MachineLearning

Two heads are better than one. This proverb describes the concept behind ensemble methods in machine learning. Let's examine why ensembles dominate ML competitions and what makes them so powerful.

Harnessing the potential of machine learning for computer vision is not a new concept but recent advances and the availability of new tools and datasets have made it more accessible to developers. In this article, Toptal Software Developer Teimur Gasanov demonstrates how you can create an app capable of identifying handwritten digits in under 30 minutes, including the API and UI.

Working with non-numerical data can be challenging, even for seasoned data scientists. To make good use of such data, it needs to be transformed. But how? In this article, Toptal Data Scientist Yaroslav Kopotilov will introduce you to embeddings and demonstrate how they can be used to visualize complex data and make it usable.

TensorFlow is one of the leading tools for training deep learning models. Outside that space, it may seem intimidating and unnecessary, but it has many creative uses—like producing highly effective adversarial input for black-box AI systems.

Pre-trained models are making waves in the deep learning world. Using massive pre-training datasets, these NLP models bring previously unheard-of feats of AI within the reach of app developers.

For those working with AI, the future is certainly exciting. At the same time, there is a general sense that AI suffers from one pesky flaw: AI in its current state can be unpredictably unreliable.

IMDb ratings have genre bias: For example, dramas tend to score higher. Removing common feature bias and keeping unique characteristics, it's possible to create a new, refined score based on IMDb information.

Supervised learning is the key to computer vision and deep learning. However, what happens when you don’t have access to large, human-labeled datasets? In this article, Toptal Computer Vision Developer Urwa Muaz demonstrates the potential of semi-supervised image classification using unlabeled datasets.

For a successful natural language processing project, collecting and preparing data, building resilient pipelines, and getting "model ready" can easily take months of effort even with the most talented engineers. But what if we could reduce the data required to a fraction? In this article, we’ll cover how transfer learning is making world-class models open source and introduce BERT (bidirectional encoder representations from transformers). BERT is the most powerful NLP “tool” to date. We’ll explore how it works and why it will change the way companies execute NLP projects.