- 10
- 10 642
Kaze Wong
Приєднався 30 чер 2020
MINDS seminar @JHU 03-09-2024 Lessons learned from Machine learning in Gravitational wave
MINDS seminar @JHU 03-09-2024 Lessons learned from Machine learning in Gravitational wave
Переглядів: 48
Відео
Success and failure of ML in gravitational wave: Carl-Zeiss-Stiftung-Summer-School 2023
Переглядів 304Рік тому
This is an invited lecture I gave at the CZS summer school in Heidelberg, Germany. I talked about some of my personal opinion on how we should use machine learning techniques in science, using gravitational wave as an example.
Simulating blackholes backward: backpropagating gravitational-wave events.
Переглядів 1672 роки тому
This video explains some of the gist of my paper "Backward Population Synthesis: Mapping the Evolutionary History of Gravitational-Wave Progenitors". Note that this is not supposed to represent the exact work in the paper, but the idea behind the paper. For scientifically accurate messages, please see the original paper. Link to arXiv paper: arxiv.org/abs/2206.04062 Github: github.com/kazewong/...
Can we teach a machine what if the universe is different? The CAMELS project data release
Переглядів 8832 роки тому
How would a universe with different physics look like? The CAMELS project is a set of simulations that contains 4,233 simulated universe with different cosmological and astrophysical assumptions, that can be used as a "textbook" to train machines to learn about the universe. Find out more about the CAMELS project on www.camel-simulations.org/ A feature article about the data release www.simonsf...
Astrophysical simulations in blender tutorial - Importing data and shading
Переглядів 7803 роки тому
First in a tutorial series for visualizing astrophysical simulations in blender. This video is a sprint guide for importing volumetric data in blender and giving them basic shading. Instead of watching it and getting bored, it works best if you download the data file and follow along. The material I use are hosted on github.com/kazewong/Blender_volume_tutorial . The simulation we used is a lowe...
Visualizing cosmological simulation in blender tutorial (Final render)
Переглядів 3503 роки тому
This is how the final render should look like following the tutorial hosted github.com/kazewong/Blender_volume_tutorial. We use Blender (www.blender.org/) and openVdb (www.openvdb.org/) to visualize a snapshot from a cosmological simulation of the IllustrisTNG project ( www.tng-project.org/data/) at present day (redshift 0). Since this is associated with a introductory tutorial, we chose a low ...
Density estimation with normalizing flow in a minute
Переглядів 8 тис.4 роки тому
Normalizing flow is a generative deep neural network which can output a probability density function describing your data, by training a series of tunable transforms in the form of a neural network. This can be useful for many purposes such as simulation-based inference. For example in one of my recent papers ( arxiv.org/abs/2011.03564 ), we used a masked autoregressive flow network (arxiv.org/...
This is fantastic and thank you. Did you ever do a tutorial about how to convert the data into vdb? I'm kind of stuck there now.
@Monomorphic github.com/kazewong/Blender_tutorial/blob/main/block_hole_disk%2Fnpz_to_vdb.py Here is a script I use to process some specific data from a numpy array, other than you have to prepare the data cube yourself, the rest should be pretty general
@@physicskaze Thank you for the reply. I have your example working already without problems. I am trying to download a higher resolution .npz from IllustrisTNG. When I download the TNG300-1Dark I get 5.2 GB of .hdf5 files. DOn't really know what to do with those. Where did you get the numpy array files? Do I need to convert the .hdf5 files?
@@Monomorphic That TNG file is a specific data set we want to visualize, and there are a lot of different technique to convert the point cloud dataset in the .hdf5 to a 3D numpy array.
@@physicskaze Can you point me towards a very large, very high resolution volume file of the cosmic web that I could use? My purpose is to create in Blender a very detailed fly-though of the cosmic filament structure. Lots of intricate tendrils. No need for it to be animated, just a still volume snapshot is fine. Pretty much like the sample volume file you provided here except much larger volume and very high resolution? Please let me know. Thank you.
But how does it works for one variable? The flow must be monotonic function and thus will not give us many maxima and minima as seen here.
Podrias ayudarme con un tutorial o manual, por favor
great visualization!
so is this what's happening when generative models say they are predicting by sampling a probability distribution? does that mean the machine learning model is essentially learning the mean and standard deviation of the data?
I've tried to grasp this concept for a while now, and now I finally get it - thank you!
super cool 👍thanks for the explanation
Your first link with to your paper (2007.10350) is broken
Thanks! this is fixed now
Yes, integrate python code to produce data while live updating the visualization.
Hey I downloaded the blend file from your github page but it's not rendering anything, all I can see in it is a simple volume shader. Is this correct? If you are making the file available it would be great to check it out :)
Nice man
Fantastic.
So I can do something something with CSv data?
You should be able load the csv data and fill the 3D array with those data, then output it as vdb. As long as you can write your data into numpy array, it should be fine.
This looks amazing!
Oh Good. Im looking forward to this. A polite request. Please use a normalise filter in an audio app to get the audio volume consistent. I am wearing headphones so others are not disturbed. Its impossible to find a comfortable volume level. TIA.
Thanks for the suggestion :)
This is a really good intuitive explanation.