ClearSignal

NOTES: Retrieval-Augmented Generation for Large Language Models

papers Featured

NOTES: Retrieval-Augmented Generation for Large Language Models

Here is a short compilation of bullet points gathered while reading the paper " Retrieval-Augmented Generation for Large Language Models: A Survey". The authors did a reasonably good job summarizing the landscape of current RAG systems.

papers Featured

Yolo-X Notes

The main motivation behind YOLOX was to update the YOLO series with the recent advancements at the time, particularly anchor-free detection.

Multi-task Loss

Multi-task Loss

This is a short review of the paper titled "Multi-Task Learning Using Uncertainty to Weigh Losses for Scene Geometry and Semantics" by Kendall et al, 2018 .

Polynomial fitting using Pytorch

Polynomial fitting using Pytorch

Here is a short tutorial on how to fit polynomials using pytorch.

Affine augmentations

Affine augmentations

Old notes on affine augmentations. Enjoy!

Migrate your blog from Ghost(Pro) to Digital ocean

Migrate your blog from Ghost(Pro) to Digital ocean

I recently decided to migrate my ghost blog from ghost(pro) subscription to a digital ocean droplet. Primary reasons for the migration were: 1. Inability to modify themes with a basic subscription 2. Limited integrations. Here are the steps involved in the migration: Downloading content from Ghost(pro) website * Navigate

SVD

SVD (Singular Value Decomposition) is one of my favorite topics in linear algebra. It's almost magical to factorize any matrix into a product of two orthogonal matrices and a diagonal matrix. Derivation In simple words, any matrix 'A' can be decomposed into a product of 3

PolyLoss : A new framework for loss functions

papers Featured

PolyLoss : A new framework for loss functions

Original Paper: [https://arxiv.org/pdf/2204.12511.pdf] Authors: Zhaoqi Leng, Mingxing Tan , Chenxi Liu , Ekin Dogus Cubuk , Xiaojie Shi, Shuyang Cheng ,Dragomir Anguelov [Waymo And Google] Summary Authors propose a new framework of loss functions, motivated by the Taylor series expansion of commonly used functions like cross entropy

Short NLP Overview!

Short NLP Overview!

**Natural language processing (NLP)** is a branch of science sitting at the intersection of computer science, artificial intelligence, and computational linguistics. To put it simply, it gives computers the ability to read, understand and generate text and speech just like we do

My New Desktop!

My New Desktop!

If you're into gaming and deep learning, you need to own a GPU. For years I was working with an older GPU (GTX 960M), but I thought it was time to upgrade. But wait, should I get a laptop, a pre-built desktop, or just build my own ? It

EfficientNet (V1 & V2) - A smart heuristic

EfficientNet (V1 & V2) - A smart heuristic

EfficientNet tries to come up with a smart heuristic to scale a CNN, relating resolution, width, and depth of a CNN.

Designing and shipping a ML Feature

Designing and shipping a ML Feature

What exactly are we trying to accomplish? Will the new model architecture really be a game-changer? How much impact will this new dataset have on the model accuracy? These are some questions a typical machine learning team deals with on a daily basis. But it is important to not get

Semantic Segmentation - DeepLab V3+

Semantic Segmentation - DeepLab V3+

Semantic Segmentation Semantic segmentation involves partitioning/marking regions in the image belonging to different objects/classes. Deep learning methods have made a remarkable improvement in this field within the past few years. This short article summarises DeepLab V3+, an elegant extension of DeepLab v3 proposed by the same authors (Chen

CPU vs GPU vs TPU

CPU vs GPU vs TPU

CPU vs GPU vs TPU

Detection/Classification metrics

Detection/Classification metrics

Problem Statement 1. Once we've trained multiple detection/classification models, how to choose the best model ? 2. Once we've chosen the best model, how to choose the optimum operating point (the best threshold) ? Solution There are two metrics which can help us here. If you'