Asymptotically Efficient Estimation of the Derivative of the Invariant Density

Size: px
Start display at page:

Download "Asymptotically Efficient Estimation of the Derivative of the Invariant Density"

Transcription

1 Statistical Inerence or Stochastic Processes 6: 89 17, 3. 3 Kluwer cademic Publishers. Printed in the Netherlands. 89 symptotically Eicient Estimation o the Derivative o the Invariant Density NK S. DLLYN and YUY. KUTOYNTS Laboratoire de Statistique et Processus, Université du Maine, venue Messiaen, 785 Le Mans, Cedex 9, France. s: arnak.dalalyan.etu@univ-lemans.r, kutoyants@univ-lemans.r eceived in inal orm 17 pril 1; ccepted 3 May ) bstract. The problem o estimation o the derivative o the invariant density is considered or a one-dimensional ergodic diusion process. The lower minimax bound on the L -type risk o all estimators is proposed and an asymptotically eicient up to the constant) in the sense o this bound kernel-type estimator is constructed. MS 1999 Subject Classiication: 6M5, 6G7, 6G. Key words: ergodic diusion, invariant density, derivative estimation, asymptotical eiciency, Pinsker s constant. 1. Introduction In this work we consider a diusion process X given by the stochastic dierential equation dx t = SX t ) dt + dw t, X, t, 1) where W t denotes a standard Wiener process deined on a probability space,f, P) and the initial value X is a random variable independent o the Wiener process. We suppose that the ollowing condition is uliled: { x } GS) = exp Sv)dv dx <. ) By this condition the process 1) is recurrent positive, that is, there exists an invariant probability distribution with the density unction S y) = 1 { y } GS) exp Sv)dv and the process {X t,t } has ergodic properties 3, 5].

2 9 NK S. DLLYN ND YUY. KUTOYNTS The trend coeicient S ) is supposed to be unknown, so the invariant density S ) is unknown unction, too. The problem o estimation o this density by observations X T = {X t, t T } as T has been considered by many authors see 1,, 9, 1, 1] and reerences therein). Particularly, in 9] a lower bound on the minimax risk o all estimators is proposed and several estimators localtime, kernel-type, unbiased) attaining this bound in the limit are constructed. These estimators are called asymptotically eicient when the time o observation tends to ininity. It is also shown that they are T consistent and asymptotically normal. There are several open problems closely related with the results described above. First, to ind an estimator o the density which is asymptotically eicient in the second order, say, as it is done in distribution unction estimation problem in i.i.d. case by Golubev and Levit 8]. Second, to construct an asymptotically eicient estimator o the trend coeicient S ). Third, to ind an asymptotically eicient estimator o the derivative o the invariant density. One o the possible approaches in these problems is the so-called minimax approach that we develop in this paper. emark that there exists a close relation between the problems cited above. In act, the trend coeicient can be written as Sx) = S x) S x). It can be easily shown that this unction can be estimated with a rate depending on the smoothness o S ). Say, i the unction S ) is k-times continuously dierentiable, then the best possible rate has to be T k/k+1). Thereore, i one considers the estimator o Sx) constructed via some eicient estimators o S x) and S x), then, roughly speaking, the error o estimation o the density Sx) is asymptotically smaller than the error o S x) estimation. Thus, the main problem is to construct an asymptotically eicient estimator o the derivative S x). The present work is devoted to the problem o this derivative S ) estimation. similar problem was discussed by Lucas 11] in the case where X T is stationary and continuous but not necessarily a Markov process. The author obtained some conditions under which a T -convergent kernel-type estimator exists, but pointed out that the class o processes satisying these conditions is quite narrow. We present below two results concerning the same problem. The irst one is the lower minimax bound on the L -type risk o all estimators Section ). The second provides a kernel-type estimator, which is asymptotically eicient in the sense o this bound Section 3). We inish by concluding remarks where we propose some possible generalizations o these results Section 4). Note that the statement o the problem and the results obtained in this paper lie in the ramework o Pinsker s bound approach 14] see 13] or the discussion and reerences) and the schemes o the proos ollow the work by Golubev and Levit 8] see as well the paper by Schipper 15]). The main diiculty in our case, which does not exist in the case o density estimation by i.i.d. data, is the ollowing. When we take sup o the risk on the class o unctions S ) satisying condition ),

