Siddhesh Rane

https://lnkd.in/g9xx7Ztd Unification of all fundamental forces using Alena Tensor In this work we established new framework to include gravity into QFT by doing it in flat curvature but introducing interaction with gravity using Alena tensor instead which will be proved to be equivalent to contracted christofel symbol as we have proposed before. The Unified Theory That All Seek The universe hides its forces in its threads, Gravity, immense, its path widespread. Electromagnetic fields entwined, Nuclear siblings with power combined. Scientists dream deep in their hearts, To unite all forces, their paths chart. Alena Tensor, a bridge to start, A journey to unify every part. In smooth and curved simplicity, The Christoffel symbols hold the key. Each sign guides the light to see, A theory for the universe, in harmony. All four forces join as one, Bound together, under the sun. In this theory so brightly spun, The truth will shine when the work is done! Die Vereinheitlichte Theorie, die alle suchen Das Universum verbirgt seine Kräfte im Band, Die Schwerkraft ist groß und mächtig bekannt. Elektromagnetische Felder verwoben klar, Kernkräfte vereint, so wunderbar. Die Wissenschaft träumt tief und rein, Alle Kräfte zu einen, das Ziel muss sein. Alena Tensor, der Brücke Schein, Ein Pfad zur Einheit, so licht und fein. In glatter und gebogener Symmetrie, Zeigen Christoffel-Symbole die Melodie. Jedes Zeichen führt das Licht zur Theorie, Ein Universum in perfekter Harmonie. Alle vier Kräfte in einem vereint, Zusammengeführt, wie der Himmel scheint. In dieser Theorie, die so hell erscheint, Wird die Wahrheit leuchten, wenn das Werk gemeint!

1

1 Comment

Hi folks 👋, have you ever heard of OCCAM? It's a simple, yet powerful language designed specifically for concurrent and distributed systems created in 1983. 🛠️   Here's a quick example of OCCAM code:   CHAN OF INT channel;   PROC producer()  IS    VAR INT data := 0;    WHILE TRUE      SEQ        data := data + 1;        out channel data;    ENDWHILE  ENDPROC   PROC consumer()  IS    VAR INT data;    WHILE TRUE      SEQ        in channel data;        output "Received: ", data, NL;    ENDWHILE  ENDPROC   PROC main()  IS    PAR      producer();      consumer();    ENDPAR  ENDPROC   This code demonstrates a simple producer-consumer pattern using OCCAM's channels for communication between processes.   💡 Why choose OCCAM?   Clean and concise syntax ✍️: OCCAM's syntax is designed to be easy to read and write. Explicit concurrency 🔄: OCCAM makes concurrency a first-class concept, making it ideal for distributed systems. Strong type system 🛡️: OCCAM's strong type system helps prevent errors and improve code reliability.   Let me know in the comments what you think 👇   #OCCAM #ProgrammingLanguage #DistributedSystems #Concurrency #SoftwareDevelopment 🚀

20

5 Comments

Recently I met up with an old course mate, and he shared some insightful advice about his transition to tech after two years in microchip manufacturing 💡 There are some key actions he took that I found relevant for many career switchers out there: 1️⃣ 𝐁𝐮𝐢𝐥𝐝 𝐩𝐫𝐨𝐣𝐞𝐜𝐭𝐬 𝐚𝐜𝐭𝐢𝐯𝐞𝐥𝐲 - Although he wasn't in the industry, he built numerous side projects (machine learning + web apps) to keep his skills sharp. He even created a YouTube channel to share his knowledge with a broader audience. Teaching is one of the best ways to learn, and through this channel, he grew tremendously, both in terms of his technical skills and learning abilities. 2️⃣ 𝐃𝐞𝐝𝐢𝐜𝐚𝐭𝐞𝐝 𝐟𝐨𝐜𝐮𝐬 & 𝐜𝐚𝐫𝐞𝐞𝐫 𝐛𝐫𝐞𝐚𝐤 - As much as he would have liked to spend more time learning computer science, the truth is, he could only dedicate limited hours to the subject every week. After weighing his situation from different aspects, he decided to resign and pursue a master's degree to further improve his skills. Personally, I don't think a degree is necessary to succeed in computer science, but for him, it was a good time to channel all his energy into what he loves and have more time each week to build his computer science skillset. He could spend more time building his portfolio and preparing for interviews. 3️⃣ 𝐁𝐞𝐢𝐧𝐠 𝐬𝐭𝐫𝐚𝐭𝐞𝐠𝐢𝐜 𝐰𝐢𝐭𝐡 𝐲𝐨𝐮𝐫 𝐜𝐡𝐨𝐢𝐜𝐞 - When deciding which stream to join for his master's, the key question he asked was how he could stand out from others who were already 2-3 years into the industry. After doing some research, he decided to pursue AI/ML for his master's. He wanted to explore the MLOps area, which requires a combination of software engineering and AI skills, an area with relatively lower competition in the market. Fast forward to today, he is currently working as an MLOps engineer at a bank. He enjoys his current work a lot and is happier than ever, having achieved his goal after 2-3 years of a bumpy career exploration journey. I find his story very inspiring and a good learning reference for some of you who are currently doing career switches. Do you have any learnings or advice that you can share? Share them in the comment below ! ----- 🔔 I post content about software engineering, follow me on LinkedIn and subscribe to my newsletter for more content like this. #softwareengineering #career

