Mateusz Buda (@budamat) header
Mateusz Buda (@budamat) profile picture

Mateusz Buda

@budamat

Machine learning research engineer

 Warsaw, Poland
mateuszbuda.github.io
Joined April 2015
  • Tweets320
  • Following66
  • Followers104
2Media
Photo by Mateusz Buda (@budamat)Photo by Mateusz Buda (@budamat)
Mateusz Buda retweeted
andrew gao (@itsandrewgao) profile picture
andrew gao
@itsandrewgao10 Aug 2023
Go to Google Scholar and look up ‘As an AI language model” -“ChatGPT”’
Image from andrew gao (@itsandrewgao)
476
7,449
2,378
66,785
9,738,767
Mateusz Buda retweeted
Krzysztof Geras (@kjgeras) profile picture
Krzysztof Geras
@kjgeras25 Jul 2023
Moje trzy grosze na temat tego, co obecnie dzieje się w AI...
Maciej Kawecki - This Is IT (@kawecki_maciej) profile pictureMaciej Kawecki - This Is IT
@kawecki_maciej25 Jul 2023
Wywiad, który mówi "nie" spiskowym teoriom o AI. Jeden z najwybitniejszych polskich naukowców, tworzący od lat sztuczne sieci neuronowe ratujące ludzkie życia. Człowiek, którego algorytm tworzony latami na New York University Langone Health dziennie ratuje dziś setki kobiet. Setki! Profesor Krzysztof Geras to profesor Uniwersytetu Nowojorskiego, gdzie opracował algorytm najskuteczniej na świecie diagnozujący raka piersi. Przeprowadzając badania, skuteczność jego algorytmu zapewniły mu między innymi ogromne zasoby danych polskich pacjentów. Efekt jego prac dotrze też do polski. Robił doktorat na University of Edinburgh. Jest autorem ponad 50 recenzowanych artykułów naukowych dotyczących obszaru sztucznej inteligencji, cytowanych na całym świecie ponad 3 500 razy. Krzysztof stworzył i opublikował w IEEE Explore pionierską pracę wskazującą na moc danych obrazowych w diagnostyce nowotworowej. Otrzymał grand badawczy od Gordon and Betty Moore Foundation założonej przez twórcę Intela, dzięki któremu to stypendium skuteczność jego algorytmu będzie mogła ratować chorych na inne nowotwory. W 2022 roku został laureatem nagrody im. S. Lema za całokształt badań naukowych. Pan profesor poświęcił mi w Nowym Jorku czas, by odpowiedzieć w wywiadzie na najbardziej nurtujące mnie pytania związane z rozwojem AI. Pierwszy człowiek z którym rozmawiałem, tak racjonalnie mówiący, że jesteśmy daleko od stworzenia AI zdolnej do zagrożenia ludzkości. Wskazuje, że cały czas bilans korzyści z rozwoju tej technologii jest wyższy, niż bilans wad. Tylko o wadach mówi się głośniej. Mówi, że AI przypomina broń jądrową ale nie w perspektywie zagrożeń, ale sposobu dostępu do niej na świecie. Krzysztof, dziękuję za poświęcony mi czas. To był dla mnie zaszczyt. Polecam rozmowę, a link zamieszczam poniżej. Inny głos w dyskusji. Krzysztof pokazuje, jak technologia potrafi ratować i dawać szanse. Wywiad: youtu.be/grm8dTpa5yY
Image from Maciej Kawecki - This Is IT (@kawecki_maciej)
3
1
2
10
2,954
ml-stat-util: Statistical functions based on bootstrapping for computing confidence intervals and p-values comparing machine learning models and human readers Lang: Jupyter Notebook ⭐️ 42 Author: @budamat #MachineLearning github.com/mateuszbuda/ml-st…
GitHub - mateuszbuda/ml-stat-util: Statistical functions based on bootstrapping for computing...

GitHub - mateuszbuda/ml-stat-util: Statistical functions based on bootstrapping for computing...

Statistical functions based on bootstrapping for computing confidence intervals and p-values comparing machine learning models and human readers - mateuszbuda/ml-stat-util

github.com
2
3
298
new blog post: all about transformers! covering the intuition behind an "attention only" network, building up an understanding of scaled dot product attention, and detailing the original transform architecture. let me know what you think! 🙏 jeremyjordan.me/transformer-…
Understanding the Transformer architecture for neural networks

Understanding the Transformer architecture for neural networks

The attention mechanism allows us to merge a variable-length sequence of vectors into a fixed-size context vector. What if we could use this mechanism to entirely replace recurrence for sequential...

jeremyjordan.me
4
99
1
376
41,538
Mateusz Buda retweeted
Nick Konz (@nick_konz) profile picture
Nick Konz@nick_konz23 Feb 2023
I'm excited to share that the results of the Digital Breast Tomosynthesis Tumor Detection Challenge (DBTex) have officially been published in our @JAMANetworkOpen paper, with our public benchmark, dataset, and detection algorithms' code (links 👇)! jamanetwork.com/journals/jam… (1/n)
Artificial Intelligence to Detect Lesions in Digital Breast Tomosynthesis

Artificial Intelligence to Detect Lesions in Digital Breast Tomosynthesis

This diagnostic study describes a grand challenge for using artificial intelligence (AI) to detect cancer in digital breast tomosynthesis images.

jamanetwork.com
2
11
3
42
39,077
8 teams in the international DBTex competition produced #AI algorithms with high sensitivity to detect lesions on digital breast tomosynthesis (#DBT) images, including model code, datasets, and benchmarks for future researchers. ja.ma/3EvSG9i
5
5
1,103
ml-stat-util: Statistical functions based on bootstrapping for computing confidence intervals and p-values comparing machine learning models and human readers Lang: Jupyter Notebook ⭐️ 39 Author: @budamat #MachineLearning github.com/mateuszbuda/ml-st…
GitHub - mateuszbuda/ml-stat-util: Statistical functions based on bootstrapping for computing...

