Algorithms

TbD-net solves the visual reasoning problem by breaking it down to a chain of subtasks. The answer to each subtask is shown in heat maps highlighting the objects of interest, allowing analysts to see the network's thought process.

Artificial intelligence system uses transparent, human-like reasoning to solve problems

Model from MIT Lincoln Laboratory Intelligence and Decision Technologies Group sets a new standard for understanding how a neural network makes decisions.

September 11, 2018

MIT researchers have developed an algorithm that traces intersections in sketches without error. This could save digital artists significant time and frustration when vectorizing an image for animation, marketing logos, and other applications.

Smoothing out sketches’ rough edges

MIT-developed tool improves automated image vectorization, saving digital artists time and effort.

September 11, 2018

Members of a team developing Adaptable Interpretable Machine Learning at Lincoln Laboratory are: (l-r) Melva James, Stephanie Carnell, Jonathan Su, and Neela Kaushik.

Taking machine thinking out of the black box

Adaptable Interpretable Machine Learning project is redesigning machine learning models so humans can understand what computers are thinking.

September 5, 2018

CSAIL researchers have developed a visualization tool for CAD that, for the first time, lets users instead interactively explore all designs that best fit multiple, often-conflicting performance tradeoffs, in real time.

Design tool reveals a product’s many possible performance tradeoffs

Users can quickly visualize designs that optimize multiple parameters at once.

August 15, 2018

MIT researchers aim to improve the quality of life for patients suffering from glioblastoma, the most aggressive form of brain cancer, with a machine-learning model that makes chemotherapy and radiotherapy dosing regimens less toxic but still as effective as human-designed regimens.

Artificial intelligence model “learns” from patient data to make cancer treatment less toxic

Machine-learning system determines the fewest, smallest doses that could still shrink brain tumors.

August 9, 2018

Computer scientists from MIT and IPRI propose a new cryptographic system to improve accountability of government surveillance while maintaining enough confidentiality for the police to do their jobs.

Holding law-enforcement accountable for electronic surveillance

CSAIL system encourages government transparency using cryptography on a public log of wiretap requests.

August 8, 2018

MIT researchers have developed a machine-learning model that better selects molecule candidates for therapeutics, while also allowing for automated modification of the molecular structure for higher potency. The innovation has potential to speed up drug development.

Automating molecule design to speed up drug development

Machine-learning model could help chemists make molecules with higher potencies, much more quickly.

July 6, 2018

A new AI system can look at an image and determine which set of pixels are responsible for making specific sets of soundwaves.

An AI system for editing music in videos

Given a video of a musical performance, CSAIL’s deep-learning system can make individual instruments louder or softer.

July 5, 2018

MIT’s Cheetah 3 robot can climb stairs and step over obstacles without the help of cameras or visual sensors.

“Blind” Cheetah 3 robot can climb stairs littered with obstacles

Improved design may be used for exploring disaster zones and other dangerous or inaccessible environments.

July 4, 2018

A system developed at MIT allows a human supervisor to correct a robot's mistakes using gestures and brainwaves.

How to control robots with brainwaves and hand gestures

Computer Science and Artificial Intelligence Laboratory system enables people to correct robot mistakes on multiple-choice tasks.

June 20, 2018

A new computer chip, smaller than a U.S. dime and shown here with a quarter for scale, helps miniature drones navigate in flight.

Chip upgrade helps miniature drones navigate

Low-power design will allow devices as small as a honeybee to determine their location while flying.

June 20, 2018

MIT researchers describe a machine-learning algorithm that can register brain scans and other 3-D images more than 1,000 times more quickly using novel learning techniques.

Faster analysis of medical images

Algorithm makes the process of comparing 3-D scans up to 1,000 times faster.

June 18, 2018

A team of MIT engineers has developed a “smart power outlet” in the form of a device that can analyze electrical current usage from a single or multiple outlets.

MIT engineers build smart power outlet

Design can “learn” to identify plugged-in appliances, distinguish dangerous electrical spikes from benign ones.

June 15, 2018

Teaching robots how to move objects

PhD candidate and Amazon Robotics Challenge winner Maria Bauza helps to improve how robots interact with the world.

June 12, 2018

Using wireless signals, RF-Pose could serve as a health care system to monitor patients' movements from the other side of a wall.

Artificial intelligence senses people through walls

Wireless smart-home system from the Computer Science and Artificial Intelligence Laboratory could monitor diseases and help the elderly “age in place.”