25

2 Comments

The parameters of compiler optimizations interact in complex and non-trivial ways, and finding optimal combinations remains an open research challenge. There are several approaches to discovering effective combinations of these optimization parameters. Recently, Michael Canesche (Cadence Design Systems) and Gaurav Verma (Stony Brook University) have had promising results using auto-tuning with coordinate descent. To visualize this process, imagine that each optimization represents a 'dimension' in a search space. The combination of all optimizations and their respective parameters forms a vector in this space. Now, consider that each point in this space is mapped to a real number representing the performance of a kernel (assuming a fixed input). Coordinate descent is a search technique that assumes the performance curve is convex. Based on this assumption, it aims to find the optimal point by varying one optimization parameter at a time. Explore as a Storm, Exploit as a Raindrop: On the Benefit of Fine-Tuning Kernel Schedulers with Coordinate Descent: https://lnkd.in/dgxxfdYR The Droplet Search Algorithm for Kernel Scheduling: https://lnkd.in/dBfQid9z

77

Happy to share that our latest paper is out! In this study, we delve into the world of Slow Feature Analysis, a powerful technique to transform measured data into uncorrelated signals ranging from slow to fast. We've introduced a novel approach that goes beyond traditional methods, addressing the challenge of nonstationary and oscillating features. Our semi-supervised encoder-decoder architecture incorporates a statistical preference for these characteristics, paving the way for more accurate modeling. Curious about the results? We put our approach to the test on both simulated and real industrial processes, with promising outcomes. Thanks to my supervisor, Prof. Biao Huang. Read more about our findings in the full paper! https://lnkd.in/gK6naG5U

61

2 Comments

"Recently the standard GNN has been combined with ideas from other ML models to develop exciting innovative applications. One such development is the integration of GNN with sequential models — Spatio-Temporal GNN that is able to capture both the temporal and spatial (hence the name) dependences of data, this alone could be applied to a number of challenges/problems in industry/research." #graphneuralnetworks #GNN #research #machinelearning #deeplearning #research #drugdiscovery

37

How much should you trust your compiler? For most developers, the compiler is something you take for granted. If there are bugs, they must surely originate from your code. Except that isn’t quite right. Compilers can have many kinds of bugs. From the obvious ones (performance regressions, compilation errors), to more insidious translation bugs. Compilers can have translation bugs where the source and translated programs exhibit different observable behaviors. Translation Validation (TV) attempts to provide guarantees on the correctness of compilation. Intrigued? We had Prof. Supratik Chakraborty from Indian Institute of Technology, Bombay at the Innovations In Compiler Technology Workshop, where he gave a talk on Translation Validation. You can check it out here: https://lnkd.in/dXUmZ-s3 Follow our channel to know about the bleeding edge in Compiler Technology: https://lnkd.in/dtVC39Yp #IICT #IICT2024 #Compilers #CompilerTechnology #LLVM #gcc #clang #FormalVerification #software #cplusplus #c #programming

