https://blog.sksoumik.com/tags/machine-learning/ Recent content in machine learning on Sadman Kabir Soumik Hugo -- gohugo.io en Copyright © 2022, Sadman Kabir Soumik Sun, 14 Dec 2025 18:00:00 +0000 https://blog.sksoumik.com/artificial-intelligence/how-llm-post-training-works/ Sun, 14 Dec 2025 18:00:00 +0000 https://blog.sksoumik.com/artificial-intelligence/how-llm-post-training-works/ Training a large language model usually happens in more than one stage. The first stage is pretraining. This is where the model learns language, facts, patterns, coding concepts, reasoning patterns, and many other things from a very large amount of data. But a pretrained model is not automatically a good assistant. It might know a lot, but it may still struggle to follow instructions, answer in the right format, understand human preferences, or behave the way we want. https://blog.sksoumik.com/artificial-intelligence/different-types-of-llms-architectures/ Tue, 18 Mar 2025 00:00:00 +0000 https://blog.sksoumik.com/artificial-intelligence/different-types-of-llms-architectures/ Let's first break it down: what exactly are large language models (LLMs), why do we call them 'large,' and how are they different from other types of language models? An LLM is a machine learning model trained on massive amounts of text using transformer-based architectures (or their variations). These models can generate, understand, and process human-like text, making them useful for tasks like translation, summarization, reasoning, coding etc,. How is an LLM different from other language models? https://blog.sksoumik.com/artificial-intelligence/airflow-machine-learning-pipeline/ Fri, 17 Feb 2023 00:00:00 +0000 https://blog.sksoumik.com/artificial-intelligence/airflow-machine-learning-pipeline/ Author: Sadman Kabir Soumik Let's first understand what's MLOps. What is MLOps? MLOps (Machine Learning Operations) is a set of practices and tools used to manage the entire lifecycle of machine learning models. MLOps includes everything from data preparation and model training to deployment, monitoring, and ongoing maintenance. The primary goal of MLOps is to create a streamlined and automated process for deploying and managing machine learning models at scale. https://blog.sksoumik.com/artificial-intelligence/from-rnn-to-transformers-without-math/ Mon, 30 Jan 2023 00:00:00 +0000 https://blog.sksoumik.com/artificial-intelligence/from-rnn-to-transformers-without-math/ Transformer-based models are a types of neural network architecture that uses self-attention mechanisms to process input data. They were introduced in the paper "Attention Is All You Need" by Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N. Gomez, Lukasz Kaiser, Illia Polosukhin in 2017, and have since become a popular choice for many natural language processing task. Prerequisite: Before going further, I assume that you have a basic understanding of how neural networks work. https://blog.sksoumik.com/artificial-intelligence/removing-background-selfie-segmentation/ Mon, 16 Jan 2023 00:00:00 +0000 https://blog.sksoumik.com/artificial-intelligence/removing-background-selfie-segmentation/ Project Goal The project aims to perform image segmentation on selfie images, see how we can blur the image's background, and even replace the background with some other solid colour like black. We will use a framework called MediaPipe to accomplish this task. About MediaPipe MediaPipe is an open-source framework developed by Google that allows developers to build and deploy cross-platform multimodal machine learning models. The framework provides a set of reusable components for tasks such as object detection, hand tracking, facial landmark detection, and gesture recognition. https://blog.sksoumik.com/artificial-intelligence/frequently-asked-data-science-interview-qestions/ Tue, 15 Nov 2022 00:00:00 +0000 https://blog.sksoumik.com/artificial-intelligence/frequently-asked-data-science-interview-qestions/ Can you explain the bias-variance trade-off and how it relates to model performance? Machine learning and statistics have a fundamental concept that requires balancing the model's bias and variance, known as the bias-variance trade-off. These two types of errors can affect a model's performance. Bias, a type of error, occurs when a model makes assumptions about the data that are too simplistic. High bias means the model is too simple to capture the underlying patterns. https://blog.sksoumik.com/artificial-intelligence/common-machine-learning-algorithms/ Wed, 26 Oct 2022 00:00:00 +0000 https://blog.sksoumik.com/artificial-intelligence/common-machine-learning-algorithms/ Before jumping into Deep Learning, one must know the classical/traditional Machine Learning algorithms, because understanding traditional machine learning algorithms can provide a strong foundation in machine learning concepts. These algorithms often involve simple, intuitive concepts that can be helpful in understanding more complex deep learning models. Traditional machine learning algorithms can be faster to train and easier to interpret than deep learning models. This can be particularly useful in situations where you need to make quick decisions or where it's important to understand the reasoning behind a model's predictions. https://blog.sksoumik.com/artificial-intelligence/machine-learning-model-compression-techniques/ Mon, 10 Oct 2022 00:00:00 +0000 https://blog.sksoumik.com/artificial-intelligence/machine-learning-model-compression-techniques/ There are 4 main approaches you can consider for model compression. They are: Quantization Pruning Knowledge Distillation Low-Rank Factorization Quantization Quantization is the most general and commonly used model compression