3 ESTIMTION OF THE DEIVTIVE 91 we touch the case GS) =, which corresponds to the null recurrent process, and in this case we have no more the law o large numbers. Thereore to prove the asymptotic eiciency o the estimators we need to control the divergence o this integral. It is done using three conditions 3) 5), given below. To ormulate the main results we need some notation. Let C T be the space equipped with the topology o the uniorm convergence) o all continuous unctions deined on,t] and let B T be the σ -algebra o its Borelian subsets. The process X T = {X t, t T } induces a probability measure on the space C T, B T ). This measure and the mathematical expectation with respect to this measure will be denoted by P S and E S, respectively. In the cases where the unction S ) depends on a parameter, we will write P and E instead o P S and E S. The set o all trend coeicients S ), which provide the weak) existence, uniqueness and ergodicity o X t see 3]) will be denoted by. We suppose that the initial value X has the density S ). Under this condition {X t,t} is a stationary process. We consider the nonparametric problem o S ) derivative unction estimation rom continuous path observations and give the exact asymptotics o the quadratic minimax risk. That is, we ind a constant C) such that lim in sup T k/k+1) E S ϑ T x) S x)) dx = C), T ϑ T S where the inimum is taken over the class o all possible estimators. To deine the set we need some additional notation. Let χ denote the indicatoroanevent and F S ) be the probability distribution o the random variable ξ with density unction S ),thatis, F S x) = Pξ x). Throughout this paper the constants do not depending on T and S) will be denoted by C. The parameter space is deined in the ollowing way. Let us ix a real number >and an integer k and deine the ellipsoid k,) = {S ) k+1) S ] } x) dx. We can deine now the ininite-dimensional parameter space as the set o all S k,), such that or a constant D does not depending on S) the ollowing conditions are uliled: 1. There exists apositive constant B does not depending on S) such that sup B S x) ] dx<d, 3) B>B sup B B>B x >B x <B S x) ] ES ξ χ {y>x} F S y) S y) dy] dx<d. 4)

4 9 NK S. DLLYN ND YUY. KUTOYNTS. The ollowing estimate holds: S x) ] ] χ{ξ>x} F S ξ) ES dx<d. 5) S ξ) O course, the set depends on the constants k,, D and B. In the sequel this set will be denoted by D.. Lower Bound In this section we will establish a lower bound on the minimax risk in sup T ϑ T, S ) = in sup E S ϑ T x) S x)) dx, 6) ϑ T S ϑ T S where the in is taken over all estimators o the derivative. THEOEM 1. The ollowing inequality is true lim lim in in sup T k/k+1) T ϑ T, S ) Pk, ), 7) D T ϑ T S D where the Pinsker constant Pk, ) is deined by ) πk + 1)k + 1) k/k+1) Pk, ) = k + 1) 1/k+1). 8) 4k Proo. To prove this inequality we minorate the minimax risk 6) by the minimax risk over a properly chosen parametric amily. The dimension o this amily depends on T and tends to ininity when T. Thus, let us introduce the ollowing parameterizations S x) = i e i x)u x ), x, ], 9) i L where = ln1 + T)and Ux) is k + 1 times dierentiable increasing unction vanishing or x and equal to 1 or x 1. The unctions {e i } i Z are the elements o the trigonometric basis o L, ], thatis, ) πix sin i i>, e i x) = 1 1 i i =, ) πix cos i i<. The positive number L = L T will be chosen later. In the outside o the interval, ] we put S x) = S x) = sgnx)k + ) x ) k+1.

