purposes, we use in Figures 5, 7 equidistant x values as this reduces noise and emphasizes the effect of ADA parameters more. Number of augmentations: For each anchor matrix A and γ we can add n new samples to the dataset. The addition of more augmented samples may not be beneficial as the optimization may overfit the approximations in the augmented data model in Equation 10. In the Cosine data model this is specifically problematic when X is close to multiples of π as depicted in Figure 7. Additionally, we provide a baseline and an augmented model fit in Figure 8 with different number of augmentations. As it is standard practice to use stochastic gradient descent methods for optimizing a regressor, we suggest applying ADA on each minibatch instead of the entire dataset. This avoids choosing a fixed numbers of augmentations. Furthermore, it adds diversity to the "mixing" behavior of ADA, because the samples that are being mixed change in each iteration. This paper is available on arxiv under CC0 1.0 DEED license. Authors: (1) Nora Schneider, Computer Science Department, ETH Zurich, Zurich, Switzerland (nschneide@student.ethz.ch); (2) Shirin Goshtasbpour, Computer Science Department, ETH Zurich, Zurich, Switzerland and Swiss Data Science Center, Zurich, Switzerland (shirin.goshtasbpour@inf.ethz.ch); (3) Fernando Perez-Cruz, Computer Science Department, ETH Zurich, Zurich, Switzerland and Swiss Data Science Center, Zurich, Switzerland (fernando.perezcruz@sdsc.ethz.ch). Authors: Authors: (1) Nora Schneider, Computer Science Department, ETH Zurich, Zurich, Switzerland (nschneide@student.ethz.ch); (2) Shirin Goshtasbpour, Computer Science Department, ETH Zurich, Zurich, Switzerland and Swiss Data Science Center, Zurich, Switzerland (shirin.goshtasbpour@inf.ethz.ch); (3) Fernando Perez-Cruz, Computer Science Department, ETH Zurich, Zurich, Switzerland and Swiss Data Science Center, Zurich, Switzerland (fernando.perezcruz@sdsc.ethz.ch). Table of Links Abstract and 1 Introduction Abstract and 1 Introduction 2 Background 2.1 Data Augmentation 2.1 Data Augmentation 2.2 Anchor Regression 2.2 Anchor Regression 3 Anchor Data Augmentation 3 Anchor Data Augmentation 3.1 Comparison to C-Mixup and 3.2 Preserving nonlinear data structure 3.1 Comparison to C-Mixup and 3.2 Preserving nonlinear data structure 3.3 Algorithm 3.3 Algorithm 4 Experiments and 4.1 Linear synthetic data 4 Experiments and 4.1 Linear synthetic data 4.2 Housing nonlinear regression 4.2 Housing nonlinear regression 4.3 In-distribution Generalization 4.3 In-distribution Generalization 4.4 Out-of-distribution Robustness 4.4 Out-of-distribution Robustness 5 Conclusion, Broader Impact, and References 5 Conclusion, Broader Impact, and References A Additional information for Anchor Data Augmentation A Additional information for Anchor Data Augmentation B Experiments B Experiments A Additional information for Anchor Data Augmentation A.1 Derivation of ADA for nonlinear data A.2 Additional information on hyperparameters of ADA In this section, we illustrate in a simple 1D example (i.e. cosine data used in Figure 1) how changes in the hyperparameter values modify the data and affect the achieved estimation. Additionally, we show in Appendix B.4 how ADA performance on real-world data is impacted by changes in the hyperparameter values. Anchor Matrices and Locality: Anchor variable A is assumed to be the exogenous variable that generates heterogeneity in the target and has an approximately linear relation with (X, y) (see AR loss in Equation 3). It is recommended to choose the variable relying on expert knowledge about the features that the target has a higher dependence on or is possibly misrepresented in the dataset so that we encourage the robustness of the trained model against this type of discrepancy. After deciding the features, one way to construct the anchor matrix A is to partition the dataset according to the similarity of the features, using for example binning or clustering algorithms. Then we can fill the rows of A with a one-hot encoding of the partition index that each sample belongs to. Anchor Matrices and Locality: A A We use the following nonlinear Cosine data model as a running example to demonstrate more clearly how A is constructed and affects the augmentation procedure. A For illustration purposes, we use in Figures 5, 7 equidistant x values as this reduces noise and emphasizes the effect of ADA parameters more. Number of augmentations: For each anchor matrix A and γ we can add n new samples to the dataset. The addition of more augmented samples may not be beneficial as the optimization may overfit the approximations in the augmented data model in Equation 10. In the Cosine data model this is specifically problematic when X is close to multiples of π as depicted in Figure 7. Additionally, we provide a baseline and an augmented model fit in Figure 8 with different number of augmentations. Number of augmentations: As it is standard practice to use stochastic gradient descent methods for optimizing a regressor, we suggest applying ADA on each minibatch instead of the entire dataset. This avoids choosing a fixed numbers of augmentations. Furthermore, it adds diversity to the "mixing" behavior of ADA, because the samples that are being mixed change in each iteration. This paper is available on arxiv under CC0 1.0 DEED license. This paper is available on arxiv under CC0 1.0 DEED license. available on arxiv

How Hyperparameter Tuning Enhances Anchor Data Augmentation for Robust Regression

Stories, News & Mentions about ILLUSTRIAIO

Illustriaio’s Stories on HackerNoon