text
stringlengths
16
3.88k
source
stringlengths
60
201
C� | (A� · nˆ), where ˆn = A, � C). The latter identity can also be checked � · (B� × C� ) = det( � B, � C . So volume = A C/ B� × � |B� × � | directly using components.
https://ocw.mit.edu/courses/18-02-multivariable-calculus-fall-2007/649253ba60d11b0598cc58e9dcf58142_lec_week1.pdf
INTRODUCTION Very roughly speaking, representation theory studies symmetry in linear spaces. It is a beautiful mathematical subject which has many applications, ranging from number theory and combinatorics to geometry, probability theory, quantum mechanics and quantum field theory. Representation theory was born in ...
https://ocw.mit.edu/courses/18-712-introduction-to-representation-theory-fall-2010/64ac0a7f209d31929d393a80e3ab85bb_MIT18_712F10_intro.pdf
mostly follow [FH], with the exception of the sections discussing quivers, which follow [BGP]. We also recommend the comprehensive textbook [CR]. The notes should be accessible to students with a strong background in linear algebra and a basic knowledge of abstract algebra. Acknowledgements. The authors are grateful...
https://ocw.mit.edu/courses/18-712-introduction-to-representation-theory-fall-2010/64ac0a7f209d31929d393a80e3ab85bb_MIT18_712F10_intro.pdf
MASSACHUSETTS INSTITUTE OF TECHNOLOGY Physics Department Physics 8.07: Electromagnetism II Prof. Alan Guth September 5, 2012 LECTURE NOTES 1 VECTOR ANALYSIS DEFINITION: A vector is a quantity that has magnitude and direction. Examples: displacement, velocity, acceleration, force, momentum, electric and magnet...
https://ocw.mit.edu/courses/8-07-electromagnetism-ii-fall-2012/64baa9cb5023c49bc9147bee5416b6ec_MIT8_07F12_ln1.pdf
In fact, one can show that if AA · BA obeys rotational invariance and the distributive law, then A AA · B = const|AA||BA | cos θ . We’ll come back to this later. 8.07 LECTURE NOTES 1, FALL 2012 p. 3 Cross product of two vectors: AA × BA ≡ |AA||BA | sin θ ˆn , where ˆn is a unit vector perpendicular to AA a...
https://ocw.mit.edu/courses/8-07-electromagnetism-ii-fall-2012/64baa9cb5023c49bc9147bee5416b6ec_MIT8_07F12_ln1.pdf
Ax + Bx)ˆex + (Ay + By )ˆey + (Az + Bz)ˆez . 8.07 LECTURE NOTES 1, FALL 2012 p. 4 Vector dot product: AA · BA = (Axeˆx + Ay eˆy + Az eˆz ) · (Bxeˆx + By eˆy + Bzeˆz ) . eˆx · eˆx = ˆey · eˆy = ˆez · eˆz = 1 , and ˆex · eˆy = 0, as does the dot product of any two basis vectors. . A BA = AxBx + AyBy + AzBz .. A...
https://ocw.mit.edu/courses/8-07-electromagnetism-ii-fall-2012/64baa9cb5023c49bc9147bee5416b6ec_MIT8_07F12_ln1.pdf
product of any two distinct basis vectors must vanish, so AA · BA = const (AxBx + Ay By + Az Bz) . This had better be equivalent to const|AA||BA | cos θ, but we can see it explicitly by using the rotational invariance to orient AA along the pos­ itive x-axis, so Ax = |AA|, and Ay = Az = 0. Then the above formula g...
https://ocw.mit.edu/courses/8-07-electromagnetism-ii-fall-2012/64baa9cb5023c49bc9147bee5416b6ec_MIT8_07F12_ln1.pdf
://ocw.mit.edu 8.07 Electromagnetism II (cid:41)(cid:68)(cid:79)(cid:79)(cid:3)(cid:21)(cid:19)(cid:20)(cid:21) For information about citing these materials or our Terms of Use, visit: http://ocw.mit.edu/terms.
https://ocw.mit.edu/courses/8-07-electromagnetism-ii-fall-2012/64baa9cb5023c49bc9147bee5416b6ec_MIT8_07F12_ln1.pdf
Type Classes and Subtyping Armando Solar-Lezama Computer Science and Artificial Intelligence Laboratory MIT October 5, 2015 With content by Arvind and Adam Chlipala. Used with permission. October 5, 2015 L08-1 Hindley-Milner gives us generic functions • Can generalize a type if the function makes no assump...
https://ocw.mit.edu/courses/6-820-fundamentals-of-program-analysis-fall-2015/64d576d7d140cb9fbbd23364c2c2b453_MIT6_820F15_L08.pdf
) :: a -> a -> a fromInteger :: Integer -> a } October 5, 2015 L08-5 Generalized Functions w/ “class” types matrixMul :: Num a -> Mat a -> Mat a -> Mat a dft :: Num a -> Vec a -> Vec a -> Vec a • All of the numeric aspects of the type has been isolated to the Num type – For each type, we built a num...
https://ocw.mit.edu/courses/6-820-fundamentals-of-program-analysis-fall-2015/64d576d7d140cb9fbbd23364c2c2b453_MIT6_820F15_L08.pdf
October 5, 2015 L08-8 Laws for a type class • A type class often has laws associated with it – E.g., + in Num should be associate and commutative • These laws are not checked or ensured by the compiler; the programmer has to ensure that the implementation of each instance correctly follows the law more o...
https://ocw.mit.edu/courses/6-820-fundamentals-of-program-analysis-fall-2015/64d576d7d140cb9fbbd23364c2c2b453_MIT6_820F15_L08.pdf
super.append(ls); } int howManyAppends() { return numAppends; } } L08-15 Subtyping as Graph Search float List Arrow from A to B to indicate that B is a “direct” subtype of A int Nil Cons LoggingCons Q: How do we decide if A is a subtype of B? A: Graph reachability! (Ea...
https://ocw.mit.edu/courses/6-820-fundamentals-of-program-analysis-fall-2015/64d576d7d140cb9fbbd23364c2c2b453_MIT6_820F15_L08.pdf
�� L08-20 Sanity-Checking the Principle Primitive rule: int ≤ float ✔Any integer N can be treated as N.0, with no loss of meaning. Primitive rule: float ≤ int ✗E.g., “%” operator defined for int but not float. Primitive rule: int ≤ int → int ✗Can't call an int! L08-21 From Nominal to Structural A str...
