Chaitanya Tumula’s Post

Excited to be sharing another blog post, this time diving into the topic of regularisation in neural networks! Some popular regularisation techiniques can initially seem quite bizarre, but after reading about their central importance in deep learning I wanted to dig deeper for myself. I give both experimental and theoretical justifications for some common regularisation techniques, which I hope makes the topic both interesting and accessible. Excited to hear your thoughts! https://lnkd.in/d-6prCCJ

Very interesting new paper about the future of Bayesian Deep Learning. It seems the wind is beginning to change but people are not quite yet ready to abandon neural networks completely. "In the current landscape of deep learning research, there is a predominant emphasis on achieving high predictive accuracy in supervised tasks involving large image and language datasets. However, a broader perspective reveals a multitude of overlooked metrics, tasks, and data types, such as uncertainty, active and continual learning, and scientific data, that demand attention. Bayesian deep learning (BDL) constitutes a promising avenue, offering advantages across these diverse settings. This paper posits that BDL can elevate the capabilities of deep learning. It revisits the strengths of BDL, acknowledges existing challenges, and highlights some exciting research avenues aimed at addressing these obstacles. Looking ahead, the discussion focuses on possible ways to combine large-scale foundation models with BDL to unlock their full potential." https://lnkd.in/d-Xyjz2e

The more you understand about deep learning, the more obvious it becomes to you that these networks are shockingly good at fitting any dataset and modelling any data distribution. https://lnkd.in/edqQbWQ8

Fantastic! IEEE is releasing a special journal issue on the Mathematics of Deep Learning. This is a great opportunity for anyone interested in understanding the fundamental science behind deep learning innovations. This should be mandatory literature for non-mathematical public research institutions and their employees—especially those who dare to claim they are building AI without fundamental mathematical knowledge. #mathematicsofdeeplearning #MathematicsOfAI #AIEthics #AcademicResearch #SignalProcessing

What are the different types of machine learning? Classical machine learning is typically categorized by the manner in which algorithms improve their predictive accuracy. The four fundamental types of machine learning are: ✅ Supervised learning ✅ Unsupervised learning ✅ Semi-supervised learning ✅ Reinforcement learning The selection of an algorithm is influenced by the characteristics of the data and the problem at hand. Furthermore, many algorithms are versatile and can be applied across multiple learning paradigms. For example, deep learning architectures such as convolutional neural networks (CNNs) and recurrent neural networks (RNNs) can be employed in supervised, unsupervised, and reinforcement learning contexts, depending on the specific requirements of the task and the data available.

Graph representation learning focuses on transforming high-dimensional, sparse graph data into low-dimensional vectors. Traditional methods aim to keep the embeddings of connected nodes close, preserving the graph's structure, but they face limitations such as restricted model capacity, reliance on outdated unsupervised strategies, and the need for joint optimization with downstream tasks. Deep learning, particularly graph neural networks, has shown significant advantages over these traditional methods. This survey reviews current deep graph representation learning algorithms, introduces a new taxonomy, categorizes approaches based on neural network architectures and learning paradigms, and highlights practical applications. It also offers future research directions and challenges in the field. Paper: https://lnkd.in/gr_zktxw

A couple of new and interesting reads from my group! DeepUQ: Assessing the Aleatoric Uncertainties from two Deep Learning Methods: https://lnkd.in/gjxQzcDi Neural Network Prediction of Strong Lensing Systems with Domain Adaptation and Uncertainty Quantification: https://lnkd.in/gWh8RKMb Domain-Adaptive Neural Posterior Estimation for Strong Gravitational Lens Analysis: https://lnkd.in/g8XZbVkr Population-level Dark Energy Constraints from Strong Gravitational Lensing using Simulation-Based Inference: https://lnkd.in/gZuex3an

Two months ago, I only had a vague idea of how neural networks were constructed and used. After this course, I've been able to create a full transformer model and I am super proud of this achievement that required hours of work, watching videos, and hands-on labs practicing. If you strive to learn and be curious like me, I can only recommend this course to understand better the world of deep learning.

Mathematical vectors: the building blocks of AI systems. Learn essential vector and matrix operations that power neural networks. Perfect for ML engineers and data scientists exploring deep learning fundamentals.

Book of the week 📚 Every week, I review a book 📚 in my newsletter 📝. This week's spotlight is Deep Learning by Prof. John D. Kelleher. If you are looking for a down-to-earth and concise resource for getting started with deep learning, this book is a great choice. This pocket-size book covers the foundation of deep learning, and it includes the following topics: ✅ Deep learning history ✅ Foundations of deep learning ✅ Neural networks ✅ Convolutional and Recurrent Neural Networks (i.e., CNN and RNN) ✅ Learning function More details are available in my newsletter. ⭐️ Newsletter: https://lnkd.in/gx6T3UUP ⭐️ Data Science Channel: https://lnkd.in/g_GdP-pf #deeplearning #machinelearning #datascience #ai