Geri Dön

Short term workload prediction: The reservoir computing approach

Başlık çevirisi mevcut değil.

  1. Tez No: 918800
  2. Yazar: ALPER ALİMOĞLU
  3. Danışmanlar: Belirtilmemiş.
  4. Tez Türü: Yüksek Lisans
  5. Konular: Bilgisayar Mühendisliği Bilimleri-Bilgisayar ve Kontrol, Computer Engineering and Computer Science and Control
  6. Anahtar Kelimeler: Echo State Network, Recurrent Neural Network, Machine Learning, Recursive and n-point ahead direct Time Series Prediction
  7. Yıl: 2014
  8. Dil: İngilizce
  9. Üniversite: State Unıversıty Of New York At Bınghamton
  10. Enstitü: Yurtdışı Enstitü
  11. Ana Bilim Dalı: Belirtilmemiş.
  12. Bilim Dalı: Belirtilmemiş.
  13. Sayfa Sayısı: Belirtilmemiş.

Özet

Özet yok.

Özet (Çeviri)

'What's going to happen in the future?' Time series data holds the answers, and machine learning is the leading edge method for interpreting and using the data. Machine learning is a research field that revolves around algorithms by which a computer learns from obtained data, which is from past experiences. The machine learning technique used in this work is called reservoir computing. It is a simplified structure of the human brain that uses a randomly created, recurrent network of artificial neurons called an echo state network. Reservoir computing is dependent on a nonlinear dynamic system. This system is fed with an input sequence that is mapped to a higher dimension space. During the process of training the system, only a linear readout is altered by the state of the reservoir. The input connections and recurrent connections in the system are left unchanged. This approach greatly decreases the time required to train this recurrent neural network, but does not affect the performance of the tasks. Saving energy should not be the foremost priority when IT service demand cannot be met in accordance with the predefined SLAs. The main goal is to design the DC where the IT capacities are just at the right level to meet the instantaneous demand to save energy. However, the main challenge is to predict workload usage in advance since it takes a finite amount of time to activate the servers and to adjust the cooling. The server activation takes a few minutes so it is imperative to have techniques that predict three to a few minutes in advance. Our objective is to predict the workload for a short-term (three minutes ahead) prediction interval using the reservoir computing approach. This research is concentrated on understanding the differences between direct n-point ahead and recursive predictions using reservoir systems in short term predictions.

Benzer Tezler

  1. Short term workload prediction: The reservoir computing approach

    Kısa vadeli iş yükü tahmini: Rezervuar hesaplama yaklaşımı

    ALPER ALİMOĞLU

    Yüksek Lisans

    İngilizce

    İngilizce

    2014

    Bilgisayar Mühendisliği Bilimleri-Bilgisayar ve KontrolState Unıversıty Of New York At Bınghamton

    Mühendislik Bilimleri Ana Bilim Dalı

    PROF. DR. KANAD GHOSE

  2. Kuyruk modelleri ve analizi üzerine bir uygulama

    Başlık çevirisi yok

    DİDEM ÖZPULAT

    Yüksek Lisans

    Türkçe

    Türkçe

    1985

    İstatistikEge Üniversitesi

    Bilgisayar Bilimleri ve Mühendisliği Ana Bilim Dalı

    DOÇ. DR. FİKRET İKİZ

  3. Protozoonların kültüründe soya fasulyesinden hazırlanan besiyerlerinin kullanımı

    Using media prepared from soybean for protozoa culture

    SEMRA ÖZÇELİK

    Yüksek Lisans

    Türkçe

    Türkçe

    1984

    Jeoloji MühendisliğiCumhuriyet Üniversitesi

    Mikrobiyoloji Ana Bilim Dalı

    DOÇ.DR. GÜLENDAME SAYGI

  4. Orta öğretimdeki sporcu öğrencilerin sorunları

    Başlık çevirisi yok

    FÜSUN ÖZBEKÇİ

    Yüksek Lisans

    Türkçe

    Türkçe

    1987

    SporUludağ Üniversitesi

    Spor Bilimleri Ana Bilim Dalı

    DOÇ. DR. SUAT ANAR

  5. Kobaylarda splenik ototransplantasyon ve postsplenektomi sepsisi

    Başlık çevirisi yok

    S.ÇAĞATAY ÇİFTER

    Tıpta Uzmanlık

    Türkçe

    Türkçe

    1987

    Genel CerrahiGazi Üniversitesi

    Genel Cerrahi Ana Bilim Dalı