https://ocw.mit.edu/courses/6-820-fundamentals-of-program-analysis-fall-2015/64d576d7d140cb9fbbd23364c2c2b453_MIT6_820F15_L08.pdf
A : int, B : float} ≤ {A : float, B : float} {A : float, B : float} ≤ {A : int, B : float} ? {A : int, B : float} ≤ {A : float} ? Depth: ∀i. 𝜏i ≤ 𝜏'i {ai : 𝜏i} ≤ {ai : 𝜏'i} Width: ∀j. ∃i. ai = a'j ∧ 𝜏i = 𝜏'j {ai : 𝜏i} ≤ {a'j : 𝜏'j} Yes! No! Yes! L08-25 Function Types Consider types 𝜏1 → 𝜏2. ...
https://ocw.mit.edu/courses/6-820-fundamentals-of-program-analysis-fall-2015/64d576d7d140cb9fbbd23364c2c2b453_MIT6_820F15_L08.pdf
Read a 𝜏 from some index. 2. Write a 𝜏 to some index. 𝜏1 ≤ 𝜏2 𝜏1[] ≤ 𝜏2[] Covariant rule: Counterexample: int[] x = new int[1]; float[] y = x; // Use subtyping here. y[0] = 1.23; int z = x[0]; // Not an int! L08-28 Arrays Consider types 𝜏[]. What operations must we support? 1. Read a 𝜏 f...
https://ocw.mit.edu/courses/6-820-fundamentals-of-program-analysis-fall-2015/64d576d7d140cb9fbbd23364c2c2b453_MIT6_820F15_L08.pdf
information about citing these materials or our Terms of Use, visit: http://ocw.mit.edu/terms.
https://ocw.mit.edu/courses/6-820-fundamentals-of-program-analysis-fall-2015/64d576d7d140cb9fbbd23364c2c2b453_MIT6_820F15_L08.pdf
6.826—Principles of Computer Systems 2002 6.826—Principles of Computer Systems 2002 19. Sequential Transactions with Caching There are many situations in which we want to make a ‘big’ action atomic, either with respect to concurrent execution of other actions (everyone else sees that the big action has either not...
https://ocw.mit.edu/courses/6-826-principles-of-computer-systems-spring-2002/64d732d98d264f9a61e381bf1358c907_19.pdf
more efficient code that allows for caching data; this is usually necessary for decent performance. Unfortunately, it complicates matters considerably. We give a rather abstract version of the caching code, and then sketch the concrete specialization that is in common use. Finally we discuss some pragmatic issues. ...
https://ocw.mit.edu/courses/6-826-principles-of-computer-systems-spring-2002/64d732d98d264f9a61e381bf1358c907_19.pdf
with Caching 2 6.826—Principles of Computer Systems 2002 6.826—Principles of Computer Systems 2002 Action Do(x := x – 1); Do(y := y + 1) Commit Crash before commit X 5 5 5 4 5 Y 5 5 5 6 5 x 5 4 4 4 5 y 5 5 6 6 5 If we want to take account of the possibility that the server (specified ...
https://ocw.mit.edu/courses/6-826-principles-of-computer-systems-spring-2002/64d732d98d264f9a61e381bf1358c907_19.pdf
ph := run >> % aborts any current trans. APROC Do(a) -> V RAISES {crashed} = << IF ph = run => VAR v | (v, vs) := a(vs); RET v [*] RAISE crashed FI >> Here is the previous example extended with the ph variable. Action Begin(); Do(x := x - 1); Do(y := y + 1) Commit Crash before commit X 5 5 5 5 4 5 Y 5...
https://ocw.mit.edu/courses/6-826-principles-of-computer-systems-spring-2002/64d732d98d264f9a61e381bf1358c907_19.pdf
% volatile state % stable log % volatile log % phase (volatile) % recovering APROC Commit() RAISES {crashed} = << IF ph = run => ss := vs; ph := idle [*] RAISE crashed FI >> EXCEPTION crashed APROC Abort () = << vs := ss; ph := idle >> APROC Crash () = << vs := ss; ph := idle >> % same as Crash % ‘aborts’ the...
https://ocw.mit.edu/courses/6-826-principles-of-computer-systems-spring-2002/64d732d98d264f9a61e381bf1358c907_19.pdf
the crash does % the recovery action vl := sl; Redo(); vs := ss; rec := false PROC Redo() = % sl = vl before this is called % replay vl, then clear sl DO vl # {} => << ss := ss + {vl.head} >>; << vl := vl.tail >> OD << sl := {} >> Action Begin(); Do(x := x - 1); Do(y := y + 1) Commit Redo: apply x:=4 Redo:...
https://ocw.mit.edu/courses/6-826-principles-of-computer-systems-spring-2002/64d732d98d264f9a61e381bf1358c907_19.pdf
l k = {} RET Here is what this code does for the previous example, assuming for simplicity that A = U. You may wish to apply the abstraction function to the state at each point and check that each action simulates an action of the spec. \/ (EXISTS k', l'| k = k' + l' /\ l' # {} /\ l' <= l /\ IsHiccups(k', l) ) be...
https://ocw.mit.edu/courses/6-826-principles-of-computer-systems-spring-2002/64d732d98d264f9a61e381bf1358c907_19.pdf
block. Then you can tell whether a U has already been applied just by comparing its VN with the VN of the block. For example, if the version number of the update is 23: Original Idempotent The disk block x: Int x: Int vn: Int The update x := x + 1 IF vn = 22 => x := x + 1; [*] SKIP FI vn := 23 Note: vn = ...
https://ocw.mit.edu/courses/6-826-principles-of-computer-systems-spring-2002/64d732d98d264f9a61e381bf1358c907_19.pdf
= << vs := ss; vl := {}; vph := idle >>; vph := sph; vl := (vph = idle => {} [*] sl); Redo(); vs := ss % what the crash does % the recovery % action PROC Redo() = % replay vl, then clear sl DO vl # {} => << ss := ss + {vl.head} >>; << vl := vl.tail >> OD DO sl # {} => << sl := sl.tail >> OD; << sph := idle; vph...
https://ocw.mit.edu/courses/6-826-principles-of-computer-systems-spring-2002/64d732d98d264f9a61e381bf1358c907_19.pdf
8 6.826—Principles of Computer Systems 2002 6.826—Principles of Computer Systems 2002 Caching We would like to have code for SequentialTr that can run fast. To this end it should: 1. Allow the volatile state vs to be cached so that the frequently used parts of it are fast to access, but not require a complete ...