5 ESTIMTION OF THE DEIVTIVE 93 It is evident that the unction S ) deined in this way is k times dierentiable and the kth derivative is continuous. Consequently, the density ) is k + 1 times continuously dierentiable. Since or any ixed, the unction S ) is locally Lipshitz and sup x xs x) x <, there exists a unique solution o 1) see 3]) corresponding to the case S = S. In the sequel we consider the parametric space Ɣ T o all sequences { i } i L such that i G σ i or all i Z such that i L.Here σ i = σ i,t = α ) k 1, i = 1) T i + with ) Tk + 1)k + 1) 1/k+1) α = α T = 11) 4kπ k and σ =. The integer L is chosen to be equal α]. LEMM 1. For any m =, 1,..., k, we have the ollowing estimates x m) sup S v) dv) C k+ T 1/4k+). 1) sup Ɣ T x,] Proo. For m =, using the inequality i G σ i and the deinitions 1) and 11), we obtain x S v) dv sup S x) sup i e i x) i x x i = i = = 4G L α ) k 1/ 1 4Gαk/)+1 Ck+ T i i=1 + T T. 1/4k+) emark that this chain o inequalities proves the bound 1) or m = 1 as well. Suppose that this bound holds or 1,,...,m 1 and prove it or m. Since all the derivatives o the unction U ) up to the order m are bounded, we have S m) x) 1 πi ) m C k+ i + T 1/4k+) i = C i m σ m i + i = C m 1 T α i=1 Cα k m 1 T + T Ck+ T 1/4k+) i k 1 α ) k/ C k+ + i T 1/4k+) Ck+ Ck+ 1/4k+) T 1/4k+).

6 94 NK S. DLLYN ND YUY. KUTOYNTS COOLLY 1. The ollowing relation holds x) = 1 + o T 1) { exp { x ) k+ } i x >, 1 i x. The proo is easy and immediately ollows rom Lemma 1. Let {ξ i } i Z be i.i.d. random variables with common probability density px) such that p ξ i <G, Eξ i ) =, Eξi ) = 1, I = x)] dx = 1 + ε, px) where ε wheng. We introduce a prior distribution on Ɣ T putting with i = σ i ε) ξ i σ i ε) = 1 + ε) T α1 ε) i k 1 or each i dierent rom. The coeicient will be deterministic and equal to. The Fisher inormation I i o this prior distribution with respect to i or i = ) is then equal to 1 + ε)/σ i ε). Since the minimax risk can be evaluated below by the Bayesian one with a correction term which is shown to be small in order), we seek now a lower bound o the Bayesian risk corresponding to a prior distribution, deinedby T ) = in ϑ T E ) + ϑ T x) x)) dx, where E is the mathematical expectation with respect to the probability measure P dx T ) d). Let ψ i, and ψ i,t be the Fourier coeicients on, ] o ) and ϑ T ), respectively, that is, ψ i, = x)e ix) dx, ψ i,t = ϑ T x)e i x) dx. Since {e i )} is an orthonormal sequence and the unction ) belongs to the Hilbert space generated by this sequence, using the Parseval s identity one has T ) in ϑ T E ϑ T x) x)) dx in ϑ T E ) ψ i,t ψ i,. i <L

7 ESTIMTION OF THE DEIVTIVE 95 By two-dimensional van Trees inequality see 4]) T ) 1 + o T 1) ) ]) E ψi, / i + ψ i, / i T E I i ) + I i ) ), 13) + I i + I i <i L where I i ) is deined by the ormula ] S I i ) = E ξ). i The elementary calculations show that x) = x) S x) + i x) E i Using Lemma 1, one can easily prove that x ] sup x) E x which gives us ξ S i v) dv x ξ ] S v) dv. i = o T 1), 14) x) = x)e i x)u x ) + o T 1). 15) i So the partial derivative o ψ,i with respect to i can be evaluated like ollows ψ i, i = = = e i x) i x) dx e i x)u x ) x) dx + 1 o T 1) e i x) x) dx + 1 o T 1). This equality and the elementary identity ei x) + e i x) = 1/ imply ψ i, i + ψ i, i = 1 + o T 1)). 16) It is not diicult to show that I i ) + I i ) = e i x) + e i x)) x) dx + O T 1) = 1 + o T 1). Now the inequality 13) can be rewritten as T ) o T 1) ) <i L σ i ε) Tσ i ε) ε). 17)

