A Fast Parallel Algorithm for HDBSCAN* Clustering

A research library for pytorch-based neural network pruning, compression, and more.

A framework for analyzing computer vision models with simulated data

Implementation of the CVPR 2019 Paper - Speech2Face: Learning the Face Behind a Voice by MIT CSAIL

The focus of recent meta-learning research has been on the development of learning algorithms that can quickly adapt to test time tasks with limited data and low computational cost. Few-shot learning is widely used as one of the standard benchmarks in meta-learning. In this work, we show that a simple baseline: learning a supervised or self-supervised representation on the meta-training set, followed by training a linear classifier on top of this representation, outperforms state-of-the-art few-shot learning methods. An additional boost can be achieved through the use of self-distillation.

The project is situated at MIT, in the Department of Mechanical Engineering and the Center for Ocean Engineering as part of the Laboratory for Autonomous Marine Sensing Systems (LAMSS). Core developers are also part of the MIT Computer Science and Artificial Intelligence Lab, (CSAIL). Core MOOS software is maintained and distributed by the Oxford Robotics Institute (ORI).

MOOS stands for "Mission Oriented Operating Suite". IvP stands for "Interval Programming". MOOS-IvP is pronounced "moose i-v-p".

This repository contains the demo code for the CVPR'17 paper Network Dissection: Quantifying Interpretability of Deep Visual Representations. You can use this code with naive Caffe, with matcaffe and pycaffe compiled. We also provide a PyTorch wrapper to apply NetDissect to probe networks in PyTorch format.

GAN Dissection is a way to inspect the internal representations of a generative adversarial network (GAN) to understand how internal units align with human-interpretable concepts. It is part of NetDissect.

Ligra is a lightweight framework for processing graphs in shared memory. It is particularly suited for implementing parallel graph traversal algorithms where only a subset of the vertices are processed in an iteration. The project was motivated by the fact that the largest publicly available real-world graphs all fit in shared memory. When graphs fit in shared-memory, processing them using Ligra can give performance improvements of up orders of magnitude compared to distributed-memory graph processing systems.

FALCONN is a library with algorithms for the nearest neighbor search problem. The algorithms in FALCONN are based on Locality-Sensitive Hashing (LSH), which is a popular class of methods for nearest neighbor search in high-dimensional spaces. The goal of FALCONN is to provide very efficient and well-tested implementations of LSH-based data structures.