In the previous post, we looked at Prior-Data Fitted Networks (PFNs): a method that sidesteps the intractability of Bayesian inference by training a Transformer to directly approximate the posterio...
Prior-Fitted Networks: Teaching Transformers to Do Bayesian Inference
In the previous post, we worked through the foundations of Bayesian inference. We saw that the posterior $p(t \mid \mathcal{D})$ is almost always intractable, and that the two main tools for approx...
From Prior to Posterior: A Guide to Bayesian Inference
Bayesian inference is one of the most principled frameworks for reasoning under uncertainty. At its core, it is about updating your beliefs when you see new data. In this post, we will build up the...
Understanding GARCH Models, A Beginner-Friendly Guide with Python Implementation
Have you ever noticed that stock prices or exchange rates tend to behave in clusters? For example, periods of calm with small price changes are often followed by periods of high activity with big j...
Exploring AI for Finance, A New Chapter in My Research Journey
As an AI researcher, I’ve always been fascinated by how machine learning and artificial intelligence can solve complex, real-world problems. Recently, I’ve found myself drawn to a new area of explo...
The Bitter Truth About AI; Why Human Ingenuity Often Loses to Computation
This blog post is written based on Rich Sutton’s article, “The Bitter Lesson”. Ah, the fascinating world of artificial intelligence! It’s a place where robots don’t exactly plot world domination ...
FTTransformer; Transformer Architecture for Tabular Datasets
Introduction If you follow my blog, you’ve probably noticed my keen interest in deep learning for tabular data. It’s not because I find tasks like predicting housing prices fascinating—I don’t! My...
Self-Supervised Representation Learning for Tabular Datasets
Self-supervised Learning for Tabular Datasets Self-supervised learning aims to learn latent representations for unlabeled datasets. It has shown to be an effective representation learning method; ...
Self-supervision and Collapsing Solutions
Collapsing Solutions in Self-supervised Learning In the previous post I explained how self-supervised learning has been established as a decent method for unsupervised representation learning. I d...
Self-Supervised Learning
Self-Supervision Introduction Self-supervised learning (SSL) is rapidly closing the gap with supervised methods. Very recently, Facebook AI Research (FAIR), one major player in broadening the hori...