Machine Learning is the study of methods for programming computers to learn. Computers are applied to a wide range of tasks, and for most of these it is relatively easy for programmers to design and implement the necessary software. However, there are many tasks for which this is difficult or impossible. These can be divided into four general categories. First, there are problems for which there exist no human experts. For example, in modern automated manufacturing facilities, there is a need to predict machine failures before they occur by analyzing sensor readings. Because the machines are new, there are no human experts who can be interviewed by a programmer to provide the knowledge necessary to build a computer system. A machine learning system can study recorded data and subsequent machine failures and learn prediction rules. Second, there are problems where human experts exist, but where they are unable to explain their expertise. This is the case in many perceptual tasks, such as speech recognition, hand-writing recognition, and natural language understanding. Virtually all humans exhibit expert-level abilities on these tasks, but none of them can describe the detailed steps that they follow as they perform them. Fortunately, humans can provide machines with examples of the inputs and correct outputs for these tasks, so machine learning algorithms can learn to map the inputs to the outputs. Third, there are problems where phenomena are changing rapidly. In finance, for example, people would like to predict the future behavior of the stock market, of consumer purchases, or of exchange rates. These behaviors change frequently, so that even if a programmer could construct a good predictive computer program, it would need to be rewritten frequently. A learning program can relieve the programmer of this burden by constantly modifying and tuning a set of learned prediction rules. Fourth, there are applications that need to be customized for each computer user separately. Consider, for example, a program to filter unwanted electronic mail messages. Different users will need different filters. It is unreasonable to expect each user to program his or her own rules, and it is infeasible to provide every user with a software engineer to keep the rules up-to-date. A machine learning system can learn which mail messages the user rejects and maintain the filtering rules automatically. Machine learning addresses many of the same research questions as the fields of statistics, data mining, and psychology, but with differences of emphasis. Statistics focuses on understanding the phenomena that have generated the data, often with the goal of testing different hypotheses about those phenomena. Data mining seeks to find patterns in the data that are understandable by people. Psychological studies of human learning aspire to understand the mechanisms underlying the various learning behaviors exhibited by people (concept learning, skill acquisition, strategy change, etc.).

Machine learning

International Technology Roadmap for Semiconductors 2003の要求清浄度について － シリコンウエハ表面と雰囲気環境に要求される清浄度, 分析方法の現状について －

A geometric program (GP) is a type of mathematical optimization problem characterized by objective and constraint functions that have a special form. Recently developed solution methods can solve even large-scale GPs extremely efficiently and reliably; at the same time a number of practical problems, particularly in circuit design, have been found to be equivalent to (or well approximated by) GPs. Putting these two together, we get effective solutions for the practical problems. The basic approach in GP modeling is to attempt to express a practical problem, such as an engineering analysis or design problem, in GP format. In the best case, this formulation is exact; when this is not possible, we settle for an approximate formulation. This tutorial paper collects together in one place the basic background material needed to do GP modeling. We start with the basic definitions and facts, and some methods used to transform problems into GP format. We show how to recognize functions and problems compatible with GP, and how to approximate functions or data in a form compatible with GP (when this is possible). We give some simple and representative examples, and also describe some common extensions of GP, along with methods for solving (or approximately solving) them.

/pdf/a-tutorial-on-geometric-programming-47sdznbzxw.pdf

A tutorial on geometric programming

This article provides a practical introduction to the design trade-offs of the currently available 3D IC technology options. It begins with an overview of techniques, such as wire bonding, microbumps, through vias, and contactless interconnection, comparing them in terms of vertical density and practical limits to their use. We then present a high-level discussion of the pros and cons of 3D technologies, with an analysis relating the number of transistors on a chip to the vertical interconnect density using estimates based on Rent's rule. Next, we provide a more detailed design example of inductively coupled interconnects, with measured results of a system fabricated in a 0.35-/spl mu/m technology and an analysis of misalignment and crosstalk tolerances. Lastly, we present a case study of a fast Fourier transform (FFT) placed and routed in a 0.18-/spl mu/m through-via silicon-on-insulator (SOI) technology, comparing the 3D design to a traditional 2D approach in terms of wire length and critical-path delay.