https://ocw.mit.edu/courses/6-826-principles-of-computer-systems-spring-2002/64d732d98d264f9a61e381bf1358c907_19.pdf
stable state, which no longer exists. So there’s an incentive to limit the amount by which the background process runs behind. Normally the volatile state consists of entries in the cache. Although the abstract code below does not make this explicit, the cache usually contains the most recent values of variables, th...
https://ocw.mit.edu/courses/6-826-principles-of-computer-systems-spring-2002/64d732d98d264f9a61e381bf1358c907_19.pdf
can abort a transaction after installing some parts of the cache to the stable state, we have to follow the “write ahead log” or WAL rule, which says that before a cache entry can be installed, all the actions that affected that entry (and therefore all their undo’s) must be in the stable log. Although we don’t wan...
https://ocw.mit.edu/courses/6-826-principles-of-computer-systems-spring-2002/64d732d98d264f9a61e381bf1358c907_19.pdf
2002 6.826—Principles of Computer Systems 2002 UndoLog : sequence of undo actions (not updates) The essential step is installing a cache update into the stable state. This is an internal action, so it must not change the abstract stable or volatile state. As we shall see, there are many ways to satisfy this requi...
https://ocw.mit.edu/courses/6-826-principles-of-computer-systems-spring-2002/64d732d98d264f9a61e381bf1358c907_19.pdf
:= L{} undoing := false % Stable State % Stable Log % Stable Undo Log % Cache (dirty part) % Volatile Log % Volatile Undo Log % Volatile PHase % Permanent Log Note that there are two logs, called L and UL (for undo log). A L records groups of updates; the difference between an update U and an action A is th...
https://ocw.mit.edu/courses/6-826-principles-of-computer-systems-spring-2002/64d732d98d264f9a61e381bf1358c907_19.pdf
of the undo action to vl/sl. For example, after un1, un2, and un3 have been processed, ul might be un0 un1 un2 cancel un3 cancel cancel = un0 un1 un2 cancel cancel = un0 un1 cancel = un0 Of course many other encodings are possible, but this one is simple and fits in well with the rest of the code. Examples of a...
https://ocw.mit.edu/courses/6-826-principles-of-computer-systems-spring-2002/64d732d98d264f9a61e381bf1358c907_19.pdf
vl)–(sul+vul) SequentialTr.ph = (~undoing => vph idle) % INVARIANTS [1] (ALL l1, l2 | sl = l1 + {commit} + l2 /\ ~ commit IN l2 ==> ss + l1 = ss + sl - sul ) % Stable undos cancel % uncommitted tail of sl [2] ss + sl = ss + sl + vl - vul % Volatile undos cancel vl [3] ~ undoing ==> ss + sl + vl = ss ++ c % C...
https://ocw.mit.edu/courses/6-826-principles-of-computer-systems-spring-2002/64d732d98d264f9a61e381bf1358c907_19.pdf
ss ++ {w} + sl + sl = ss % Any cache entry can be installed VAR ul := sul + vul, i := 0 | DO ul # {} => VAR un := ul.last | ul := ul.reml; IF un=cancel => i := i+1 [*] i>0 => i := i-1 [*] Apply(un, cancel) FI OD; undoing := false % Every entry in sul + vul has a cancel, and everything is undone in vs. % Background...
https://ocw.mit.edu/courses/6-826-principles-of-computer-systems-spring-2002/64d732d98d264f9a61e381bf1358c907_19.pdf
w} = ss => c := c - {w} >> APROC Truncate() = << VAR l1, l2 | % Move some of sl to pl sl = l1 + l2 /\ ss + l2 = ss + sl => pl := pl + l1; sl := l2 >> % Media recovery The idea is to reconstruct the stable state ss from the permanent log pl by redoing all the updates, starting with a fixed initial ss. Details are l...
https://ocw.mit.edu/courses/6-826-principles-of-computer-systems-spring-2002/64d732d98d264f9a61e381bf1358c907_19.pdf
) [] VAR ul1, un, ul2 | un # cancel /\ ul = ul1 + {un, cancel} + ul2 => RET UndoLog(s, ul1 + ul2) FI A cache is a set of commuting update functions. When we combine two caches c1 and c2, as we do in apply, we want the total effect of c1 and c2, and all the updates still have to commute and be atomic updates. The Com...
https://ocw.mit.edu/courses/6-826-principles-of-computer-systems-spring-2002/64d732d98d264f9a61e381bf1358c907_19.pdf
if the existing systems don’t implement two-phase commit, you can still build a multi-level system on them. • Often you can get more concurrency by allowing lower level transactions to commit and release their locks. For example, a B-tree typically holds locks on whole disk blocks to maintain the integrity of the i...
https://ocw.mit.edu/courses/6-826-principles-of-computer-systems-spring-2002/64d732d98d264f9a61e381bf1358c907_19.pdf
’s are also used to code the WAL guard in Install and the guard in Truncate. It’s OK to install a W if the LSN of the last entry in sl is at least as big as the n of the W. It’s OK to drop a U from the front of sl if every uninstalled W in the cache has a bigger LSN. The simplest case is a block equal to a single di...
https://ocw.mit.edu/courses/6-826-principles-of-computer-systems-spring-2002/64d732d98d264f9a61e381bf1358c907_19.pdf
u.ba ==> w commutes with u, because u only looks at u.ba. Stated precisely: (ALL s | (ALL ba | ba # u.ba ==> u(s)(ba) = s(ba) /\ (ALL s' | s(u.ba) = s'(u.ba) ==> u(s)(u.ba) = u(s')(u.ba)) ) So the guard in Install testing whether w is already installed is just (EXISTS u | u IN vl /\ u.ba = w.a) because in Do we get...
https://ocw.mit.edu/courses/6-826-principles-of-computer-systems-spring-2002/64d732d98d264f9a61e381bf1358c907_19.pdf
current one, or to issue a stop payment order for a check that was printed erroneously. Or it might require manual intervention, for instance, calling up the bank customer and asking what really happened in yesterday’s ATM transaction. Usually compensation is not perfect. Because compensation is complicated and imp...
