• úvod
  • témata
  • události
  • tržiště
  • diskuze
  • nástěnka
  • přihlásit
    registrace
    ztracené heslo?
    RUDOLFMachine Learning | Strojové učení | In Machines We Trust | Víra v mechanickou bestii
    JINX
    JINX --- ---
    Mohl bych se prosim zeptat jaky mate nazor na neuronku ktera ma 100% uspesnost pri klasifikaci? IMHO je preucena protoze kdyz snizim trenovaci dataset z 10000 zaznamu radove mene (treba 2000) dostanu uspesnost cca 96 procent. Pouzivate nekdo nejake testovaci kriterium kdy zjistit zda neuronka je preucena a kdy ne?
    UETOYO
    UETOYO --- ---
    V tomhle článku jsou asi dobré postřehy... je i docela čerstvý, ale s GPU se to mění každý měsíc...: http://timdettmers.com/2017/04/09/which-gpu-for-deep-learning/
    UETOYO
    UETOYO --- ---
    Které karty jsou teď nejlepší v poměru cena/výkon pro domácí experimenty s ML?
    Případně má nejaký smysl kupovat např. dvě karty?
    DEEFHA
    DEEFHA --- ---
    Pokud to sem nepatří, tak mě prosím smažte: nějak jsem neměl o víkendu co dělat a vzniklo z toho webové rozhraní k výstupům open_nsfw Caffe modelu, který asi není potřeba dlouze představovat. Takže - SFWchk! :-)
    Is it clean? Or is it pr0n? Let the A.I. decides! - SFWchk
    https://sfwchk.com
    RUDOLF
    RUDOLF --- ---
    Dneska na coursera

    Applied Machine Learning in Python
    https://www.coursera.org/learn/python-machine-learning

    About this course: This course will introduce the learner to applied machine learning, focusing more on the techniques and methods than on the statistics behind these methods. The course will start with a discussion of how machine learning is different than descriptive statistics, and introduce the scikit learn toolkit. The issue of dimensionality of data will be discussed, and the task of clustering data, as well as evaluating those clusters, will be tackled. Supervised approaches for creating predictive models will be described, and learners will be able to apply the scikit learn predictive modelling methods while understanding process issues related to data generalizability (e.g. cross validation, overfitting). The course will end with a look at more advanced techniques, such as building ensembles, and practical limitations of predictive models. By the end of this course, students will be able to identify the difference between a supervised (classification) and unsupervised (clustering) technique, identify which technique they need to apply for a particular dataset and need, engineer features to meet that need, and write python code to carry out an analysis.
    SOPTIK
    SOPTIK --- ---
    JINX: Podle me do tensorboardu muzes ukladat prakticky cokoli, popr si klidne vsechny feature mapy logovat. Otazka je, zda te to zajima behem uceni nebo az pak pri predikcich, pak taky jestli chces jen vizualizovat vahy (filtry) nebo uz konkretni vysledky konvoluci se vstupnim obrazkem ... Pokud ti jde o analyzu uceni, lepsi je si v tensorboardu zobrazovat ty histogramy a statistiku vah, jak se meni v case :-)
    JINX
    JINX --- ---
    SOPTIK: jo to vypada docela pekne, popremyslim jestli dava smysl to do Kerasu prepsat. Puvodne jsem tedy myslel ze kdyz ulozim natrenovany model, tak jej pak jen prozenu nejakym nastrojem ktery mi pozadovane vystupy vyplivne.
    SOPTIK
    SOPTIK --- ---
    JINX: pokud pouzivas keras tak pekny je toto https://github.com/keplr-io/quiver
    JINX
    JINX --- ---
    Mohl bych se prosim zeptat jakym zpusobem lze vizualizovat konvolucni vrstvy, pripadne celou naucenou sit? (zkousel jsem podle zdrojaku ruznych examplu vyexportovat summary do logu a pak pouzit tensorboard - nicmene tady se vubec nic nezobrazovalo)
    RUDOLF
    RUDOLF --- ---
    Prague Artificial Intelligence & Deep Learning (Prague)| Meetup
    https://www.meetup.com/Prague-Artificial-Intelligence-Deep-Learning/
    HANT
    HANT --- ---
    Češi pomohli vytvořit umělou inteligenci s intuicí. Už není jako běžný počítač a nabízí revoluci napříč obory | Hospodářské noviny (IHNED.cz)
    http://domaci.ihned.cz/...genci-s-intuici-uz-nemysli-jako-bezny-pocitac-nabizi-revoluci-napric-obory

    DeepStack: Expert-level artificial intelligence in heads-up no-limit poker
    http://science.sciencemag.org/...urce=sciencemagazine&utm_medium=twitter&utm_campaign=moravcik-11471
    LUDWIG_
    LUDWIG_ --- ---
    Rocket AI: 2016’s Most Notorious AI Launch and the Problem with AI Hype
    https://medium.com/...et-ai-2016s-most-notorious-ai-launch-and-the-problem-with-ai-hype-d7908013f8c9
    HANT
    HANT --- ---
    Better Strategies 5: A Short-Term Machine Learning System – The Financial Hacker
    http://www.financial-hacker.com/...ld-better-strategies-part-5-developing-a-machine-learning-system/
    KALISNIK
    KALISNIK --- ---
    TRAVIX: The Room v budoucnosti!
    RUDOLF
    RUDOLF --- ---
    TRAVIX: význam výrazu automatické psaní už nikdy nebude stejný
    TRAVIX
    TRAVIX --- ---
    RUDOLF: Si to nech příště vygenerovat:
    Some Like It Bot | FiveThirtyEight
    https://fivethirtyeight.com/features/some-like-it-bot/
    RUDOLF
    RUDOLF --- ---
    Open Sourcing a Deep Learning Solution for Detecting NSFW Images

    GitHub - yahoo/open_nsfw: code for running Model and code for Not Suitable for Work (NSFW) classification using deep neural network Caffe models
    https://github.com/yahoo/open_nsfw

    Open Sourcing a Deep Learning Solution for... | Yahoo Engineering
    https://yahooeng.tumblr.com/post/151148689421/open-sourcing-a-deep-learning-solution-for
    RUDOLF
    RUDOLF --- ---
    RUDOLF: éé píšu jako prase
    RUDOLF
    RUDOLF --- ---
    nedávno přes živě.cz, postavený na účení samplů, takže při syntéze jsou slyšet mlaskání, nádechy apod.

    tady ukázky:

    WaveNet: A Generative Model for Raw Audio | DeepMind
    https://deepmind.com/blog/wavenet-generative-model-raw-audio/

    This paper introduces WaveNet, a deep neural network for generating raw audio waveforms. The model is fully probabilistic and autoregressive, with the predic- tive distribution for each audio sample conditioned on all previous ones; nonethe- less we show that it can be efficiently trained on data with tens of thousands of samples per second of audio. When applied to text-to-speech, it yields state-of- the-art performance, with human listeners rating it as significantly more natural sounding than the best parametric and concatenative systems for both English and Mandarin. A single WaveNet can capture the characteristics of many different speakers with equal fidelity, and can switch between them by conditioning on the speaker identity. When trained to model music, we find that it generates novel and often highly realistic musical fragments. We also show that it can be employed as a discriminative model, returning promising results for phoneme recognition.


    https://arxiv.org/pdf/1609.03499.pdf
    Kliknutím sem můžete změnit nastavení reklam