Demystifying 3D ICs: the pros and cons of going vertical

IEEE transactions on computer-aided design of integrated circuits and systems : a publication of the IEEE Circuits and Systems Society

http://users.ece.gatech.edu/limsk/www/pdfs/integration97.pdf

Performance optimization of VLSI interconnect layout

In this paper, we study the interconnect layout optimization problem under a higher-order RLC model to optimize not just delay, but also waveform for RLC circuits with non-monotone signal response. We propose a unified approach that considers topology optimization, wiresizing optimization, and waveform optimization simultaneously. Our algorithm considers a large class of routing topologies, ranging from shortest-path Steiner trees to bounded-radius Steiner trees and Steiner routings. We construct a set of required-arrival-time Steiner trees or RATS-trees, providing a smooth trade-off among signal delay, waveform, and routing area. Using a new incremental moment computation algorithm, we interleave topology construction with moment computation to facilitate accurate delay calculation and evaluation of waveform quality. Experimental results show that our algorithm is able to construct a set of topologies providing a smooth trade-off among signal delay, signal settling time, voltage overshoot, and routing cost.

/pdf/interconnect-design-for-deep-submicron-ics-1wpzzxlew8.pdf

Interconnect design for deep submicron ICs

We investigate the problem of decoupling capacitance (decap) allocation for power supply noise suppression at floorplan level. First, we assume that a floorplan is given and consider the decap placement as a postfloorplan step. Second, we consider the decap placement as an integral part of a floorplanning methodology (noise-aware floorplanning). In both cases, the objective is to minimize the floorplan area while suppressing the power supply noise below the specified limit. Experimental results on MCNC benchmark circuits show that, for postfloorplan decap placement, the white space allocated for decap is about 6%-9% of the chip area for the 0.25-/spl mu/m technology. The power-supply noise is kept below the specified limit. Compared to postfloorplan approach, the peak power-supply noise can be reduced by as much as 40% and the decap budget can be reduced by as much as 21% by using noise-aware floorplanning methodology. The total area is also reduced due to the reduced total decap budget gained from reduced power supply noise.

/pdf/decoupling-capacitance-allocation-and-its-application-to-4diqm5otf0.pdf

Decoupling capacitance allocation and its application to power-supply noise-aware floorplanning

Circuit interconnect has become a substantial obstacle in the design of high performance systems. In this paper we explore a new routing paradigm that strikes at the root of the interconnect problem by reducing wire lengths directly. We present a non-Manhattan Steiner tree heuristic, obtaining wire length reductions of much as 17% on average, when compared to rectilinear topologies. Moreover, we present a graph-based interconnect optimization algorithm, called the GRATS-tree algorithm, which allows performance optimization beyond what can be obtained through wire length reduction alone. The two tree construction algorithms are integrated into a new global router that allows large scale non-Manhattan design. Although we consider circuit placements performed under rectilinear objectives, our global router can reduce maximum congestion levels by as much as 20%. In general we find that the non-Manhattan approach requires additional Steiner points and bends; realization of non-Manhattan routing structures requires additional vias. We observe that the increase in via cost is much less dramatic than might be expected; the benefits of wire length reduction may outweigh the additional via cost.

https://docs.lib.purdue.edu/cgi/viewcontent.cgi?article=1039&context=ecetr

Manhattan or non-Manhattan?: a study of alternative VLSI routing architectures

We study the minimum-cost bounded-skew routing tree problem under the pathlength (linear) and Elmore delay models. This problem captures several engineering tradeoffs in the design of routing topologies with controlled skew. Our bounded-skew routing algorithm, called the BST/DME algorithm, extends the DME algorithm for exact zero-skew trees via the concept of a merging region. For a prescribed topology, BST/DME constructs a bounded-skew tree (BST) in two phases: (i) a bottom-up phase to construct a binary tree of merging regions which represent the loci of possible embedding points of the internal nodes, and (ii) a top-down phase to determine the exact locations of the internal nodes. We present two approaches to construct the merging regions: (i) the Boundary Merging and Embedding (BME) method which utilizes merging points that are restricted to the boundaries of merging regions, and (ii) the Interior Merging and Embedding (IME) algorithm which employs a sampling strategy and a dynamic programming-based selection technique to consider merging points that are interior to, as well as on the boundary of, the merging regions. When the topology is not prescribed, we propose a new Greedy-BST/DME algorithm which combines the merging region computation with topology generation. The Greedy-BST/DME algorithm very closely matches the best known heuristics for the zero-skew case and for the unbounded-skew case (i.e., the Steiner minimal tree problem). Experimental results show that our BST algorithms can produce a set of routing solutions with smooth skew and wire length tradeoffs.

/pdf/bounded-skew-clock-and-steiner-routing-4e9xw2e3tg.pdf

Cheng-Kok Koh

Papers

Performance optimization of VLSI interconnect layout

Interconnect design for deep submicron ICs

Decoupling capacitance allocation and its application to power-supply noise-aware floorplanning

Manhattan or non-Manhattan?: a study of alternative VLSI routing architectures

Bounded-skew clock and Steiner routing