Ensemble methods are considered the state-of-the art solution for many machine learning challenges. Such methods improve the predictive performance of a single model by training multiple models and combining their predictions. This paper introduce the concept of ensemble learning, reviews traditional, novel and state-of-the-art ensemble methods and discusses current challenges and trends in the field.
Machine learning has a huge potential to improve products, processes and research. But machines usually don’t give an explanation for their predictions, which hurts trust and creates a barrier for the adoption of machine learning. This book is about making machine learning models and their decisions interpretable. Machine learning models are already used to choose the best advertisement for you, it filters out spam from your emails and it even assesses risk in the judicial system which ultimately can have consequences for your freedom. Can everyone trust the learned model? The model might perform well on the training data, but are the learned associations general enough to transfer to new data? Are there some oddities in the training data which the machine learning model dutifully picked up? This book will give you an overview over techniques that you can use to make black boxes as transparent as possible and make their predictions interpretable. The first part of the book introduces simple, interpretable models and instructions how to do the interpretation. The later chapters focus on general model-agnostics tools that help analysing complex models and making their decisions interpretable. In an ideal future, machines will be able to explain their decisions and the algorithmic age we are moving towards will be as human as possible. This books is recommended for machine learning practitioners, data scientists, statisticians and anyone else interested in making machine decisions more human.
The purpose of this technical report is two-fold. First of all, it introduces a suite of challenging continuous control tasks (integrated with OpenAI Gym) based on currently existing robotics hardware. The tasks include pushing, sliding and pick & place with a Fetch robotic arm as well as in-hand object manipulation with a Shadow Dexterous Hand. All tasks have sparse binary rewards and follow a Multi-Goal Reinforcement Learning (RL) framework in which an agent is told what to do using an additional input. The second part of the paper presents a set of concrete research ideas for improving RL algorithms, most of which are related to Multi-Goal RL and Hindsight Experience Replay.
Big Data revolution led to the explosion in Data Centers, which are consuming energy at increasingly higher rate. This blog reviews 2 standard methods for improving data center efficiency and argues that 3rd method – Machine Learning – is the best solution.
Data analytics is rapidly changing the face of manufacturing as we know it. At Mango, we’re seeing companies using their data effectively to gain an advantage over competitors. These companies are using data science to properly set up and control manufacturing For example, automatically adjusting parameters for specific parts/production lines to decrease wastage and meet demand. Research has shown that 68% of manufacturers were already investing in data science to achieve a range of improvements. This means that more than 30% of manufacturers still haven’t adopted a data-driven approach and are therefore not yet working leaner, smarter, improving yields and reducing costs for an increased bottom line. We know that manufacturing is an asset-intensive industry and companies need to move fast, be more innovative and work smart in order to be competitive. To remain ahead of the game, manufacturers need to adopt a different way of thinking when it comes to data. However, any transition from the industrial to the digital age can be both daunting and a minefield.
In this 4th post of my series on Deep Learning from first principles in Python, R and Octave – Part 4, I explore the details of creating a multi-class classifier using the Softmax activation unit in a neural network.
I’ve just updated The Popularity of Data Science Software to reflect my take on Gartner’s 2018 report, Magic Quadrant for Data Science and Machine Learning Platforms.
The world of using static tooling for packaging, provisioning, deployments, and monitoring, APM and log management will be over. With Docker adoption, the Cloud and API driven approaches and micro-services to deploying applications at a large scale, ensuring high reliability, requires an excellent take. So, it’s essential to include creative managing tools for cloud instead of reinventing the wheel every time. With the rise of ML and AI, more DevOps tooling vendors are incorporating intelligence with their offerings for further simplifying the task of engineers.
Getting started in data science can be overwhelming, especially when you consider the variety of concepts and techniques a data scienctist needs to master in order to do her job effectively. Even the term ‘data science’ can be somewhat nebulous, and as the field gains popularity it seems to lose definition.