She is by far the best-looking authority on black holes. This was fun to listen to
@wolpumba40992 ай бұрын
*Summary* This presentation introduces Estimagic, a Python package for nonlinear optimization, highlighting its usefulness for complex estimation problems and statistical inference. *Key takeaways:* * *Problem (**0:40**):* Many professionals (e.g., epidemiologists, statisticians, engineers) face nonlinear optimization challenges but lack the expertise and interest to delve into optimization algorithms. * *Existing Solutions' Shortcomings (**1:25**):* Traditional optimization libraries like SciPy can be challenging for non-experts due to cryptic errors, boilerplate code, and difficulties switching between libraries with different interfaces. * *Estimagic's Solution (**3:51**):* * *Unified Interface:* Wraps optimizers from various libraries (SciPy, NLopt, etc.) under a common, user-friendly interface inspired by SciPy.minimize. * *Flexible Parameter Handling:* Accepts various parameter data types (NumPy arrays, Pandas objects, lists, dictionaries, etc.) and allows for nested structures, eliminating the need to manually unpack parameters. * *Diagnostic Tools (**7:46**):* Provides visualization tools (e.g., Criterion plot) to compare the performance of different optimization algorithms easily. * *Global Optimization (**9:32**):* Includes algorithms for global optimization (genetic algorithms, multi-start framework) to handle objective functions with multiple local optima. * *Other Features (**11:09**):* * Numerical derivative calculation * Maximization capabilities * Support for bounds and constraints * Jax compatibility for accelerated computations and automatic differentiation * Enhanced error handling for numerical stability * Algorithms for nonlinear least squares problems * *Team (**12:45**):* The speaker acknowledges the contributions of various individuals and institutions in developing and supporting Estimagic. * *Roadmap (**13:29**):* Future development plans include simplifying the addition of optimizers, creating algorithms for noisy optimization, and improving the real-time dashboard. * *Call to Action (**14:07**):* Users are encouraged to contribute, report issues, and request features on the Estimagic GitHub page. i used gemini 1.5 pro to summarize the transcript
@Molox154 ай бұрын
The API is very similar to lpyspark. In fact I don't think it would be a hassle to convert existing pipelines to polars.
@oflasch6 ай бұрын
Excellent work! Solara is a joy to work with. Many thanks for creating it!
@iutubtivi10 ай бұрын
I have no knowledge or time to do benchmarking but, I was using pandas' "append" to combine about 8000 CSV files (about 10 GB in total) and it was taking almost an hour and a half, i decided to try polars, according to stack overflow i could use, concat, vstack, or extend, i randomly chose "vstack", and it did the same workload in less than 1 minute, same computer, same python version, same everything, all i had to do was modify the script a little bit, for example remove "index = False" when exporting the resulting (huge) dataframe to CSV.
@SuilujChannel10 ай бұрын
impressive!
@adityaaware984410 ай бұрын
Am I being super Geeky cz im getting super excited by this😅😅
@tilkesh Жыл бұрын
Thanks
@dafmdev Жыл бұрын
At the beginning of the presentation he showed the performance of using only Pandas. How is the performance using Rust?
@mindreendtre Жыл бұрын
1 order of magnitude was suggested, but it would be interesting to see a comparison.
@Jononor Жыл бұрын
Very good progress. Looking forward to the day where predict_proba() is the default people use - not predict(). The automatic 0.5 decision threshold is basically never the right one!
@nilsflaschel49 Жыл бұрын
Impressive, how scikit-learn is progressing, especially in the imbalanced learning area!
@krumarnaudov Жыл бұрын
Love the move to industry use-cases!
@sergeybrutspark Жыл бұрын
Be the first to like ))) nice topic
@DiegoAndresAlvarezMarin2 жыл бұрын
It is a shame that we missed this material. An advanced talk with a horrible sound and image :( Can you please post the link to the slides?
@Sigma000002 жыл бұрын
Any updates on this work?
@juozasdautartas85662 жыл бұрын
Very useful library!
@pau19763 жыл бұрын
Thanks, very useful.
@simonjemand31493 жыл бұрын
Thank you for your nice talk! Roughly, how much longer do c-functions take to be implemented compared to their Numba counterpart?
@nix99problems3 жыл бұрын
Badly captured video. Can't see the projected screen
@DrIlyas-sq7pz3 жыл бұрын
Thank you. I am using QuTip package in which i am making some logic gates. I follow the instructions In the user guide it is said that i need to install imagemagick but by conda or pip it does not install image magic but i installed wand.I already installed imagmagick on my pc manually. No after all the imports etc and after writing commands for gate making, i need to write gate.png to display the gate but it keep giving error below. Plz let me know what is this error and how can i display some result as image file in python. I will be grateful. I am using jupyter.
