Projects

SeedBank — collection of interactive machine learning examples by Google. Examples of "seeds" – Variational auto-encoder for music, compare GAN, DeepDream etc. Zero Setup, GPU included.

Albumentations – fast image augmentation library by Kaggle Grandmasters (e.g., motion blur, perspective for street view, random crop shift, brightness etc).

Seq2Seq-Vis – a visual debugging tool for Sequence-to-Sequence models.

Geomstats – Python package for Riemannian geometry in machine Learning.

Research Papers and Books

"Bandit Algorithms" – Book Draft by Tor Lattimore and Csaba Szepesvari. Cambridge University Press.

"Towards Understanding the Role of Over-Parametrization in Generalization of Neural Networks". Present a new capacity bound that decreases with the increasing number of hidden units and may explain better generalization of larger networks.

"ClariNet: Parallel Wave Generation in End-to-End Text-to-Speech". First fully end-to-end TTS model that directly converts text into waveform with a single neural network.

"Through-Wall Human Pose Estimation Using Radio Signals" – WiFi frequencies traverse walls and reflect off the human body. Deep Neural network parses such radio signals to estimate 2D poses. MIT CSAIL.

"Deep Video Portraits" – photo-realistic reanimation of portrait videos. Transfer full 3D head pose, facial expression, eye gaze and blinking.

Posts, Articles, Tutorials

Multi-Modal Methods: Image Captioning – recent intersections between computer vision and natural language processing.

The Annotated Transformer – annotated version of the paper “Attention is All You Need” in the form of a line-by-line implementation by Harvard NLP group.

Papers with Code – machine learning papers with code.

How to explain gradient boosting – explains the intuition, provides visualizations for model construction, explains the mathematics as simply as possible.

Transfer Learning in Natural Language Processing by Intel AI Academy.

Video Lectures and Talks

CVPR 2018: Main Conference, Tutorials, Workshops