https://ocw.mit.edu/courses/6-826-principles-of-computer-systems-spring-2002/64d732d98d264f9a61e381bf1358c907_19.pdf
18.465 notes, March 29, 2005, revised May 2 The spatial median In one dimension, for any probability distribution function F with a finite first mo- ment, the medians are exactly the values of m for which ∫ |x −m|dF (x) is minimized, using a definition allowing an interval of medians on which the distribution function ...
https://ocw.mit.edu/courses/18-465-topics-in-statistics-nonparametrics-and-robustness-spring-2005/64fb1ad786d5fc7157b0f30b430e7c9a_spatialmedian.pdf
be called strictly convex if whenever x (cid:6)= y ∈ C and 0 < λ < 1 we have f (λx + (1 − λ)y) < λf (x) + (1 − λ)f (y). It’s easily seen that a function f on an interval is convex if its second derivative is nonnegative and strictly convex if its second derivative is strictly positive. For example, on R1 , f (x) = x...
https://ocw.mit.edu/courses/18-465-topics-in-statistics-nonparametrics-and-robustness-spring-2005/64fb1ad786d5fc7157b0f30b430e7c9a_spatialmedian.pdf
0) := |s − x| − |s0 − x| ≤ |s − s0|, g(s) := M (s, P, s0) is always finite. Clearly, it’s continuous in s, and goes to ∞ as s → ∞ for fixed s0. Thus the infimum of M (s, P, s0) is attained, and a spatial median always exists. Changing s0 only adds a constant to the integral, so the minimization doesn’t depend on s0. ...
https://ocw.mit.edu/courses/18-465-topics-in-statistics-nonparametrics-and-robustness-spring-2005/64fb1ad786d5fc7157b0f30b430e7c9a_spatialmedian.pdf
Notes. Haldane (1948) proved uniqueness of the spatial median in Rk, k ≥ 2. (In Haldane’s proof, note that d2R/dx2 > 0 unless yr = 0 for all r, in which case all the observations are on a line.) Haldane gives the proof in detail for a finite sample (empirical measure). The device of taking |s − x| − |s0 − x| in plac...
https://ocw.mit.edu/courses/18-465-topics-in-statistics-nonparametrics-and-robustness-spring-2005/64fb1ad786d5fc7157b0f30b430e7c9a_spatialmedian.pdf
Engineering Risk Benefit Analysis 1.155, 2.943, 3.577, 6.938, 10.816, 13.621, 16.862, 22.82, ESD.72, ESD.721 DA 2. The Value of Perfect Information George E. Apostolakis Massachusetts Institute of Technology Spring 2007 DA 2. The Value of Perfect Information 1 Recall the evaluation of the survey results (Slide 14, DA...
https://ocw.mit.edu/courses/esd-72-engineering-risk-benefit-analysis-spring-2007/6515ae688d570abf0946ec1059c38644_da2.pdf
/s)=0.706 P(L3/s)=0.294 P(L4/s)=0.000 1.000 P[L2 materializes/survey says L2] = 0.706, because the survey is not perfect. DA 2. The Value of Perfect Information 4 Bayes’ Theorem for the Clairvoyant P[L2 materializes/ CV says L2] = CVsaysL L/ 2 2 materializ L(xP)es materializ )es 2 = (P CVsaysL L/ i 2 materializ L(xP)...
https://ocw.mit.edu/courses/esd-72-engineering-risk-benefit-analysis-spring-2007/6515ae688d570abf0946ec1059c38644_da2.pdf
.0 α=$195K L2 $300K L3 $100K L4 -$100K L1 $150K DA 2. The Value of Perfect Information 8 The value of alpha α: EMV, if the terminal decision is to be made with perfect information at no cost. α = 0.3x300 + 0.5x150 + 0.2x150 = $195K DA 2. The Value of Perfect Information 9 The value of beta • What is the EMV without ...
https://ocw.mit.edu/courses/esd-72-engineering-risk-benefit-analysis-spring-2007/6515ae688d570abf0946ec1059c38644_da2.pdf
considering whether to actually do it. DA 2. The Value of Perfect Information 13 Summary and Observations • We have developed single-attribute, multi-stage sequential Decision Trees. • The model is useful to a single decision maker. • Decision Criterion: Maximize the EMV. • Maximizing the EMV is not the best decisi...
https://ocw.mit.edu/courses/esd-72-engineering-risk-benefit-analysis-spring-2007/6515ae688d570abf0946ec1059c38644_da2.pdf
6.012 - Microelectronic Devices and Circuits - Fall 2005 Lecture 3-1 Lecture 3 - Semiconductor Physics (II) Carrier Transport September 15, 2005 Contents: 1. Thermal motion 2. Carrier drift 3. Carrier diffusion Reading assignment: Howe and Sodini, Ch. 2, §§2.4-2.6 6.012 - Microelectronic Devices and Circuits ...
https://ocw.mit.edu/courses/6-012-microelectronic-devices-and-circuits-fall-2005/652b369381b25855a652d351077fe3b5_lecture3anotat.pdf
devices 6.012 - Microelectronic Devices and Circuits - Fall 2005 Lecture 3-5 2. Carrier Drift Apply electric field to semiconductor: E ≡ electric field [V /cm] ⇒ net force on carrier F = ±qE E - Between collisions, carriers accelerate in direction of field: v(t) = at = − qE t mn for electrons v(t) = qE t mp ...
https://ocw.mit.edu/courses/6-012-microelectronic-devices-and-circuits-fall-2005/652b369381b25855a652d351077fe3b5_lecture3anotat.pdf
• for medium and high doping level, µ limited by colli- sions with ionized impurities • holes ”heavier” than electrons: → for same doping level, µn > µp 6.012 - Microelectronic Devices and Circuits - Fall 2005 Lecture 3-8 Drift current Net velocity of charged particles ⇒ electric current: Drift current densi...
https://ocw.mit.edu/courses/6-012-microelectronic-devices-and-circuits-fall-2005/652b369381b25855a652d351077fe3b5_lecture3anotat.pdf
1E-3 1E-4 p-Si n-Si 1E+12 1E+13 1E+14 1E+15 1E+16 1E+17 1E+18 1E+19 1E+20 1E+21 Doping (cm-3) 6.012 - Microelectronic Devices and Circuits - Fall 2005 Lecture 3-11 Numerical example: • Si with Nd = 3 × 1016 cm−3 at 300 K µn (cid:2) 1000 cm2/V · s ρn (cid:2) 0.21 Ω · cm • apply |E| = 1 kV /cm |vdn| (cid...