8 96 NK S. DLLYN ND YUY. KUTOYNTS Simple calculations show that the last sum is equal to 4kα1 ε) Tk + 1) 1 + o T 1)). eplacing α by expression 11) we get the inequality T ) T k/k+1) Pk,)1 + o T 1))1 ε) with ) πk + 1)k + 1) k/k+1) Pk,)= k + 1) 1/k+1). 18) 4k emark 1. It is important to emphasize that the values 1) and 11) are the solutions o the maximization problem related with the unctional y) = y i Ty i> i + over the set { Ek, ) = y = y i ) ) i> πi k y i }. i> This choice is based on relation 17) and Lemma. Now we prove some lemmas which will permit us to complete the proo o the lower bound. LEMM. There exists a constant D>such that the unctions { S ) } Ɣ T satisy conditions 3) 5), i T is suiciently large. Proo. We show irstly that there exists a constant D such that u) ] ] χ{ξ>u} F ξ) E du<d 19) ξ) sup Ɣ T or all T>T. emark that or z>y>the ollowing inequality is true z ) k+ y ) k+ z y)y ) k+1, it ollows that 1 F y) y) y This leads us to the inequality 1 F ξ) u)e ξ) exp { z y)y ) k+1} dz = ) χ {ξ>u}] du 1 y ) k+1. u) E χ {ξ>u} ξ ) k+ ] du u) u ) k du<1 )

9 ESTIMTION OF THE DEIVTIVE 97 or T large enough. In the same way, the inequality y ) k+ x ) k+1 y ) or y x, implies that Hence u) u F y) y) dy u u) dy y) 1 u ). k+1 u e u )k+1 y u) dy = C F ξ) u)e ξ) ) χ {<ξ<u}] du C u) u) 1 du u) du = C ) < 1 1) i T is large enough. Then, we have and 1 F y) y) dy 1 1 y) dy = o T 1) ) u)e F ξ) ξ) ) χ { 1<ξ<}] du 8 Proceeding like in the proo o ), one can show that Thus, 1 F y) dy y) sup F y) y 1 y) 1 = ) k+ u) du <C. ) ) F ξ) u)e χ {ξ< }] du 1 3) ξ) or T large enough. Combining the inequalities ) 3), we obtain ) χ{ξ>u} F ξ) ] u)e du C. 4) ξ) In the same way it can be shown that ) χ{ξ>u} F ξ) ] u)e du C. 5) ξ)

10 98 NK S. DLLYN ND YUY. KUTOYNTS For u, ], using Lemma 1 we have u) = o T 1) and ) χ{ξ>u} F ξ) ] E ξ) ) 1 F ξ) E χ {ξ>+1}] + ξ) ) 1 + E χ { 1 ξ +1}] + ξ) + E F ξ) ξ) ) χ {ξ< 1}] C. Hence ) χ{ξ>u} F ξ) ] u)e du C ξ) and the inequality 19) is proved. We pass now to the proo o the ollowing estimate u) ] E ξ χ {y>u} F y) y) dy] du<d. 6) To prove it, one has to consider the cases ξ<, ξ, ], ξ, u] and ξ > u. In the irst, third and ourth cases we obtain 6) proceeding like in the proo o 19). For the second one, we have u) ] E ξ C 4 u) ] ] χ {y>u} F y) dy χ { ξ } du y) ] 1 y) dy du u) ] du = D. 7) So we proved 6) which implies the estimate 4). It remains to check the condition 3). We suppose that >andb>. I >B/, then B x) ] dx x) ] dx = C D B.