GitHub - mateuszbuda/ml-stat-util: Statistical functions based on bootstrapping for computing...

Statistical functions based on bootstrapping for computing confidence intervals and p-values comparing machine learning models and human readers - mateuszbuda/ml-stat-util

github.com
1
1
196
Mateusz Buda retweeted
Scraping Fish (@ScrapingF) profile picture
Scraping Fish@ScrapingF5 May 2022
Checkout a new post on our blog: Scraping Airbnb - scrapingfish.com/blog/scrapi… #Airbnb #webscraping
3
3
Class imbalance, or having significantly higher number of examples in a training set than other classes, can have adverse effects on training deep learning models in achieving optimal classification performance. There are several methods to manage class imbalance.. #AI
1
2
1
3
Class imbalance, or having significantly higher number of examples in a training set than other classes, can have adverse effects on training deep learning models in achieving optimal classification performance. There are several methods to manage class imbalance.. #AI
1
4
4
Development and evaluation of deep learning algorithms for breast cancer screening has been limited by a lack of well-annotated, large-scale publicly available data sets. This paper by #DAIR’s Dr. Maciej Mazurowski @Maciej and colleagues @budamat @nianyili @JosephLo16 provide...
1
4
10
We’re in the midst of a broad movement toward data sharing and advanced data analytics in medical research. @writemed reports on a great example in breast cancer screening. #DataSharing @budamat @MazurowskiPhD wb.md/3sHQq7O
1
4
1
9
Mateusz Buda retweeted
Medscape (@Medscape) profile picture
Medscape
@Medscape19 Aug 2021
A curated dataset of over 22K digital breast tomosynthesis images should help advance the AI algorithms used for breast cancer imaging. ms.spr.ly/6011n0FMJ
Data Sharing to Improve AI Used in Breast-Imaging Research

Data Sharing to Improve AI Used in Breast-Imaging Research

The large-scale sharing of data is a key step toward transparency in science, say researchers who have made a large dataset of DBT images available to aid development of machine-learning algorithms.

medscape.com
7
2
9
In this study, the publicly available data set, alongside the deep learning model, could significantly advance the research on machine learning tools in breast cancer screening and medical imaging in general. @budamat @MazurowskiPhD @DukeRadiology ja.ma/2VZ5th4
4
2
10
NEW PUBLIC DATASET ALERT! We made a new dataset of medical images available for machine learning including 5,610 digital breast tomosynthesis studies for 5,060 patients. This translates into 22,032 reconstructed volumes. Annotations are included. sites.duke.edu/mazurowski/re… 1/8
4
40
7
122
🚨New Paper Alert🚨 The top 3 teams of the first DBTex challenge (@NYUImaging, @IBMResearch, @ViCOROB) share the lessons we learned on a short paper in @NatMachIntell! We discuss how we trained AI models on 3D mammograms and room for improvement. rdcu.be/cpbRj 1/n
Image from Jungkyu Park (@jpatrickpark)
1
7
1
20
Short paper in Nature Machine Intelligence @NatMachIntell on our participation in the 1st DBT lesion detection challenge, joint collaboration with the winners of the challenge @nyuimaging and @IBMResearch, @kjgeras @ViCOROB @UdGRecerca rdcu.be/cpbRj
3
10
NEW MACHINE LEARNING CHALLENGE! We just launched DBTex2 challenge to develop machine learning algorithms that detect abnormalities in digital breast tomosynthesis images. The challenge comes with a publicly available dataset for more than 5000 patients. spie-aapm-nci-dair.westus2.c…
4
35
2
118
Mateusz Buda retweeted
Krzysztof Geras (@kjgeras) profile picture
Krzysztof Geras
@kjgeras7 May 2021
I am looking for a postdoc to work on deep learning methods for computer vision! Particularly: - reinforcement learning - vision transformers - Bayesian neural networks - causal inference - transfer learning - self-supervised learning Full info here: cs.nyu.edu/~kgeras/job_2021.…
3
24
2
93
Mateusz Buda retweeted
Mike Schroepfer (@schrep) profile picture
Mike Schroepfer
@schrep3 May 2021
.@AndrewYNg great at distilling down where AI is. Paper sayin AI trained at Stanford can read X-Ray's better than Humans @ Stanford - but same system doesn't work at hospital down the street. Generalization unsolved spectrum.ieee.org/view-from-…
Andrew Ng X-Rays the AI Hype

Andrew Ng X-Rays the AI Hype

AI pioneer says machine learning may work on test sets, but that’s a long way from real world use

spectrum.ieee.org
5
31
6
146
Load more