Deep Hunt — Issue #63
Here are the highlights from a super busy week — NeurIPS 2018 is here; The intertwined quest for understanding biological intelligence and creating artificial intelligence; Predicting real-time availability of 200 million grocery items in North American stores; Bag of Tricks for Image Classification with Convolutional Neural Networks
News
The biggest annual machine learning conference — NeuRIPS (which changed its short form from NIPS) is well underway this week at Montreal with a ton of announcements and presentations and talks and posters. Deep Hunt will publish a separate edition on it later.
Articles
The intertwined quest for understanding biological intelligence and creating artificial intelligence
Surya Ganguli writes about his personal vision of how neuroscience, psychology, AI, physics and mathematics and other fields can work together to both understand biological intelligence and create artificial intelligence!
Tutorials, Tools and Tips
This site is a really cool community driven initiative to keep track of the state of the art in various tasks and also includes datasets, metrics with categorization.
Adding Diversity to Images with Open Images Extended
At NeurRIPS ’18 Google announced Open Images Extended, a new branch of Google’s Open Images dataset, which is intended to be a collection of complementary datasets with additional images and/or annotations that better represent global diversity.
Predicting real-time availability of 200 million grocery items in North American stores
Instacart team takes us through how they built their grocery inventory prediction model with details on the various components of the architecture.
Research
Bag of Tricks for Image Classification with Convolutional Neural Networks
This paper examines a collection of training procedure refinements and empirically evaluates their impact on the final model accuracy through ablation studies. Turns out that were able to improve various CNN models significantly which also lead to better transfer learning performance in other application domains such as object detection and semantic segmentation.
UMAP: Uniform Manifold Approximation and Projection for Dimension Reduction
Uniform Manifold Approximation and Projection (UMAP) is the new alternative to t-SNE that many researchers are exploring for dimension reduction.
If you like what you are reading, please follow and recommend to your friends or give a shoutout on Twitter! I’m glad to hear your suggestions and recommendations @deephunt_in or in comments below!