11 ESTIMTION OF THE DEIVTIVE 99 For B/, we have B x) ] dx C x ) k+ e x )k+ dx C = D B B x ) k+3 e x )k+ dx B ) k+ y e y dy D B ) e y dy D e B ) <De B < D B. 8) This inequality completes the proo o Lemma. LEMM 3. The ollowing relation holds { } 1 πi ) k C T = Ɣ T i <1 ε) i Z { } S D 9) i T is large enough. Proo. Since C T is a subset o Ɣ T, the conditions 3) 5) are satisied or any C T. So, only the condition sup C T k+1) x) ] dx needs to be checked. Note that k+1) x) = S k) x) + P S k 1) x),..., S x) )] x), where Pz 1,...,z k ) is a polynomial. By Lemma 1 S m) x) = o T 1), m =, 1,...,k 1. So, on the one hand, P S k 1) x),..., S x) )] = ot 1) or x, ]. On the other hand, using the orthonormality o the trigonometric basis e i and the deinition o C T, we obtain 4 S k) x) x) ] 1 + o T 1) dx = = 1 + o T 1) i Z < 1 ε) 1 + o T 1) ) S k) x) ] dx πi ) k i

12 1 NK S. DLLYN ND YUY. KUTOYNTS or any C T. Consequently, k+1) x) ] dx = 4 S k) x) x) ] dx1 + ot 1)) 1 ε)1 + o T 1)). 3) It can be easily checked that k+1) x) ] dx C. 31) x > Combining 3) and 31) we obtain k+1) x) ] dx 1 ε) + ot 1). This completes the proo o Lemma 3. LEMM 4. The probability o the event S consequently S D ) = ot 1 ). Proo. The relation 9) implies S D ) C T ). D is exponentially small and emark now that 1 πi ) ] k E i i Z = 1 πi ) kσi ε) i Z = 1 ε) k o T 1) ) 1 + ε). Hence, or T suiciently large 1 πi ) ] k E i 1 ε). i Z The Hoeding s inequality gives us the ollowing upper bound { } 1 πi ) k C T ) = i 1 ε) i Z { } 1 πi ) k i Ei ) ε1 ε) i Z exp { ε 1 ε) }, Q

13 ESTIMTION OF THE DEIVTIVE 11 where Q = So we have G4 4 i Z πi ) 4kσ i C T i <α CT α 4k+1 = CT 1/k+1). i 4k α i S D ) C T ) exp{ CT 1/k+1) }. ) k 1 C i k α k T i <α TheLemma4isproved. Now everything is ready to inish the proo o Theorem 1. Note irst that we can consider only those estimators ϑ T or which T ϑ T, ) = E ϑ T x) x)) dx<1 with ) = S ) and S is the unction S corresponding to the value =. The set o all estimators satisying this inequality will be denoted by W T. For the other estimators the result is evident. We have the ollowing obvious inequalities: in ϑ T W T sup T ϑ T, S ) S D in ϑ T W T sup T ϑ T, S ) S D Ɣ T in T ϑ T,)d) ϑ T W T Ɣ T sup T ϑ T,)d) ϑ T W T Ɣ T \ D T ) sup T ϑ T,)d). ϑ T W T Ɣ T \ D We have already ound a lower bound or the irst term. The second term can be bounded as ollows sup ϑ T W T Ɣ T \ D T ϑ T,)d) 8 sup Ɣ T + )S D ). It ollows rom Lemma 1 that the L norm o is bounded uniormly on Ɣ T. Consequently, using the Lemma 4 we obtain in ϑ T W T Thereore sup T ϑ T, S ) T ) ot 1 ). S D lim lim in in sup T k/k+1) T ϑ T, S ) Pk,)1 ε). D T ϑ T S D