@DrIlyas-sq7pz3 жыл бұрын
--------------------------------------------------------------------------- CalledProcessError Traceback (most recent call last) <ipython-input-9-a760fe8d2360> in <module> 1 q = QubitCircuit(2, reverse_states=False) 2 q.add_gate("CSIGN", controls=[0], targets=[1]) ----> 3 q.png ~\anaconda3\lib\site-packages\qutip-4.6.0.dev0-py3.8-win-amd64.egg\qutip\qip\circuit.py in png(self) 1759 @property 1760 def png(self): -> 1761 return DisplayImage(self._raw_png(), embed=True) 1762 1763 def _raw_svg(self): ~\anaconda3\lib\site-packages\qutip-4.6.0.dev0-py3.8-win-amd64.egg\qutip\qip\circuit.py in _raw_png(self) 1752 1753 def _raw_png(self): -> 1754 return _latex.image_from_latex(self.latex_code(), "png") 1755 1756 if 'png' in _latex.CONVERTERS: ~\anaconda3\lib\site-packages\qutip-4.6.0.dev0-py3.8-win-amd64.egg\qutip\qip\circuit_latex.py in image_from_latex(code, file_type) 225 _run_command((_pdflatex, '-interaction', 'batchmode', 226 filename)) --> 227 _crop_pdf(filename + ".pdf") 228 if file_type in _MISSING_CONVERTERS: 229 dependency = _MISSING_CONVERTERS[file_type] ~\anaconda3\lib\site-packages\qutip-4.6.0.dev0-py3.8-win-amd64.egg\qutip\qip\circuit_latex.py in _crop_pdf(filename) 113 """Crop the pdf file `filename` in place.""" 114 temporary = ".tmp." + filename --> 115 _run_command((_pdfcrop, filename, temporary)) 116 # Windows does not allow renaming to an existing file (but unix does). 117 _force_remove(filename) ~\anaconda3\lib\subprocess.py in run(input, capture_output, timeout, check, *popenargs, **kwargs) 510 retcode = process.poll() 511 if check and retcode: --> 512 raise CalledProcessError(retcode, process.args, 513 output=stdout, stderr=stderr) 514 return CompletedProcess(process.args, retcode, stdout, stderr) CalledProcessError: Command '('pdfcrop', 'qcirc.pdf', '.tmp.qcirc.pdf')' returned non-zero exit status 1.
@martindominikus88253 жыл бұрын
this is amazing, thank you!
@KarlVonBismark4 жыл бұрын
Excellent
@gakhov4 жыл бұрын
Here you can find the slides from the talk: www.slideshare.net/gakhov/exceeding-classical-probabilistic-data-structures-in-data-intensive-applications
@drew51345 жыл бұрын
Python's arithmetic operators, should be b = exp(-((x-mu)**2/(2*sigma**2))) b thanks for the great video.
@pawhidden50545 жыл бұрын
Is there any code online available in this lecture?
@alexwickstrom3034 жыл бұрын
consulting.behnel.de/training/espy2017/
@anujmishra44122 жыл бұрын
@@alexwickstrom303 Thank you so much!
@pintukr235 жыл бұрын
I am using dask on kubernetes cluster on AKS (azure). I am getting some fixed number of workers . It is not zero. Where and what should I change?
@kulkarniajinkya5 жыл бұрын
Hey, very informative! Can you share the notebook?
@MsHappyburger5 жыл бұрын
@@bornabesic8315 thanks
@prativadas47945 жыл бұрын
can we have the ppt ?
@StefanoBorini5 жыл бұрын
hey, that's me! :)
@mundaka876 жыл бұрын
You can find more information on "Scientific computing for quantum technology" and the other talks of EuroScipy 2018 at gist.github.com/carmelom/e4c898f2305f726aad744d0dd9bdf1a0.
@GeospatialTechHeaven6 жыл бұрын
kzbin.info/www/bejne/anLaiJywj5plm6s
@MarioHari6 жыл бұрын
Watching at 1.5x speed seems perfectly normal. Else it's super slow
Tutorial notebooks can be found here github.com/leosartaj/euroscipy-sympy-tutorial
@whereistejas6 жыл бұрын
bad audio!
@mariazamyatina22786 жыл бұрын
Is it possible to download the Jupyter Notebook he is using somewhere?
@yazu006 жыл бұрын
I think this link works: github.com/MridulS/euroscipy-2017-networkx
@luiscerna70716 жыл бұрын
Just install the tool for your operating system from: jupyter.org/install then from your command prompt you need to hit the command : "jupyter notebook" I think it is that same command for any operating system. That will open the Jupyter Notebook environment on your browser and that will link to your prompt. In the environment opened in you browser you will see your home directory and from there you can access your files or create Jupyter Notebook projects, like the one from this video. Jupyter Notebook projects posses .ipynb extension. You will need to install Networkx and Matplotlib for this examples.
@TheXplorian6 жыл бұрын
@20:10 why had he made a sigmoid layer before the softmax layer, what i understand about logistic regression, is that it doesn't have a sigmoid activation
@theultimatereductionist75926 жыл бұрын
Fantastic talk!
@ferniquepierre49646 жыл бұрын
See the summary at www.euroscipy.org/2017/descriptions/19828.html