Jaswant B’s Post

Advancing time series analysis with multi-granularity guided diffusion model; An algorithm-system co-design for fast, scalable MoE inference; What makes a search metric successful in large-scale settings; learning to solve PDEs without simulated data. https://msft.it/6045lyVfD

💡 Ever wondered how algorithms handle growing data? In my latest blog, I break down the essentials of computational complexity—think Linear, Log-Linear, Quadratic, and Exponential time, all with simple examples you can relate to. #Algorithms #BigONotation #CodingMadeSimple #SoftwareDevelopment #Tech #Trends

Struggling to make LLMs generalize to out-of-distribution data? Your frustration might be alleviated by an ICLR 2024 paper, which rigorously proved that learning an approximate causal model is a necessity for OOD generalization. The result has far broader implications, as detailed in Section 4. One thing I'm speculating is that learning a causal model is the only way to be sample efficient and is the fundamental difference between LLMs and human brains. #artificialintelligence #machinelearning #deeplearning https://lnkd.in/eUPFa3fw

From bad to worse: a new slimmed down Whisper model is now available open source and has the potential to be run locally. "Trained on >5M hours of labeled data, ... a finetuned version of a pruned ... exact same model, ... way faster, at the expense of a minor quality degradation." https://lnkd.in/e7ec5xPu

Support Vector Machines (SVMs) are a robust and versatile supervised machine learning algorithm widely used for both classification and regression tasks. They are particularly effective in high-dimensional spaces and excel at handling complex datasets. By leveraging kernel functions, SVMs can map data into higher-dimensional feature spaces, uncovering intricate patterns and relationships that may not be immediately apparent. Discover more about how SVMs are transforming data analysis here: Learn More. 🚀 https://lnkd.in/eS22YXm7 . #MachineLearning #SVM #DataScience #AIInnovation

Feature scaling is a cornerstone of effective machine learning, addressing challenges posed by datasets with features on different scales. In my latest blog, I explore: ✅ The critical role of feature scaling in improving model accuracy and fairness. ✅ Advanced techniques like Min-Max Scaling and Standardization. ✅ Strategic considerations for selecting the optimal approach based on the dataset. Dive in to learn how scaling can elevate your preprocessing pipeline and drive better outcomes. I look forward to your perspectives and insights.

From Microsoft Research: GraphRAG (Graphs + Retrieval Augmented Generation) is a technique for richly understanding text datasets by combining text extraction, network analysis, and LLM prompting and summarization into a single end-to-end system -

Vector similarity search is how machines compare things like images or text, finding similar items even in massive datasets. As datasets grow and use cases become even more complex, SOAR (by Google Research) ensures ScaNN (open-source vector search library) can deliver the scalability and performance ML applications demand 🎉 . How SOAR Transforms Vector Search    •   Redundancy for Speed: SOAR leverages carefully crafted redundancy to make approximate nearest neighbor searches even faster within ScaNN.    •   Efficient and Scalable: This boost in performance comes at minimal cost to index size and other important metrics. Check out the entire paper here 👉 https://lnkd.in/g2SPjVy3 #ML

📣 Analyze data 18x faster with the power of machine learning enabled BD ElastiGate™ Auto-Gating Algorithm Technology. Learn more at: https://bit.ly/3VfERVh

Understanding the run time of #ML algorithms is important because it helps us: - Build a core understanding of an algorithm. - Understand the data-specific conditions that allow us to use an algorithm. For instance, using SVM or t-SNE on large datasets is infeasible because of their polynomial relation with data size. Similarly, using OLS on a high-dimensional dataset makes no sense because its run-time grows cubically with total features.