Hiyansh Chandel’s Post

Red HOT - 29Oct Publication; Deep Learning with JAX, Google's numerical computing library. quote "The JAX numerical computing library tackles the core performance challenges at the heart of deep learning and other scientific computing tasks. By combining Google’s Accelerated Linear Algebra platform (XLA) with a hyper-optimized version of NumPy and a variety of other high-performance features, JAX delivers a huge performance boost in low-level computations and transformations." end-quote. Publisher: Manning, https://lnkd.in/ewEApm_R #investments, #ai, #machinelearning, #python, #trading, #investments, #cfa, #statistics, #python, #hedgefund, #quantitativefinance, #deeplearning, #quant, #money, #forex, #alternativeinvestments, #gold

In computer vision, I have explored, observed and utilized various types of tracker for various applications. Generally we classify trackers into two types. 1. classical trackers 2. deep learning based trackers. 1.Classical trackers are filter based but they come with limitations pro and cons. Classical trackers include KCF, CSRT, Median Flow, MIL, GOTURN etc. We use Median Flow where the movement is not rapid and it is based on Lucas-kanade filter. Some of them don't handle occlusion very well. 2. Deep learning based trackers as a name suggests a tracker train on some dataset like GOT-10k, VOT18 etc. they are excellent for Computer vision applications but require a lot of resources. For me, Dimp and Tomp are my fav tracker based on pytracking repository. Below is the result of DaSiamrpn Tracker. I wish to write more on this topic like backbone and architecture of trackers and models but the post will get quite long and boring. #computervision #opencv #machinelearning #python #datascience

Sharing some Data Science useful topics under #Learn&share to best practice: Pytorch: Pytorch is a highly flexible and open-source machine learning framework that is widely used for developing neural network models. It offers modularity and a huge ecosystem of tools for handling various types of data, such as text, audio, vision, and tabular data. With GPU and TPU support, you can accelerate your model training by 10X. For Example, almost everything in a PyTorch neural network comes from torch.nn: ** nn.Module contains the larger building blocks (layers) ** nn.Parameter contains the smaller parameters like weights and biases (put these together to make nn.Module(s)) ** forward() tells the larger blocks how to make calculations on inputs (tensors full of data) within nn.Module(s) ** torch.optim contains optimization methods on how to improve the parameters within nn.Parameter to better represent input data. #DataScience #Learn&Share #PyTorch #Tensor #Python #LearningTopics

In rapidly growing AI in real life, mathematicians must equiped strong computer science skills. As, AI based on machine learnig and machine learning is the intersection of mathematics and computer scienc.  I believe for true innovation in ML mathematician must be equiped with computaional frameworks. I recomend expertise in python for machine learning. So, learn python and train machines.  #AI #MachineLearning #DataScience #Collaboration #Tech #Mathematics #Python

Thrilled to announce that I've successfully completed the Supervised Machine Learning: Regression and Classification course by DeepLearning.AI and Stanford University with the final grade of 93.7%. During this course, I delved deep into Logistic Regression and applied Gradient Descent, exploring various mathematical implementations of the cost and loss functions. I also used different python libraries like NumPy,Pandas,Scikit-learn for applying the mathematical implementations of the models developed.Additionally, I gained a comprehensive understanding of classification techniques, including overfitting and underfitting models, and the importance of regularization. Excited to leverage these new skills in tackling real-world challenges! #MachineLearning #AI #LogisticRegression #GradientDescent #ContinuousLearning #Coursera

Started building a complete neural network implementation using numpy and high school mathematics. The project is inspired by Sentdex, and I am trying to build upon it by capturing every detail for every step involved in both the feed forward network (forward pass / forward propagation) which primarily utilizes linear algebra and the backward propagation which makes use of calculus. I would highly recommend checking out the Github repository of the same for better understanding of how neural networks work under the hood. The the project is W. I. P and will keep adding to it as we move forward. Repost for others to learn how basic mathematical concepts can build such beautiful models that can learn almost anything! Neural Step Engine - https://lnkd.in/eHju5qqc #neuralnetworks #artificialintelligence #deeplearning #machinelearning #python

Thrilled to announce that I’ve completed the Computer Vision course at the Information Technology Institute (ITI) It was an incredible journey diving deep into topics like Python for Machine Learning NumPy Linear Algebra Probability and Statistics for ML Data Visualization and Exploration Classic Machine Learning Introduction to Deep Learning and Introduction to Computer Vision. This course was a game-changer, enhancing my skills and knowledge in AI and Computer Vision. #AI #ComputerVision #DeepLearning #MachineLearning #ITI