https://ocw.mit.edu/courses/6-012-microelectronic-devices-and-circuits-fall-2005/652b369381b25855a652d351077fe3b5_lecture3anotat.pdf
Fn = −Dn dx dp Fp = −Dp dx Dn ≡ electron diffusion coefficient [cm2/s] Dp ≡ hole diffusion coefficient [cm2/s] D measures the ease of carrier diffusion in response to a concentration gradient: D ↑ ⇒ F dif f ↑. D limited by vibrating lattice atoms and ionized dopants 6.012 - Microelectronic Devices and Circuits - Fall 2...
https://ocw.mit.edu/courses/6-012-microelectronic-devices-and-circuits-fall-2005/652b369381b25855a652d351077fe3b5_lecture3anotat.pdf
current can flow by drift and diffusion sepa- rately. Total current: Jn = J drif t + J dif f = qnµnE + qDn n n dn dx Jp = J drif t + J dif f p p = qpµpE − qDp dp dx And Jtotal = Jn + Jp 6.012 - Microelectronic Devices and Circuits - Fall 2005 Lecture 3-17 Summary: relationship between v, F , and J In semicon...
https://ocw.mit.edu/courses/6-012-microelectronic-devices-and-circuits-fall-2005/652b369381b25855a652d351077fe3b5_lecture3anotat.pdf
Clocking Figure by MIT OCW. 6.884 - Spring 2005 2/18/05 L06 – Clocks 1 Why Clocks and Storage Elements? Inputs Combinational Logic Outputs Want to reuse combinational logic from cycle to cycle 6.884 - Spring 2005 2/18/05 L06 – Clocks 2 Digital Systems Timing Conventions ƒ All digital systems need a conventi...
https://ocw.mit.edu/courses/6-884-complex-digital-systems-spring-2005/6532a35551bb8707fa017de5e53c66c7_l06_clocks.pdf
a Latch Latches are a mux, clock selects either data or output value Q CMOS Transmission Gate Latch D 0 1 CLK Optional input buffer D’ D CLK CLK Usually have local inverter to generate CLK Q Q Optional output buffer CLK Parallel N and P transistors act as switch, called a “transmission gate” 6.884 - ...
https://ocw.mit.edu/courses/6-884-complex-digital-systems-spring-2005/6532a35551bb8707fa017de5e53c66c7_l06_clocks.pdf
ailing Setup CLK CLK Q D CLK CLK If data arrives too close to clock edge, it won’t set up the feedback loop before clock closes the input transmission gate. 6.884 - Spring 2005 2/18/05 L06 – Clocks 10 The Hold Time Race CLK CLK Q D CLK CLK Added clock buffers to demonstrate positive hold time on this ...
https://ocw.mit.edu/courses/6-884-complex-digital-systems-spring-2005/6532a35551bb8707fa017de5e53c66c7_l06_clocks.pdf
– only found in high performance custom devices 6.884 - Spring 2005 2/18/05 L06 – Clocks 14 Flip-Flop Timing Parameters Clock D Q TCQmin TCQmax Tsetup Thold ƒ TCQmin/TCQmax – propagation in→out at clock edge ƒ Tsetup/Thold – define window around rising clock edge during which data must be steady to be samp...
https://ocw.mit.edu/courses/6-884-complex-digital-systems-spring-2005/6532a35551bb8707fa017de5e53c66c7_l06_clocks.pdf
ps directly, no local buffers 6.884 - Spring 2005 2/18/05 L06 – Clocks 18 H-Trees ƒ Recursive pattern to distribute signals uniformly with equal delay over area ƒ Uses much less power than grid, but has more skew ƒ In practice, an approximate H-tree is used at the top level (has to route around functional blocks)...
https://ocw.mit.edu/courses/6-884-complex-digital-systems-spring-2005/6532a35551bb8707fa017de5e53c66c7_l06_clocks.pdf
Oscillator Circuit Divide by N 6.884 - Spring 2005 2/18/05 L06 – Clocks 23 Intel Itanium Clock Distribution Global Distribution Regional Distribution Local Distribution GCLK Regional Grid RCD RCD DLCLK OTB Reference clock CLKP CLKN VCC /2 PLL Main clock DSK DSK DSK DSK = Active deskew circuits,...
https://ocw.mit.edu/courses/6-884-complex-digital-systems-spring-2005/6532a35551bb8707fa017de5e53c66c7_l06_clocks.pdf
– worst case is when CLK2 is earlier/later than CLK1 ƒ Fast path timing constraint TCQmin + TPmin ≥ Thold + Tskew – worst case is when CLK2 is earlier/later than CLK1 6.884 - Spring 2005 2/18/05 L06 – Clocks 27
https://ocw.mit.edu/courses/6-884-complex-digital-systems-spring-2005/6532a35551bb8707fa017de5e53c66c7_l06_clocks.pdf
6.S897/HST.956 Machine Learning for Healthcare Lecture 1: What makes healthcare unique? Instructors: David Sontag, Peter Szolovits The Problem Healthcare costs in the US amount to over $3 trillion and are rapidly rising. The US has some of the best clinicians in the world, but there are still many cases of chronic...
https://ocw.mit.edu/courses/6-s897-machine-learning-for-healthcare-spring-2019/655ed887d3a9c977a81864aec4139ac8_MIT6_S897S19_lec1note.pdf
not generalize well across different populations. The RX Project is an AI designed for automated knowledge acquisition. Figure 1 displays the discovery system that combines empirical data with a knowledge base that combines with researchers to generate and evaluate hypotheses about causal relationships to create new ...
https://ocw.mit.edu/courses/6-s897-machine-learning-for-healthcare-spring-2019/655ed887d3a9c977a81864aec4139ac8_MIT6_S897S19_lec1note.pdf
This highlights a theme for the rest of the course: that new policy can open the door to innovation. 2.2 Data New medical datasets are now publicly available. Mimic, the only publicly available EHR dataset, was created out of MIT, consisting of intensive care unit patient records. From the MIMIC Physiotnet website...
https://ocw.mit.edu/courses/6-s897-machine-learning-for-healthcare-spring-2019/655ed887d3a9c977a81864aec4139ac8_MIT6_S897S19_lec1note.pdf
Resources (FHIR) is a standardized API protocol for sending EHR data. Observational Health and Data Science Informatics (OHDSI) is a common data model for EHR data. 2.4 Advances in Machine Learning Breakthroughs in machine learning for tasks such as image recognition and language translation have inspired new confi...
