Energy, Learning, and Optimization
Nothing quite matches the delight of finding a 100x speedup in an optimization problem. Since my first year of grad school, I’ve been astounded by how small tweaks can produce huge performance differences. This post explores two main approaches to making linear programming problems faster: simplifying the problem formulation, » Read More
When my girlfriend was shopping for a new car in late 2024, I advocated getting an Electric Vehicle- until I realized that the effective carbon abatement cost would be $1,200 / tonne CO2-e abated, more than twice the cost of carbon capture and sequestration (per Charm Industrial). » Read More
Early blockchains served as decentralized ledgers: creating trust by guaranteeing that each new entry securely built on previous entries. Smart contracting systems like Ethereum added intelligence to on-chain operations, making it possible to run simple computer scripts. » Read More
PhinisheD! PhDone! I got my lollipop- after five years of classes, research, and lab presentations, I’ve finally finished the journey! And before I forget what it’s like to be in the swamp, I wanted to jot down thoughts about how I got through it. » Read More
I just spoke with Valentina Nakic, who is writing a thesis about the future of the blockchain+energy space, and was struck by the realization that there are some basic hurdles that will define the next 10-20 years of blockchain applications. » Read More
This tutorial walks through an example of how to implement ADMM on a blockchain, and complements my Devcon3 talk. Github repository is here
Goal: Use a smart contract to coordinate decentralized optimization¶
This can model federated machine learning or federated optimization problems
We will use the Alternating Direction Method of Multipliers, » Read More
Our blockchain/microgrid research might become a reality: working with the Blockchain@Berkeley team, we’ve submitted a proposal to the Wanxiang Global Blockchain Challenge, an international competition for blockchain infrastructure ideas. If all goes well, this research might turn into the actual metering and scheduling system for a city that is being planned in China. » Read More
Slick algorithms and whiz-bang math are nice to have, but sometimes nothing speaks as well as a hardware demo. And for energy nerds, this means flashing lights. Rather than wrangling an Arduino and working with LEDs, another option is to get a quick start using ‘smart’ » Read More
Blockchains are fascinating- a platform for robust, immutable, and trustless data storage and computation. But they’re also cumbersome- slow, inefficient, and expensive. So when are blockchains the right solution for a problem? Blockchains are not a magical new type of computing. » Read More
I help organize Cleanweb Berkeley, a student group on-campus that raises awareness of cleantech careers for data scientists, and facilitates research dealing with data and energy/electricity systems.
Over the years, we’ve tried a number of different formats; here are a few of the lessons we’ve learned:
» Read More