Gradient Boosting explained [demonstration]
div.demo-wrapper { width: 800px; margin: auto; } div.visualization-wrapper { font-size: 14px; margin-top: 0.5em; margin-bottom: 0.5em; } div.visualization { margin: auto; /*width: 800px;*/ }...
View ArticleGradient Boosting Interactive Playground
Dataset to classify: Prediction: ↑ train loss: test loss: Decision functions of first trees tree depth: learning rate: rotate dataset: rotate trees show gradients on hover use...
View ArticleMLHEP 2016 lectures slides
This year my team at Yandex organized MLHEP (Machine Learning in High Energy Physics) summer school in Lund, Sweden. There were two tracks: basic and advanced, lasting for three days + 2 days on neural...
View ArticleOccam razor vs. machine learning
Whenever possible, substitute constructions out of known entities for inferences to unknown entities Occam's razor (Russell's version) If Russell was studying Machine Learning our days, he’d probably...
View ArticleJupyter (IPython) notebooks features
Jupyter (IPython) notebooks features¶It is very flexible tool to create readable analyses, because one can keep code, images, comments, formula and plots together: Jupyter is quite extensible,...
View ArticleHamiltonian Monte Carlo explained
div.demo-wrapper { width: 800px; margin: auto; } div.visualization-wrapper { font-size: 14px; margin-top: 0.5em; margin-bottom: 0.5em; } div.controls { margin: auto; padding: 10px 5px;...
View ArticleMachine Learning in Science and Industry slides
Recently we (Alex & Tatiana) were invited to give lectures about machine learning at GradDays — an event that is organized twice a year at the Heidelberg University (Germany’s oldest university)....
View ArticleMachine learning applied to showers in the OPERA
Abstract: in this post I explain how Machine Learning tools can be applied to particle physics. I'll discuss what is a particle shower, and when it appears in the OPERA and how clustering and...
View ArticleClustering applied to showers in the OPERA
Abstract: in this post I discuss clustering: techniques that form this method and some peculiarities of using clustering in practice. This post continues previous one about the OPERA. Update: now you...
View ArticleEinops — a new style of deep learning code
Recently I’ve open-sourced einops — a new (and better) way to write deep learning code. Einops introduces a new notation and new operations. It perfectly complements existing frameworks (pytorch,...
View ArticleTwin training: trick for better model comparisons
Abstract: Frequently comparing deep learning models? A simple way to improve comparison is discussed here, this trick becomes specially handy when comparing segmentation models.Reliable comparison of...
View ArticleNumbers that lie to you
Either you don’t understand what they mean or how those were cooked.Scientist achieved 92% quality in the problem of …Our system filters out 99% of spam …New vaccine/technique is N% more efficient…This...
View ArticleJupyter (IPython) notebooks features
Jupyter (IPython) notebooks features¶It is very flexible tool to create readable analyses, because one can keep code, images, comments, formula and plots together:Jupyter is quite extensible, supports...
View ArticleHamiltonian Monte Carlo explained
div.demo-wrapper { width: 800px; margin: auto; } div.visualization-wrapper { font-size: 14px; margin-top: 0.5em; margin-bottom: 0.5em; } div.controls { margin: auto; padding: 10px 5px;...
View ArticleMachine Learning in Science and Industry slides
Recently we (Alex & Tatiana) were invited to give lectures about machine learning at GradDays— an event that is organized twice a year at the Heidelberg University (Germany’s oldest...
View ArticleMachine learning applied to showers in the OPERA
Abstract: in this post I explain how Machine Learning tools can be applied to particle physics. I'll discuss what is a particle shower, and when it appears in the OPERA and how clustering and...
View ArticleClustering applied to showers in the OPERA
Abstract: in this post I discuss clustering: techniques that form this method and some peculiarities of using clustering in practice. This post continues previous one about the OPERA. Update: now you...
View ArticleEinops — a new style of deep learning code
Recently I’ve open-sourced einops— a new (and better) way to write deep learning code.Einops introduces a new notation and new operations.It perfectly complements existing frameworks (pytorch,...
View ArticleTwin training: trick for better model comparisons
Abstract: Frequently comparing deep learning models? A simple way to improve comparison is discussed here, this trick becomes specially handy when comparing segmentation models.Reliable comparison of...
View ArticleNumbers that lie to you
Either you don’t understand what they mean or how those were cooked.Scientists achieved 92% quality in the problem of …Our system filters out 99% of spam …New vaccine/technique is 40% more efficient...
View Article