https://ocw.mit.edu/courses/6-s897-machine-learning-for-healthcare-spring-2019/655ed887d3a9c977a81864aec4139ac8_MIT6_S897S19_lec1note.pdf
. 3.1 Imagining the emergency department of the future Professor Sontag has been working with Beth Israel Deaconess Medical Center to use technology in their Emergency Department. The ER is an interesting setting because it faces extreme constraints on a daily basis, these include: limited resources, time-sensitiv...
https://ocw.mit.edu/courses/6-s897-machine-learning-for-healthcare-spring-2019/655ed887d3a9c977a81864aec4139ac8_MIT6_S897S19_lec1note.pdf
practices. Professor Sontag believes this could be particularly useful in academic medical centers to aid in the training of new doctors and in less populated areas where doctors might need to cover a broader set of conditions. 3.2 Efficient healthcare workflows Machine learning can make many healthcare workflows more ...
https://ocw.mit.edu/courses/6-s897-machine-learning-for-healthcare-spring-2019/655ed887d3a9c977a81864aec4139ac8_MIT6_S897S19_lec1note.pdf
, patient data is particularly sensitive, so de-identification is necessary and negotiating data sharing agreements is time-consuming. Due to the complexity of the healthcare system, there is often missing data and different training data vs testing data distributions. Production EHR systems might be difficult to work w...
https://ocw.mit.edu/courses/6-s897-machine-learning-for-healthcare-spring-2019/655ed887d3a9c977a81864aec4139ac8_MIT6_S897S19_lec1note.pdf
.05225, 2017. [SA75] Edward H Shortliffe and Stanton G Axline. Computer-based consultations in clinical therapeutics: Explanation and rule acquisition capabilities of the mycin. 1975. 6.S897/HST.956 Machine Learning for Healthcare — Lec1 — 6 ...
https://ocw.mit.edu/courses/6-s897-machine-learning-for-healthcare-spring-2019/655ed887d3a9c977a81864aec4139ac8_MIT6_S897S19_lec1note.pdf
Lecture 2 L t 2 Semiconductor Physics (I) Outline Intrinsic bond model : electrons and holes • Intrinsic bond model : electrons and holes • • Generation and recombination • Intrinsic semiconductor • Doping: Extrinsic semiconductor • Charge Neutrality ead g ss g Reading Assignment: e t Howe and Sodini; Chapter 2. ...
https://ocw.mit.edu/courses/6-012-microelectronic-devices-and-circuits-spring-2009/6583f0018f5bb90464ea68b550f73c95_MIT6_012S09_lec02.pdf
- 3 Simple “flattened” model of Si crystal 4 valence electrons (– 4 q), contributed by each ion silicon ion (+ 4 q) border of bulk silicon region two electrons in bond At 0K At 0K: • All bonds are satisfied – ⇒ all valence electrons engaged in bonding N “f • No “free” electrons ” t l 6.012 Lecture 2 Elec...
https://ocw.mit.edu/courses/6-012-microelectronic-devices-and-circuits-spring-2009/6583f0018f5bb90464ea68b550f73c95_MIT6_012S09_lec02.pdf
bringing together electron and hole • Releases energy in thermal or optical form • Recombination rate: • 1 recombination event requires 1 electron + 1 R = [cm−3 • s−1 ] hole ⇒ R ∝ n • p Generation and recombination most likely at surfaces where periodic crystalline structure is broken 6.012 Lecture 2 Electroni...
https://ocw.mit.edu/courses/6-012-microelectronic-devices-and-circuits-spring-2009/6583f0018f5bb90464ea68b550f73c95_MIT6_012S09_lec02.pdf
34 Zn Ga Ge As Se 48 49 50 51 52 CdCd InIn SnSn SbSb TeTe 6.012 Lecture 2 Electronic Devices and Circuits - 9 Doping: Donors Cont’d... • 4 electrons participate in bonding • 5th electron easy to release ⇒ – at room temperature, each donor releases 1 electron that is available for conduction 1 electr...
https://ocw.mit.edu/courses/6-012-microelectronic-devices-and-circuits-spring-2009/6583f0018f5bb90464ea68b550f73c95_MIT6_012S09_lec02.pdf
15 15 16 16 Al Si P S IIB 30 31 32 33 34 ZnZn GaGa Ge AsAs SeSe Ge 48 49 50 51 52 Cd In Sn Sb Te 6.012 Lecture 2 Electronic Devices and Circuits - 12 Doping: Acceptors Cont’d... 3 electrons participate in bonding • 3 electrons participate in bonding • 1 bonding site “unsatisfied” making it easy ...
https://ocw.mit.edu/courses/6-012-microelectronic-devices-and-circuits-spring-2009/6583f0018f5bb90464ea68b550f73c95_MIT6_012S09_lec02.pdf
charge neutrality must be satisfied In general: • ρ q po no + Nd Na ρ = q po − no + Nd − Na ( ( ) ) Let us examine this for Nd = 1017 cm-3, Na = 0 We solved this in an earlier example: n o = Nd = 1017cm−3, po = 2 2 n i Nd = 103cm−3 Hence: ρ ≠ 0 !! ρ What is wrong?? 6.012 Lecture 2 Electronic Devices a...
https://ocw.mit.edu/courses/6-012-microelectronic-devices-and-circuits-spring-2009/6583f0018f5bb90464ea68b550f73c95_MIT6_012S09_lec02.pdf
Verilog L3: Introduction to Verilog L3: Introduction to (Combinational Logic) (Combinational Logic) Acknowledgements : Rex Min Verilog References: • Samir Palnitkar, Verilog HDL, Pearson Education (2nd edition). • Donald Thomas, Philip Moorby, The Verilog Hardware Description Language, Fifth Edition, Kluwer Academic ...
https://ocw.mit.edu/courses/6-111-introductory-digital-systems-laboratory-spring-2006/65a101becabd787ab2a6514e967c300b_lec3.pdf
... always @ (posedge clk) begin count <= count + 1; end ... HDL Code “This section of code looks like a counter. My FPGA has some of those...” counter Inferred Macro (cid:132) Place-and-route: with area and/or speed in mind, choose the needed macros by location and route the interconnect M M M M M M M M M M M M M M...