method. Quantization reduces a model’s size by using fewer bits to represent its parameters. By default, most software packages use 32 bits to represent a float number (single precision floating point). If a model has 100M parameters and each requires 32 bits to store, it’ll take up 400 MB. https://blog.sksoumik.com/artificial-intelligence/multi-class-text-classification-apache-spark-mllib/ Tue, 24 May 2022 00:00:00 +0000 https://blog.sksoumik.com/artificial-intelligence/multi-class-text-classification-apache-spark-mllib/ Spark MLlib MLlib is a library for machine learning in Spark that aims to make it easy to use and scalable for practical applications. It includes tools for common ML tasks, such as classification, regression, clustering, and collaborative filtering, as well as featurization methods for feature extraction, transformation, dimensionality reduction, and selection. MLlib also provides tools for building, evaluating, and tuning ML pipelines, as well as utilities for linear algebra, statistics, and data handling. https://blog.sksoumik.com/artificial-intelligence/keyphrase-extraction-techniques-with-bert-embeddings-pos-patterns/ Thu, 19 May 2022 00:00:00 +0000 https://blog.sksoumik.com/artificial-intelligence/keyphrase-extraction-techniques-with-bert-embeddings-pos-patterns/ Keyphrases are important pieces of information that can be extracted from text documents. These are words or phrases that summarize the main ideas or topics of a text, and they can be useful for a variety of applications, such as document summarization, text classification, and information retrieval. In this blog post, we will explore how keyphrases can be extracted from text documents, and discuss some of the techniques and tools that can be used for this task. https://blog.sksoumik.com/artificial-intelligence/data-normalization-vs-standardization-machine-learning/ Sat, 12 Mar 2022 00:00:00 +0000 https://blog.sksoumik.com/artificial-intelligence/data-normalization-vs-standardization-machine-learning/ Why do we scale features? For machine learning, every dataset does not require feature scaling, and it is only needed when features have different ranges. For example, consider a data set containing two features, age(x1) and income(x2), where age ranges from 0–100, while income ranges from 0–20,000 and higher. Income is about 1,000 times larger than age and ranges from 20,000–500,000. So, these two features are in very different ranges. When we do further analysis, like multivariate linear regression, the attributed income will influence the result more due to its larger value. https://blog.sksoumik.com/artificial-intelligence/one-stage-two-stage-segmentation-difference/ Fri, 04 Mar 2022 00:00:00 +0000 https://blog.sksoumik.com/artificial-intelligence/one-stage-two-stage-segmentation-difference/ In computer vision, image segmentation refers to the process of dividing an image into distinct regions or segments, each corresponding to a different object or background. There are two main approaches to image segmentation: one-stage and two-stage. One-stage image segmentation methods aim to directly predict a segmentation mask for the entire image in a single pass. These methods are typically faster and more efficient than two-stage methods, but they may be less accurate and less flexible. https://blog.sksoumik.com/artificial-intelligence/machine_learning_research_vs_production/ Mon, 10 Jan 2022 00:00:00 +0000 https://blog.sksoumik.com/artificial-intelligence/machine_learning_research_vs_production/ There are several key differences between using machine learning for research and using it for production. One of the main differences is the focus of the work. Machine learning for research typically focuses on exploring new ideas and techniques, and on advancing the state of the art in the field. In contrast, machine learning for production focuses on building practical, real-world applications that can deliver value to organizations and individuals. https://blog.sksoumik.com/artificial-intelligence/writing-model-keras-vs-pytorch/ Thu, 09 Dec 2021 00:00:00 +0000 https://blog.sksoumik.com/artificial-intelligence/writing-model-keras-vs-pytorch/ PyTorch and Keras are both open-source deep learning frameworks, but they have some significant differences. PyTorch is a low-level framework that allows you to define your own computation graphs, while Keras is a high-level framework that provides a pre-defined set of layers and routines for building deep learning models. This means that PyTorch offers more flexibility and customization, while Keras is easier to use and more accessible to beginners. Choosing Framework - PyTorch vs. https://blog.sksoumik.com/artificial-intelligence/a_brief_introduction_on_openai_gpt_3/ Sat, 20 Nov 2021 00:00:00 +0000 https://blog.sksoumik.com/artificial-intelligence/a_brief_introduction_on_openai_gpt_3/ Author: Sadman Kabir Soumik GPT-3, or Generative Pretrained Transformer 3, is a state-of-the-art language model developed by OpenAI. It has been trained on a massive amount of text data, including books, articles, and websites, to generate coherent and relevant text based on a given context. GPT-3 is a transformer-based model, which means that it uses a type of neural network architecture called a transformer to process the input text. This allows the model to capture long-range dependencies and generate text that is more coherent and human-like than previous models. https://blog.sksoumik.com/artificial-intelligence/vanishing-gradient_problem/ Sun, 24 Oct 2021 00:00:00 +0000 https://blog.sksoumik.com/artificial-intelligence/vanishing-gradient_problem/ What is Vanishing Gradient Problem Neural networks are trained using stochastic gradient descent. This