𝗖𝗮𝗹𝗰𝘂𝗹𝗮𝘁𝗶𝗻𝗴 𝘁𝗵𝗲 𝗕𝗮𝗰𝗸𝗽𝗿𝗼𝗽𝗮𝗴𝗮𝘁𝗶𝗼𝗻 𝗕𝘂𝗶𝗹𝗱𝗶𝗻𝗴 𝗕𝗹𝗼𝗰𝗸: 𝗚𝗿𝗮𝗱𝗶𝗲𝗻𝘁𝘀 𝘄𝗶𝘁𝗵 𝘁𝗳.𝗚𝗿𝗮𝗱𝗶𝗲𝗻𝘁𝗧𝗮𝗽𝗲 In a neural network, backpropagation is essential for error minimization. 🔄 It involves calculating the partial derivatives, or gradients, of the loss function with respect to trainable parameters. 📉 But manually computing and implementing these derivatives in Python can be complex and requires mathematical knowledge. 📚 Fortunately, TensorFlow simplifies this with tf.GradientTape. 🎉 𝗪𝗵𝗮𝘁 𝗶𝘀 𝘁𝗳.𝗚𝗿𝗮𝗱𝗶𝗲𝗻𝘁𝗧𝗮𝗽𝗲? 🤔 tf.GradientTape in TensorFlow is used to calculate the gradients of computations involving tensors. 𝗛𝗼𝘄 𝗗𝗼𝗲𝘀 𝗜𝘁 𝗖𝗮𝗹𝗰𝘂𝗹𝗮𝘁𝗲 𝗚𝗿𝗮𝗱𝗶𝗲𝗻𝘁𝘀? 🤓 TensorFlow records operations applied to tensors, especially tf. Variable objects, within the scope of tf.GradientTape. When the gradient method is called on the tape, it calculates the gradient with respect to the specified inputs. 𝗘𝘅𝗮𝗺𝗽𝗹𝗲 𝗖𝗼𝗱𝗲 💻: w = tf.Variable(initial_value=5.) with tf.GradientTape() as tape: loss = tf.square(w) gradient = tape.gradient(loss, w) print(gradient.numpy()) # Output: 10.0 𝗘𝘅𝗽𝗹𝗮𝗻𝗮𝘁𝗶𝗼𝗻: 👉w is initialized as a TensorFlow variable with a value of 5. 👉tf.GradientTape records operations to compute gradients. 👉loss is defined as the square of w. 👉The gradient computes the derivative of loss with respect to w. 👉The gradient of w^2 is 2w. So for w = 5, the gradient will be 2×5 = 10. 𝗡𝗼𝘁𝗲: tf.GradientTape records operations on tf.Variable. To compute gradients for constants, explicitly tell the tape to watch them using tape.watch(const_var). #python #tensorflow #NeuralNetworks #GradientDescent #Backpropagation #DataScience #Coding #TechTips #Programming #MLTips #AIResearch #TensorFlowTips

🚀 Day 28: Mastering Non-Linear Binary Classification with TensorFlow 🚀 Today, we’re diving into Non-Linear Binary Classification, using TensorFlow to build a model that can handle complex decision boundaries. Unlike linear classification, non-linear classification is essential for tasks where the data cannot be separated by a straight line. We'll use the make_circles function from scikit-learn to create a challenging dataset and aim to achieve higher accuracy (above 99% ). In this post, we’ll cover: Generating Non-Linear Data with make_circles. Building a Non-Linear Classifier in TensorFlow. Optimizing the Model to achieve high accuracy. Visualizing Non-Linear Decision Boundaries. Check out the full code on my GitHub: TensorFlow Masters. 🚀 Let’s build models that go beyond simple boundaries! #MachineLearning #AI #DeepLearning #TensorFlow #BinaryClassification #NonLinearModels #DataScience #NeuralNetworks #ArtificialIntelligence #ML #DL #AIResearch #Python #BigData #ModelTraining #Tech #AICommunity #NonLinearDecisionBoundary #TensorFlowJourney

🔥 NumPy vs. PyTorch – Which One Should You Use? If you're working with machine learning or deep learning, you've probably come across both NumPy (numpy) and PyTorch (torch). But when should you use each? 🤔 🔹 Key Differences: ✅ NumPy: Best for general numerical computations, data manipulation, and scientific computing. ✅ PyTorch: Designed for deep learning, supports GPU acceleration, and has automatic differentiation (autograd). #Python #MachineLearning #DeepLearning #PyTorch #NumPy #AI #DataScience