https://ocw.mit.edu/courses/6-111-introductory-digital-systems-laboratory-spring-2006/65a101becabd787ab2a6514e967c300b_lec3.pdf
outbar, sel); input a, b, sel; output out, outbar; assign out = sel ? a : b; assign outbar = ~out; endmodule a b 1 0 sel out outbar (cid:132) Continuous assignments use the assign keyword (cid:132) A simple and natural way to represent combinational logic (cid:132) Conceptually, the right-hand expression is continuous...
https://ocw.mit.edu/courses/6-111-introductory-digital-systems-laboratory-spring-2006/65a101becabd787ab2a6514e967c300b_lec3.pdf
3in (out, in1, in2,in3); (cid:134) bufif1 and bufif0 are tri-state buffers (cid:132) Net represents connections between hardware elements. Nets are declared with the keyword wire. L3: 6.111 Spring 2006 Introductory Digital Systems Laboratory 7 always Procedural Assignment with always Procedural Assignment with (cid...
https://ocw.mit.edu/courses/6-111-introductory-digital-systems-laboratory-spring-2006/65a101becabd787ab2a6514e967c300b_lec3.pdf
Match Assignments MixMix--andand--Match Assignments (cid:132) Procedural and continuous assignments can (and often do) co-exist within a module (cid:132) Procedural assignments update the value of reg. The value will remain unchanged till another procedural assignment updates the variable. This is the main differenc...
https://ocw.mit.edu/courses/6-111-introductory-digital-systems-laboratory-spring-2006/65a101becabd787ab2a6514e967c300b_lec3.pdf
111 Spring 2006 Introductory Digital Systems Laboratory 11 The Power of Verilog The Power of bit Signals Verilog: : nn--bit Signals (cid:132) Multi-bit signals and buses are easy in Verilog. (cid:132) 2-to-1 multiplexer with 8-bit operands: module mux_2_to_1(a, b, out, outbar, sel); input[7:0] a, b; input sel; outpu...
https://ocw.mit.edu/courses/6-111-introductory-digital-systems-laboratory-spring-2006/65a101becabd787ab2a6514e967c300b_lec3.pdf
module maybe_mux_3to1(a, b, c, sel, out); input [1:0] sel; input a,b,c; output out; reg out; always @(a or b or c or sel) begin case (sel) 2'b00: out = a; 2'b01: out = b; 2'b10: out = c; endcase end endmodule Is this a 3-to-1 multiplexer? L3: 6.111 Spring 2006 Introductory Digital Systems Laboratory 14 Incomplete Spe...
https://ocw.mit.edu/courses/6-111-introductory-digital-systems-laboratory-spring-2006/65a101becabd787ab2a6514e967c300b_lec3.pdf
end endmodule (cid:132) …or, fully specify all branches of conditionals and assign all signals from all branches (cid:134) For each if, include else (cid:134) For each case, include default L3: 6.111 Spring 2006 Introductory Digital Systems Laboratory 16 Dangers of Verilog Dangers of : Priority Logic Verilog: Prior...
https://ocw.mit.edu/courses/6-111-introductory-digital-systems-laboratory-spring-2006/65a101becabd787ab2a6514e967c300b_lec3.pdf
if (i[1]) e = 2’b01; else if (i[2]) e = 2’b10; else if (i[3]) e = 2’b11; else e = 2’bxx; end Inferred Result: 2’b11 2’bxx 1 0 2’b10 1 0 2’b01 1 0 2’b00 1 0 e[1:0] i[3] i[2] i[1] i[0] (cid:132) if-else and case statements are interpreted very literally! Beware of unintended priority logic. L3: 6.111 Spring 2006 Introd...
https://ocw.mit.edu/courses/6-111-introductory-digital-systems-laboratory-spring-2006/65a101becabd787ab2a6514e967c300b_lec3.pdf
Example: A 32-bit ALU Function Table A[31:0] B[31:0] 32’d1 32’d1 0 1 0 1 F[0] + - * F[2:0] F2 F1 F0 Function 0 0 0 0 0 1 0 1 0 0 1 1 1 0 X A + B A + 1 A - B A - 1 A * B 00 01 10 F[2:1] R[31:0] L3: 6.111 Spring 2006 Introductory Digital Systems Laboratory 20 Module Definitions Module Definitions 2-to-1 MUX 3-to-1 MUX m...
https://ocw.mit.edu/courses/6-111-introductory-digital-systems-laboratory-spring-2006/65a101becabd787ab2a6514e967c300b_lec3.pdf
// signed arithmetic later assign prod = i0 * i1; endmodule L3: 6.111 Spring 2006 Introductory Digital Systems Laboratory 21 Level ALU Declaration TopTop--Level ALU Declaration (cid:132) Given submodules: A[31:0] B[31:0] module mux32two(i0,i1,sel,out); module mux32three(i0,i1,i2,sel,out); module add32(i0,i1,sum); modu...
https://ocw.mit.edu/courses/6-111-introductory-digital-systems-laboratory-spring-2006/65a101becabd787ab2a6514e967c300b_lec3.pdf
SE which allows simulation at different levels including Behavioral and Post-Place-and- Route L3: 6.111 Spring 2006 Introductory Digital Systems Laboratory 23 More on Module Interconnection More on Module Interconnection (cid:132) Explicit port naming allows port mappings in arbitrary order: better scaling for large...
https://ocw.mit.edu/courses/6-111-introductory-digital-systems-laboratory-spring-2006/65a101becabd787ab2a6514e967c300b_lec3.pdf
XNOR Note distinction between ~a and !a &a ~& | ~| ^ AND NAND OR NOR XOR a < b a > b a <= b a >= b a == b a != b a === b a !== b Relational [in]equality returns x when x or z in bits. Else returns 0 or 1 case [in]equality returns 0 or 1 based on bit by bit comparison L3: 6.111 Spring 2006 Introductory Digital Syst...
https://ocw.mit.edu/courses/6-111-introductory-digital-systems-laboratory-spring-2006/65a101becabd787ab2a6514e967c300b_lec3.pdf
4'b0000; b = 4'b0000; cin = 1'b0; #50; a = 4'b0101; b = 4'b1010; // sum = 1111, cout = 0 #50; a = 4'b1111; b = 4'b0001; // sum = 0000, cout = 1 a = 4'b0000; b = 4'b1111; cin = 1'b1; // sum = 0000, cout = 1 #50; a = 4'b0110; b = 4'b0001; // sum = 1000, cout = 0 #50; end // initial begin endmodule // test_adder L3: 6.111...
