How to tell whether machine-learning systems are robust enough for the real world
New method quickly detects instances when neural networks make mistakes they shouldn’t.
Wireless movement-tracking system could collect health and behavioral data
In some cases, radio frequency signals may be more useful for caregivers than cameras or other data-collection methods.
Smarter training of neural networks
MIT CSAIL project shows the neural nets we typically train contain smaller “subnetworks” that can learn just as well, and often faster.
Merging cell datasets, panorama style
Algorithm stitches multiple datasets into a single “panorama,” which could provide new insights for medical and biological studies.
New approach could accelerate efforts to catalog vast numbers of cells
Data-sampling method makes “sketches” of unwieldy biological datasets while still capturing the full diversity of cell types.
Improving security as artificial intelligence moves to smartphones
Researchers unveil a tool for making compressed deep learning models less vulnerable to attack.
Can science writing be automated?
A neural network can read scientific papers and render a plain-English summary.
Vivienne Sze wins Edgerton Faculty Award
EECS faculty member is recognized for technical innovation, educational excellence, and efforts to advance women and underrepresented minorities in her field.
Giving robots a better feel for object manipulation
Model improves a robot’s ability to mold materials into shapes and interact with liquids and solid objects.
Robots that can sort recycling
CSAIL’s "RoCycle" system uses in-hand sensors to detect if an object is paper, metal or plastic.
A novel data-compression technique for faster computer programs
Researchers free up more bandwidth by compressing “objects” within the memory hierarchy.
Machine learning moves popular data elements into a bucket of their own
Counting search queries isn’t easy, but MIT CSAIL’s new LearnedSketch system for “frequency-estimation” aims to help.
Advance boosts efficiency of flash storage in data centers
New architecture promises to cut in half the energy and physical space required to store and manage user data.
Teaching machines to reason about what they see
Researchers combine statistical and symbolic artificial intelligence techniques to speed learning and improve transparency.
Model learns how individual amino acids determine protein function
Technique could improve machine-learning tasks in protein design, drug testing, and other applications.