involves first calculating the prediction error made by the model and using the error to estimate a gradient used to update each weight in the network so that less error is made next time. This error gradient is propagated backward through the network from the output layer to the input layer. As the backpropagation algorithm advances downwards (or backward) from the output layer towards the input layer, the gradients often get smaller and smaller and approach zero, eventually leaving the weights of the initial or lower layers nearly unchanged. https://blog.sksoumik.com/artificial-intelligence/bagging-boosting-stacking-blending-bayesian-averaging/ Thu, 10 Jun 2021 00:00:00 +0000 https://blog.sksoumik.com/artificial-intelligence/bagging-boosting-stacking-blending-bayesian-averaging/ There are several types of ensemble techniques in machine learning, including: Bagging, Boosting, Stacking, Blending, Bootstrapped ensembles, Bayesian model averaging. Bagging Bagging (short for bootstrapped aggregating) is an ensemble technique that involves training multiple models on different subsets of the training data, and then averaging the predictions of the individual models to make the final prediction. This can be done with decision trees, neural networks, or any other type of model. https://blog.sksoumik.com/artificial-intelligence/comparing-random-forest-decision-tree-gradient-boosting/ Sat, 27 Mar 2021 00:00:00 +0000 https://blog.sksoumik.com/artificial-intelligence/comparing-random-forest-decision-tree-gradient-boosting/ Decision Tree, Random Forest (RF), and Gradient Boosting (GB) are three popular algorithms used for supervised learning tasks such as classification and regression. In this blog, we will compare these three algorithms in terms of their features, performance, and usability. Decision Tree is a simple and intuitive algorithm that can be used for classification and regression tasks. A Decision Tree model is built by recursively partitioning the training data into smaller and smaller subsets based on the values of the input features. https://blog.sksoumik.com/artificial-intelligence/understanding-word-embeddings/ Fri, 11 Dec 2020 00:00:00 +0000 https://blog.sksoumik.com/artificial-intelligence/understanding-word-embeddings/ Word embedding is a technique in natural language processing (NLP) where words are represented as vectors of real numbers. This allows words with similar meanings to have similar representation, and can be used in various NLP tasks such as machine translation and text classification. There are several different techniques for word embedding in natural language processing (NLP), including: TF-IDF — Term Frequency-Inverse Document Frequency TF-IDF (Term Frequency-Inverse Document Frequency) is a technique used in natural language processing to measure the importance of a word in a document. https://blog.sksoumik.com/artificial-intelligence/understanding-recurrent-neural-networks/ Sun, 25 Oct 2020 00:00:00 +0000 https://blog.sksoumik.com/artificial-intelligence/understanding-recurrent-neural-networks/ Recurrent Neural Network A recurrent neural network (RNN), is a type of neural network that can process sequential data, like text, audio, or time series data. Here's how it works: first, the RNN takes in some input data, which could be a word in a sentence, a sound wave from an audio recording, or a measurement from a sensor at a specific time. Then, the RNN processes this input and generates an output, which could be a predicted next word in a sentence, a generated audio waveform, or a predicted sensor measurement. https://blog.sksoumik.com/artificial-intelligence/types-of-recommender-systems-machine-learning/ Sat, 05 Oct 2019 00:00:00 +0000 https://blog.sksoumik.com/artificial-intelligence/types-of-recommender-systems-machine-learning/ There are several different types of recommender systems, each with its own unique characteristics and applications. Some of the most commonly used types of recommender systems include: Content-based recommender systems: These systems recommend items to users based on the characteristics of the items themselves. For example, a content-based recommender system for a movie website might recommend movies to users based on the genre, director, or actor. Collaborative filtering recommender systems: These systems make recommendations to users based on the preferences of similar users. https://blog.sksoumik.com/artificial-intelligence/tips-to-avoid-overfitting-machine-learning/ Thu, 09 May 2019 00:00:00 +0000 https://blog.sksoumik.com/artificial-intelligence/tips-to-avoid-overfitting-machine-learning/ Very deep neural networks with a massive number of parameters are very robust machine learning systems. But, in this type of huge network, overfitting is a common serious problem. Learning how to deal with overfitting is essential to mastering machine learning. The fundamental issue in machine learning is the tension between optimization and generalization. Optimization refers to adjusting a model to get the best performance possible on the training data (the learning in machine learning). https://blog.sksoumik.com/artificial-intelligence/guide-to-transfer-learning-techniques/ Mon, 21 May 2018 00:00:00 +0000 https://blog.sksoumik.com/artificial-intelligence/guide-to-transfer-learning-techniques/ Transfer learning is a technique in machine learning that allows a model trained on one task to be reused and fine-tuned for another similar task. The idea behind transfer learning is that a model that has already learned to recognize patterns in one set of data can be applied to a different but related problem, allowing the model to learn faster and with less data than if it were trained from scratch.