https://ocw.mit.edu/courses/6-111-introductory-digital-systems-laboratory-spring-2006/65a101becabd787ab2a6514e967c300b_lec3.pdf
Lectures 8 & 9 M/G/1 Queues Eytan Modiano MIT Eytan Modiano Slide 1 M/G/1 QUEUE Poisson M/G/1 General independent Service times • Poisson arrivals at rate λ • Service time has arbitrary distribution with given E[X] and E[X2] – Service times are independent and identically distributed (IID) – – E[service tim...
https://ocw.mit.edu/courses/6-263j-data-communication-networks-fall-2002/65d9ab519ec4851af812f4b89ffaeedc_Lectures8_9.pdf
1 j=i- N i • E[Wi] = E[Ri] + E[X]E[Ni] = R + NQ/µ – Here we have used PASTA property plus independent service time property • W = R + λW/µ => W = R/(1-ρ) – Using little’s formula Eytan Modiano Slide 5 What is R? (Time Average Residual Service Time) Residual Service Time R(t) X 1 X 3 X 2 X 2 X1 X 4 X...
https://ocw.mit.edu/courses/6-263j-data-communication-networks-fall-2002/65d9ab519ec4851af812f4b89ffaeedc_Lectures8_9.pdf
i Eytan Modiano Slide 7 E[Wi] = E[Ri] + E[X]E[Ni] = R + NQ/µ = R/(1-ρ) Average Residual Service Time (with vacations) Residual Service Time R(t) X 1 V 1 X 2 X 4 X 3 X1 X 2 V1 X4 time -> X 3 M(t) R = [R(t)]= t 0 1 t L(t) 2 R( τ )d τ = 1 (� X i + � t i=1 2 j=1 V 2 j ) 2 R = lim t →∞ E[...
https://ocw.mit.edu/courses/6-263j-data-communication-networks-fall-2002/65d9ab519ec4851af812f4b89ffaeedc_Lectures8_9.pdf
1/µ Each slot = one packet transmission time = 1/µ • Transmission can begin only at start of a slot • If system is empty at the start of a slot, server not available for the duration of the slot (vacation) • E[X] = E[v] = 1/µ • E[X2] = E[v2] = 1/µ2 W = λ / µ2 2(1 − λ/µ) + 1/ µ2 2 /µ = λ / µ 2(µ− λ) + 1...
https://ocw.mit.edu/courses/6-263j-data-communication-networks-fall-2002/65d9ab519ec4851af812f4b89ffaeedc_Lectures8_9.pdf
time units when there is nothing to transmit E[V] = m; E[V2] = m2. WSFDM = WFDM + E[V2]/2E[V] = WFDM + m/2 Eytan Modiano Slide 12 TDM EXAMPLE TDM Frame slot m slot 1 slot 2 . . . slot m • TDM with one packet slots is the same (a session has to wait for its own slot boundary), so W = R/(1-ρ) R = λ= E[X2] ...
https://ocw.mit.edu/courses/6-263j-data-communication-networks-fall-2002/65d9ab519ec4851af812f4b89ffaeedc_Lectures8_9.pdf
6.852: Distributed Algorithms Fall, 2009 Class 7 Today’s plan z Asynchronous systems z Formal model  I/O automata  Executions and traces  Operations: composition, hiding  Properties and proof methods:  Invariants  Simulation relations z Reading: Chapter 8 • Next: – Asynchronous network algorithms: Leader elec...
https://ocw.mit.edu/courses/6-852j-distributed-algorithms-fall-2009/666ea5c2d621cd26c6badf37a750d290_MIT6_852JF09_lec07.pdf
model. • Based on earlier algorithm [Dwork, Lynch, Stockmeyer]. • Algorithm idea: – Processes use unreliable leader election subalgorithm to choose coordinator, who tries to achieve consensus. – Coordinator decides based on active support from majority of processes. – Does not assume anything based on not receivin...
https://ocw.mit.edu/courses/6-852j-distributed-algorithms-fall-2009/666ea5c2d621cd26c6badf37a750d290_MIT6_852JF09_lec07.pdf
Automata z General mathematical modeling framework for reactive components.  Little structure---must add structure to specialize it for networks, shared-memory systems,… z Designed for describing systems in a modular way:  Supports description of individual system components, and how they compose to yield a larger...
https://ocw.mit.edu/courses/6-852j-distributed-algorithms-fall-2009/666ea5c2d621cd26c6badf37a750d290_MIT6_852JF09_lec07.pdf
Used to define fairness (give turns to all tasks).  Needed to guarantee liveness properties (e.g., the system keeps making progress, or eventually terminates). Channel automaton send(m) C receive(m) z Reliable unidirectional FIFO channel between two processes.  Fix message alphabet M. z signature  input a...
https://ocw.mit.edu/courses/6-852j-distributed-algorithms-fall-2009/666ea5c2d621cd26c6badf37a750d290_MIT6_852JF09_lec07.pdf
S s2 S s3 S s4 S s5 ... (if finite, ends in state)  s0 is a start state  (si, Si, si+1) is a step (i.e., in trans) Ȝ, send(a), a, send(b), ab, receive(a), b, receive(b), Ȝ Execution fragments z An I/O automaton executes as follows:  Start at some start state.  Repeatedly take step from current state t...
https://ocw.mit.edu/courses/6-852j-distributed-algorithms-fall-2009/666ea5c2d621cd26c6badf37a750d290_MIT6_852JF09_lec07.pdf
ynchronize on actions”). • Composing finitely many or countably infinitely many automata Ai, i  I: • Need compatibility conditions: – Internal actions aren’t shared: • int(Ai) ˆ acts(Aj) = ‡ – Only one automaton controls each output: • out(Ai) ˆ out(Aj) = ‡ – But output of one automaton can be an input of one or ...
https://ocw.mit.edu/courses/6-852j-distributed-algorithms-fall-2009/666ea5c2d621cd26c6badf37a750d290_MIT6_852JF09_lec07.pdf
“looks good” to each component, it is good overall. z Substitutivity  Can replace a component with one that implements it. Composition: Basic results Theorem 1: Projection  If D  execs(3 Ai) then D|Ai  execs(Ai) for every i.  If E  traces(3 Ai) then E|Ai  traces(Ai) for every i. Composition: Basic resul...
https://ocw.mit.edu/courses/6-852j-distributed-algorithms-fall-2009/666ea5c2d621cd26c6badf37a750d290_MIT6_852JF09_lec07.pdf