Sign in

Ed Clark

@clarkai.bsky.social
1.2K followers 594 following 53 posts

PhD student studying AI for tactical decision making in complex ocean environments @uniofbath.bsky.social Distinctly amateur cyclist, triathlete, sailor, climber scholar.google.com/citations?user=W…

PostsRepliesMedia
Ed Clark @clarkai.bsky.social · 13/02/2025
Does anybody know the level of interaction a general account on BlueSky has? Despite over 1000 followers I feel it's a lot lower than many other social platforms. Do you think this is a result of a large influx of non 'organic' growth? #meta #bluesky #askbluesky
340
Ed Clark @clarkai.bsky.social · 12/02/2025
Is there a way to shift my doomscrolling from Instagram to Bluesky? I know I'd end up more informed and positive from 10 minutes on here than on insta. Maybe some good suggestions of random fun accounts to follow?
010
Ed Clark @clarkai.bsky.social · 25/12/2024
Wallace and Grommit serving up some great PSA's on AI and automation. Doesn't just seem to play out the AI bad trope and shows to a general audience how it can go wrong quickly.
000
Ed Clark @clarkai.bsky.social · 04/12/2024
Just starting to build out my home office to have a nice space to push through the end of my PhD. What's your number one item to have in it? Can be anything from plants to monitors! 🌊🧪 #phdlife #academia
662
Ed Clark @clarkai.bsky.social · 01/12/2024
Sits down to write paper... 6 months later. This paper hasn't written itself yet no matter how many times I ask nicely... What are your tips for breaking through writing block and pushing out your paper? Could really do with some advice from the academic community! 🧪🌊
11222
Ed Clark @clarkai.bsky.social · 26/11/2024
Our oceans aren't quiet and they're getting louder. A great paper on how important sound is in the ocean and how humans are impacting it. This is not only important for the biosphere but louder oceans make bad actors harder to detect. 🧪 🌊 www.science.org/doi/10.1126/...
science.org
The soundscape of the Anthropocene ocean
Sound travels faster and farther in water than in air. Over evolutionary time, many marine organisms have come to rely on sound production, transmission, and reception for key aspects of their lives. ...
23110
Reposted by Ed Clark
Eugene Vinitsky 🍒 @eugenevinitsky.bsky.social · 23/11/2024
Kind of a broken record here but proceedings.neurips.cc/paper_files/... is totally fascinating in that it postulates two underlying, measurable structures that you can use to assess if RL will be easy or hard in an environment
e introduce the effective horizon, a property of
MDPs that controls how difficult RL is. Our analysis is mo-
tivated by Greedy Over Random Policy (GORP), a simple
Monte Carlo planning algorithm (left) that exhaustively ex-
plores action sequences of length k and then uses m random
rollouts to evaluate each leaf node. The effective horizon
combines both k and m into a single measure. We prove
sample complexity bounds based on the effective horizon that
correlate closely with the real performance of PPO, a deep
RL algorithm, on our BRIDGE dataset of 155 deterministic
MDPs (right).
815029
Ed Clark @clarkai.bsky.social · 23/11/2024
For those doing lit reviews, be aware that Scopus can now search for pre prints as well! Great for getting visibility of what's on the horizon in your field. 🧪
141
Reposted by Ed Clark
Rebecca Henderson @rebecca-henderson.bsky.social · 20/11/2024
The beginnings of a starter pack for UK-based PhD students. Please let me know if you'd like to be added! go.bsky.app/8Uj4vn7
211
Ed Clark @clarkai.bsky.social · 22/11/2024
What's your reinforcement learning stack? I use: - Custom gymnasium env - Stable baselines 3 - RL library - PPO - Algorithm - NatureCNN - Model network - wandb - run tracking - wandb sweeps - hyper-parameter optimization - tensorboard - offline tracking #AI #DRL #RL #Reinforcementlearning
2120
Ed Clark @clarkai.bsky.social · 22/11/2024
Yesterday's wintery work views. Having a large window right next to the desk has helped massively with fighting back the SAD this year. Take this post as your reminder to get some sunshine and fresh air today.
House plant in front of window with snowy background
020
Ed Clark @clarkai.bsky.social · 22/11/2024
Another must read for reinforcement learning. Answers many key questions for researchers; -Do I need multiple training runs? -How do I report model confidence? -And a great section on common mistakes to fend off reviewer 2 🧪 #DRL #reinforcementlearning #AI arxiv.org/abs/2304.01315
arxiv.org
Empirical Design in Reinforcement Learning
Empirical design in reinforcement learning is no small task. Running good experiments requires attention to detail and at times significant computational resources. While compute resources available p...
1385
Ed Clark @clarkai.bsky.social · 21/11/2024
One of my favourite DRL papers. OpenAI five, it highlights just how much compute and time is needed for training complex capable agents. Remember this is learning in a human coded environment . Training agents for real world scenarios is even more challenging. 🧪 #RL #AI arxiv.org/abs/1912.06680
arxiv.org
Dota 2 with Large Scale Deep Reinforcement Learning
On April 13th, 2019, OpenAI Five became the first AI system to defeat the world champions at an esports game. The game of Dota 2 presents novel challenges for AI systems such as long time horizons, im...
081
Ed Clark @clarkai.bsky.social · 21/11/2024
Make sure your DRL work is actually useful. A brilliant paper outlining all the things you should be considering when publishing DRL results arxiv.org/abs/1709.06560 #Reinforcementlearning #DRL #AI 🧪
arxiv.org
Deep Reinforcement Learning that Matters
In recent years, significant progress has been made in solving challenging problems across various domains using deep reinforcement learning (RL). Reproducing existing work and accurately judging the ...
3271
Ed Clark @clarkai.bsky.social · 21/11/2024
One of the most useful blog posts to read before messing around with PPO: iclr-blog-track.github.io/2022/03/25/p... #Reinforcementlearning #AI #RL
iclr-blog-track.github.io
The 37 Implementation Details of Proximal Policy Optimization · The ICLR Blog Track
120
Ed Clark @clarkai.bsky.social · 20/11/2024
Feel free to interact with me about: 🧪 - Reinforcement learning - Oceanography - Acoustic modelling - AI decision making Or more casually: - cycling 🚴 - sailing ⛵ - triathlon 🏊‍♂️ 🚴🏃 - climbing 🧗‍♀️
150
Ed Clark @clarkai.bsky.social · 20/11/2024
#introduction I'm a PhD researcher at the University of Bath looking into AI tactical decision making. I focus on using environmental data and digital twins in reinforcement learning to train models to make tactical decisions in challenging ocean environments. scholar.google.com/citations?us...
scholar.google.com
Edward Clark
‪PhD Student, University of Bath‬ - ‪‪Cited by 7‬‬ - ‪reinforcement learning‬ - ‪underwater acoustics‬ - ‪passive sonar‬
241