79

2 Comments

SIMD (Single Instruction, Multiple Data) is a parallel computing method where a single CPU instruction is used to operate on multiple data points simultaneously, significantly enhancing computational efficiency and speed. By using SIMD instructions in functions too complicated to be vectorized by the compiler, I achieved a quite noticeable speedup in the past months. Thus, I want to share a video showcasing the Selective Laser Melting (SLM) process of polydisperse titanium powder in a 2x2 mm² area, representing the largest domain I was able to simulate by now. The domain is discretized using 7 million SPH particles. The simulation ran for 10 hours on a Ryzen 7950x3D CPU, and the video was rendered using Blender. If you’re interested in how our simulation techniques can benefit your projects, please don't hesitate to contact us at blank-simulations. #SPH #multiphysics #raytracing #additivemanufacturing Method: - Smoothed Particle Hydrodynamics (SPH) - MPI-OpenMP parallelization - Dynamic workload balancing - Adaptive particle refinement Physics: - Ray tracing to model laser-material interaction - Temperature-dependent material properties - Latent heat of fusion and crystallization - Evaporation and recoil pressure - Surface tension and wetting

52

2 Comments

https://lnkd.in/gW9r4nn6 Our paper on the open-source phase-field code MicroSim was published this week, and 17 authors contributed. The National Supercomputing Mission (NSM) funded project over 3 years leading to this paper is a culmination of efforts in model formulations and code implementations in the individual groups of Phanikumar Gandham Saswata Bhattacharya Gururajan Mogadalai Venkatesh Shenoi and Abhik Choudhury. Link to the code: https://lnkd.in/gSFywzwC Kudos to the developers and the past and present students in our groups that led to this development This is also an example of how collaboration among groups can lead to something significant and valuable.

96

9 Comments

I once confused C-style pointers with C++ references. And that was earlier this year. I had been asked to write the driver for an I/O expander that'd be used to control a four-digit seven-segment display via I2C. My code would be integrated into a larger project. As someone familiar with only C, I wrote the driver for the expander in C. After testing the code and ensuring it worked properly, I was ready to integrate it with the main project. I cloned the codebase and created a branch for my work. When I opened the code files, I was shocked that they were all in C++. The syntax looked strange, so I had to start learning C++ on the job. While going through the code, I quickly noticed functions that took references as arguments. I initially thought it was an error from the code's author: I didn't know about C++ references, so I thought the author had erroneously used "&" instead of "*" to specify a pointer argument. I started researching this syntax and not long after, I discovered these were not pointers but references. I learnt that although C++ offers C-style pointers, references are safer. For example, a reference must always refer to a valid object while a pointer can be set to NULL, which could lead to safety issues like null pointer dereferencing. Additionally, references are quite straightforward to understand than pointers. References have a simple syntax. You only need to use "&" when defining reference variables or specifying reference parameter types. There's no such thing as "dereferencing a reference." For pointers, however, you'd need to take care not to use "&" and "*" wrongly. Featured image: my set-up while testing the I/O expander driver.

87

12 Comments

Simulink with super-powers...

5

🔗 Simulate PyTorch Models with Simulink! 🚀 Engineers, unlock new possibilities by simulating PyTorch and other Python-based models directly in Simulink using co-execution blocks. This integration streamlines your workflow, enabling seamless model testing and validation. 🔍 What's Inside: >> Co-execution of Python models in Simulink << >> Enhanced simulation and validation processes << 👉 Dive into the details: https://spr.ly/6041o1cOX #DeepLearning #Simulink #Python #ModelSimulation #Engineering #MATLAB #MAthWorks #PyTorch

9

Must seen for who’s interested in recent advances in robot learning for manipulation.

6

Attention in Encoder-Decoder Models: LSTM Encoder-Decoder with Attention The video: https://lnkd.in/grT6hk_G This video explains attention in neural networks! In this video, we learn LSTM-based Encoder-Decoder models with Attention, breaking down how attention mechanisms work and why they revolutionize sequence-to-sequence tasks like machine translation, text summarization, and more. 🔍 What you'll learn: -- What attention mechanisms are and why they're important -- How attention enhances Encoder-Decoder models -- A step-by-step explanation and calculation of the attention process in LSTM Encoder-Decoder models.

