Servizi per la didattica
PORTALE DELLA DIDATTICA

Approximate computing paradigms: from component to application-level

01UJWIU

A.A. 2019/20

Course Language

English

Course degree

Doctorate Research in Ingegneria Informatica E Dei Sistemi - Torino

Course structure
Teaching Hours
Lezioni 30
Teachers
Teacher Status SSD h.Les h.Ex h.Lab h.Tut Years teaching
Savino Alessandro   Ricercatore a tempo det. L.240/10 art.24-B ING-INF/05 10 0 0 0 2
Teaching assistant
Espandi

Context
SSD CFU Activities Area context
*** N/A ***    
2019/20
PERIOD: SETTEMBRE This Course introduces basic and advanced topics on Approximate Computing. The lectures will follow a bottom-up approach: from component, up to application-level. More in detail, all the three levels of existing approximate computing techniques will be presented: hardware, data and computation. In addition, the course also provides an overview on dependability aspects related with Approximate Computing. The comprehensive overview will allow students to resort to Approximate Computing to solve computational efficiency issues in non-accuracy bounded problems.
PERIOD: SEPTEMBER This Course introduces basic and advanced topics on Approximate Computing. The lectures will follow a bottom-up approach: from component, up to application-level. More in detail, all the three levels of existing approximate computing techniques will be presented: hardware, data and computation. In addition, the course also provides an overview on dependability aspects related with Approximate Computing. The comprehensive overview will allow students to resort to Approximate Computing to solve computational efficiency issues in non-accuracy bounded problems.
• General introduction o Definition(s) o Overview and classification of techniques o Metrics • Techniques for approximate computing o Data level approximation  Data representation - arithmetic  Adaptive Precision scaling  Less data & Less up-to-date data o Hardware level approximation  Exact integer and floating-point operators  Inexact operators  Voltage over-scaling and overclocking  Approximate memories  Custom floating-point & fixed-point o Computation level approximation  Computation skip (Fine grained, coarse grained)  Computation approximation (Algorithm selection, Memoization, CNN replacement) • Methods and tools for approximate computing
• General introduction o Definition(s) o Overview and classification of techniques o Metrics • Techniques for approximate computing o Data level approximation  Data representation - arithmetic  Adaptive Precision scaling  Less data & Less up-to-date data o Hardware level approximation  Exact integer and floating-point operators  Inexact operators  Voltage over-scaling and overclocking  Approximate memories  Custom floating-point & fixed-point o Computation level approximation  Computation skip (Fine grained, coarse grained)  Computation approximation (Algorithm selection, Memoization, CNN replacement) • Methods and tools for approximate computing
inizio 8 settembre dalla 10 alle 13
inizio 8 settembre dalla 10 alle 13
Modalità di esame:
Exam:
Esporta Word


© Politecnico di Torino
Corso Duca degli Abruzzi, 24 - 10129 Torino, ITALY
Contatti