Posts by Hamza FILALI

Mistral 7B: Decoding the Complexities of a Large Autoregressive Language Model

The goal of this post is to offer a clear understanding of how large language models like Mistral work ([Jiang et al., 2023]).

../../_images/mistral_architecture.png

Read more ...


Detecting Trend Changes in Time-Series Data: A Frequentist and Parametric Approach

Detecting trend changes in time-series can offer valuable insights for various applications. There are multiple ways to approach this problem, and each method comes with its own set of assumptions and intricacies.

../../_images/trend_turning.jpg

Read more ...


Why Linear Regression is Called “Regression”: regression towards mediocrity

Linear regression is a widely used statistical technique that aims to model the relationship between a dependent variable and one or more independent variables. It plays a crucial role in various fields, including economics, social sciences, healthcare, and engineering. But have you ever wondered why this powerful modeling technique is called “regression”?

../../_images/table_galton.png

Read more ...


How to reduce the amount of labelled data required for your Deep Learning model with a human in the loop?

In this blog post, we will explore how Bayesian active learning can be used to reduce the amount of labeled data required for a deep learning model with a human in the loop in the context of a classification task.

../../_images/active-learning-loop.gif

Read more ...


Dropout as a way to Model Uncertainty in Deep Learning Models

In [Gal and Ghahramani, 2016], the authors propose an easy, fast and scalable method to quantify the uncertainty of any deep learning model (including big modern architectures) as long as dropout is used during training time.

../../_images/gif_dropout_final.gif

Read more ...


Notes on Bayesian Online Change Point Detection

In this post we are going to delve into the mathematical details behind the graphical model Bayesian Online Change Point Detection introduced in [Adams and MacKay, 2007]. This model can be used to detect different type of change-points and has known many extensions over the last few years.

../../_images/bocpd_graphical_model.png

Read more ...