Lorenz Richter @NeurIPS
@lorenz_richter
Followers
397
Following
391
Media
15
Statuses
35
Postdoc at @ZuseInstitute, founder of @dida_ML
Joined February 2015
Looking forward to meeting my fantastic collaborators @DenBless94, @julberner, @cdomingoenrich, @YuanqiD and many other friends.
0
0
6
Presenting our spotlight paper on trust regions for optimal control at NeurIPS, https://t.co/YS1VkDYBEB. We show that KL-equipspaced measure transport can be interpreted as geometric annealing with adaptive step sizes, leading to major performance gains on hard control problems.
5
19
120
0
0
3
Solving control problems can be hard. This is why we introduce trust region methods, approaching them iteratively in a systematic way. In fact, this can be understood as a geometric annealing from prior to target with adaptive steps. More at @NeurIPSConf, https://t.co/YS1VkDYBEB.
1
11
60
Excited for #ICML2025 in Vancouver! On Thursday morning, I'm presenting our paper ( https://t.co/8G0YZVoTz9) on a critical issue in reinforcement learning: how to correctly handle random time horizons. We've identified incorrect formulas and offer a solution. Let's chat, write me!
2
15
117
We show that existing policy gradients introduce bias in the random time horizon setting. Our corrected formulas fix this, leading to better convergence and improved performance in practice.
0
0
3
In many real-world reinforcement learning tasks - like robotics, dialogue, or trading - episode lengths vary randomly. Yet, most policy gradient methods assume a fixed and deterministic or infinite time horizon. We close this gap with a principled derivation for random times.
1
0
3
We derive policy gradients for reinforcement learning with random time horizons in https://t.co/8G0YZVoTz9. While arguably being a typical setting in applications, it has been largely overlooked in the literature. Our adjusted formulas offer significant numerical improvements.
1
6
42
I'm especially looking forward to meeting my fantastic collaborators @julberner, @DenBless94, @junhua_c and of course many other friends.
0
1
3
On my way to #ICLR2025 in Singapore. We will present • Sequential controlled Langevin diffusions, https://t.co/FSROoxOCZA, Thu, 10am, • Underdamped diffusion bridges with applications to sampling, https://t.co/WLBk0Y5KrK, Sat, 10am + oral at FPI workshop, Mon, 10am. Let's meet!
3
21
145
Our new work https://t.co/WLBk0Y5KrK extends the theory of diffusion bridges to degenerate noise settings, including underdamped Langevin dynamics (with @DenBless94, @julberner). This enables more efficient diffusion-based sampling with substantially fewer discretization steps.
1
31
141
My friend @daspantom is looking for interns working on generative modeling for proteins. Let him know in case you are interested.
We are hiring an intern on generative models for proteins for 6-12 months and typically results in publication. Find out what science is like in a bio-ml industry research lab. Bonus: float along the Rhine to work (perfectly normal commute here): https://t.co/nj57Qn5Ygy
0
0
8
Also on open/close peer review, I analyzed the AI/ML community's interest by analyzing user activity on open statistics across different venues in the past year. The user activity pattern shows (in the figure) that the fully open peer review venue, ICLR, receives significantly
To people who think "China is surpassing the US in AI" the correct thought is "Open source models are surpassing closed ones" See ⬇️⬇️⬇️
0
1
10
0
0
4
Our new work https://t.co/CtPpH5I783 studies connections between discrete and continuous time diffusion samplers - after all, most things are very much related (GFlowNets, optimal control, path measures, PDEs). This allows us to reach faster convergence by randomized time steps.
Happy to share our latest work on #diffusion models without data: building theoretical bridges between existing methods, analysing their continuous-time asymptotics, and showing some cool practical implications. https://t.co/uZtV9Hjbx8
#MachineLearning 1/9
2
19
99
Check out our new paper on diffusion-based sampling, combining diffusion models with Sequential Monte Carlo in a principled way, https://t.co/FSROoxO5a2. It improves sampling quality and leads to more robust training. Thanks @junhua_c, @julberner, @DenBless94 for the great work!
3
5
99
Our work from last year on solving high-dimensional PDEs with tensor trains via robust BSDE-based methods has just appeared in @JmlrOrg. It's a follow-up of https://t.co/lL0566Mabz and suggests a new loss that is robust and explicit, i.e. fast.
'From continuous-time formulations to discretization schemes: tensor trains and robust regression for BSDEs and parabolic PDEs', by Lorenz Richter, Leon Sallandt, Nikolas Nüsken. https://t.co/MG57YqkUXv
#iterative #tensor #numerically
1
6
75
And back to Vienna again. I will present our work on time-continuous discrete-space diffusion models at @icmlconf (with @ludiXIVwinkler, Thu, 11:30). For the first time, we can connect those models to score-based generative modeling, see https://t.co/RMz90ezHZx. Ping me!
0
9
83