31

🚨Paper Alert 🚨 ➡️Paper Title: Evaluating Real-World Robot Manipulation Policies in Simulation 🌟Few pointers from the paper 🎯The field of robotics has made significant advances towards generalist robot manipulation policies. However, real-world evaluation of such policies is not scalable and faces reproducibility challenges, which are likely to worsen as policies broaden the spectrum of tasks they can perform. 🎯In this paper authors have demonstrated that simulation-based evaluation can be a scalable, reproducible, and reliable proxy for real-world evaluation. 🎯They identified control and visual disparities between real and simulated environments as key challenges for reliable simulated evaluation and propose approaches for mitigating these gaps without needing to craft full-fidelity digital twins of real-world environments. 🎯They then employ these approaches to create “SIMPLER”, a collection of simulated environments for manipulation policy evaluation on common real robot setups. 🎯 Through paired sim-and-real evaluations of manipulation policies, authors demonstrated strong correlation between policy performance in SIMPLER environments and in the real world. 🎯Additionally, they found that SIMPLER evaluations accurately reflect real-world policy behavior modes such as sensitivity to various distribution shifts. 🎯Lastly, authors have open-source all SIMPLER environments along with their workflow for creating new environment to facilitate research on general-purpose manipulation policies and simulated evaluation frameworks. 🏢Organization: UC San Diego, Stanford University, University of California, Berkeley, Google DeepMind 🧙Paper Authors: Xuanlin Li, Kyle Hsu, Jiayuan Gu, Karl Pertsch, Oier Mees, Homer Walke, Chuyuan (Kelly) Fu, Ishikaa Lunawat, Isabel Sieh, Sean Kirmani, Sergey Levine, Jiajun Wu, Chelsea Finn, Hao Su , Quan Vuong , Ted Xiao 1️⃣Read the Full Paper here: https://lnkd.in/gjEyPeuE 2️⃣Project Page: https://lnkd.in/gTqxuNTE 3️⃣Code: https://lnkd.in/gmiG7UNJ 🎥 Be sure to watch the attached Video-Sound on 🔊🔊 🎵Music by Yurii Semchyshyn from Pixabay Find this Valuable 💎 ? ♻️REPOST and teach your network something new Follow me 👣, Naveen Manwani, for the latest updates on Tech and AI-related news, insightful research papers, and exciting announcements.

3

🤖 Check out the MATLAB Deep Learning Hub on GitHub! What you'll find: 🧠 Examples and tutorials for deep learning concepts 🚀 Pretrained models to kickstart your project Get Started: 🖱️ Go to the MATLAB Deep Learning Hub on GitHub. 👀 Browse the resources. ⭐ Download, star, or fork your favorites and apply them to your projects! Click the link in our bio! #DeepLearning #MachineLearning #AI #MATLAB #GitHub #LearnToCode

1

Course 8/30: Computer Vision Onramp by MathWorks Today, I explored the fascinating world of computer vision with MathWorks! This course introduced me to the basics of image processing and feature detection using MATLAB. Key Takeaways: Image Analysis Essentials: Learned how to analyze and process images, understanding fundamental techniques to prepare data for computer vision tasks. Feature Detection: Discovered methods to identify and track key features within images, paving the way for object recognition. Hands-On Practice: Put theory into action by working directly in MATLAB, using practical tools to bring computer vision concepts to life. Excited to share this new certification! https://lnkd.in/gUp7QkA5 MathWorks Computer Vision Ready to take on the next challenge! 😁👍 #LearningJourney #ComputerVision #MathWorks #30DaysOfLearning

2

Encoder-Decoder: Why it still matters All of us who work in deep learning know what an encoder-decoder structure is. But the architecture is quite old at this point (especially by the standards of a field which is overhauled every 2 months). However, the base structure of an encoder-decoder is still highly relevant. This is because irrespective of whether you use diffusion of transformers (few of which only use the encoder), the basis of the architecture is inherently highly powerful. At the heart, an encoder-decoder structure takes the data inputted to its most relevant data embeddings (via the encider) and then takes it back to its original dimension (via the decoder). This is relevant, because this is essentially knowledge distillation. Not all parts of data (image, text, signal, doesn't matter) is equally relavant to a specific problem. Of course, no data is bad data, but neither is it important. Encoder-decoders overcame this problem by essentially letting a model focus on exactly what mattered. In a way, think of it similar to attention, except, in a non-structured way. This was important because when you look at these embeddings in a higher dimensional space (I use the term 'looking' very loosely because there is no way toa ctually visualize this except in very abstract mathematical terms), these all become highly structured (or, in other words, linear). That structural information becomes easier for a DL model to understand. I hope to someday finally have some time to dive down deeper into how to optimally represent these (that is the basis of why DL is still considered a black box), but esentially what I am trying to say is that just because you know what attention is, do not throw out encoder-decoders. They formed the basis of DL, and is still highly relevant. Projecting this further, ML and statistics is similarly highly relevant even today, but that is a post for another day PC: Medium (A Perfect guide to Understand Encoder Decoders in Depth with Visuals)

10