Dissertation Talk: Towards Socially and Economically Beneficial Machine Learning

310 Soda Hall
  • Wenshuo Guo
I will give an overview of my thesis work on developing learning ​algorithms that incorporate the underlying social economic ​contexts into the design from three aspects: (i) Learning with noisy input data; (ii) Learning with bandit-type user feedback; (iii) Learning under causal dynamics.

Dissertation Talk: Miniature Wireless Neural Implants

SWARM Lab/Immersion Room Cory Hall
  • Mohammad Meraj Ghanbari
The tethered nature of wired neural implants limits their lifetime to few weeks due to factors such as infection and electrode dislocation. In order to increase the lifetime of these implants and extend their application to beyond neuroscience research, we have implemented two miniature implants that are completely wireless. The design and characterization of these miniature implants are the...

EECS Colloquium: What is a Codec Avatar?

306 (HP Auditorium) Soda Hall
  • Yaser Sheikh, Meta Reality Lab
Long-distance communication is a revolution that has slowly unfolded over the past 150 years, sparked by the invention of the telegraph, followed by telephonic communication, and most recently by the broad usage of video calling. The advent of 3D displays, such as VR headsets and AR glasses, presents the opportunity for the next step in this revolution: telepresence. In this talk, I will present...

Dissertation Talk: Object and Scene Reconstruction Using Neural Radiance Fields

8019 Berkeley Way West
  • Matt Tancik
Generating photorealistic 3D representations of scenes is important for many downstream vision tasks. In this talk I will discuss Neural Radiance Fields (NeRFs), a representation that enables novel view synthesis of complex scenes by optimizing an underlying continuous volumetric scene function using a sparse set of input views. In the past two years these representations have received interest...

Dissertation Talk: The Design and Implementation of User-Schedulable Languages

  • Alex Reinking
This thesis details an emerging class of programming language designs, called user-schedulable languages, that provide a safe and productive performance engineering environment for modern, heterogeneous hardware. The defining trait of user-schedulable languages is the division of program specification into two key parts: the algorithm, which defines functionally...

Dissertation Talk: Network Optimization Algorithms and Applications to Molecular Biology

  • Alex Khodaverdian
In this talk, I pursue several problems in molecular biology through abstraction into network optimization algorithms. In the first half of this talk I consider sub-network optimization within a known dynamic network. In these instances, I introduce the notion of Condition and Time Conditioned networks, in which a network may dynamically change over time (i.e. vertices or edges). In our first...

Dissertation Talk: Neural Prosthetic with In-sensor Shared Control

Open Forum 2108 Allston Way (Berkeley Wireless Research Center)
  • Alisha Menon, UC Berkeley
In this talk, a hardware-efficient shared control scheme for assistive myoelectric prosthetic devices is presented.

Dissertation Talk: Resource-Constrained Sensing as a Utility

  • Joshua Adkins
Cloud computing revolutionized the ease with which we can build, deploy, and scale distributed computing services. These advances, however, did not extend to the physically distributed and resource-constrained computers deployed throughout the world to collect data. Despite the many benefits that would come from including these sensors in a more dynamic distributed system, their resource...
Climate change requires an intricately complex energy transition, and innovations in alternative energy generators require innovations in supporting technologies. Demand Response is one such important supporting technology, and automated prices may increase is effectiveness. We present two new RL environments that broadcast prices at different levels of the grid to test transactive control of...