14 1 NK S. DLLYN ND YUY. KUTOYNTS This completes the proo o inequality 7), since ε can be taken as small as we want. 3. Eiciency o a Kernel-type Estimator Now we have a lower bound or the minimax risk, to prove its optimality we have to ind an estimator achieving it. So, we investigate in this section the behavior o the ollowing estimator T ϑ K,T x) = K T x X t )χ { Xt <B T T } dx t, where K T ) is a kernel-type unction and B T is a positive number. Let us denote with and K T x) = α,t K xα,t ) 3) K x) = 1 π 1 1 u k ) cosux) du ) πtk + 1)k + 1) 1/k+1) α,t =. 4k We show in this section that, or B T = T, the estimator ϑ K,T is asymptotically eicient, that is, this estimator achieves the lower bound obtained in the previous section. This means that the ormula 3) gives the optimal kernel in the problem o the irst derivative o the invariant density estimation. Particularly, i k =, then the optimal kernel has the ollowing orm K sin x x cos x) x) =. πx 3 For x = this unction is equal to /3. THEOEM. We have lim lim sup in sup T k/k+1) T ϑ T, S ) Pk,). D T ϑ T S D Proo. It is evident that in ϑ T sup T ϑ T, S ) in S D K sup T ϑ K,T, S ). S D Hence, it is suicient to evaluate the risk T K, S ) = E S ϑ K,T x) S x)) dx,

15 ESTIMTION OF THE DEIVTIVE 13 where S ) D is the unknown trend coeicient. To evaluate this risk we will use the Fourier transormations. Let us denote ϕ S λ) = e iλx S x) dx, ϕ K,T λ) = e iλx ϑ K,T x) dx, T ϕ K λ) = e iλx K T x) dx, ϕ T λ) = 1 T By Parseval s identity T K, S ) = 1 π E S ϕ K,T λ) ϕ S λ) dλ. e iλx t χ { Xt <B T } dx t. s the estimator ϑ K,T is a convolution, its Fourier transorm is product o two Fourier transorms. Indeed ϕ K,T λ) = T e iλx K T x X t )χ { Xt <B T T } dx t dx = T e iλx K T x X t ) dx χ { Xt <B T T } dx t T = e iλx t ϕ K λ)χ { Xt <B T T } dx t = ϕ K λ) ϕ T λ). So the quadratic risk can be rewritten as T K, S ) = 1 π E S ϕ K λ)ϕ T λ) ϕ S λ) dλ = 1 E ϕk S λ)ϕ T λ) ϕ S λ) dλ π = ϕ K λ) Var S ϕ T λ)] dλ + π + 1 ϕk λ)e S ϕ T λ) ϕ S λ) dλ. π For the mathematical expectation o ϕ T λ),wehave E S ϕ T λ) = 1 T ] T E S e iλx t SX t )χ { Xt <B T } dt = E S e iλξ ] Sξ)χ { ξ <BT } = 1 ϕ Sλ) 1 e iλu S u) du. u >B T The ollowing lemma describes the behavior o the bias and variance terms and tells us how to choose B T.

16 14 NK S. DLLYN ND YUY. KUTOYNTS LEMM 5. I the unction K T ) is such that ϕ K λ) 1 or each λ, then there exists a constant C = C D such that, or any S D, ϕ K λ) ϕk Var S ϕ T λ)] dλ + C + CB ] T, T T T ϕk λ)e S ϕ T λ)] ϕ S λ) ϕk dλ 1)ϕ S + C ], where denotes the L -norm over. Proo. We prove here only the irst inequality, the second one can be proved in the same way. Note that Itô ormula gives us the ollowing representation with ϕ T λ) E S ϕ T λ)] = H S λ, X T ) H S λ, X ) + T + e iλx t g S λ, X t ) ] dw t BT g S λ, y) = e iλu S u) χ {u<y} F S y) du B T S y) and H S λ, x) = x g Sλ, y) dy. Consequently, Var S ϕ T λ)] = T HS E S λ, X T ) H S λ, X ) + T + e iλx t χ { Xt <B T } g S λ, X t ) ) dw t. Using the triangle inequality we get ϕ K λ) Var S ϕ T λ)] dλ ) 3 33) with 1 = 1 T ϕ T K λ) E S e iλx t χ { Xt <B T } dw t dλ 1 ϕ K λ) dλ, T = 1 ϕ T K λ) E HS S λ, ξ) dλ, 3 = 1 T ϕ T K λ) E S g S λ, X t ) dw t dλ. B T

17 ESTIMTION OF THE DEIVTIVE 15 To evaluate the term we use the act that ϕ K λ) is less than 1. Thus, according Parseval s identity and condition 4) we have 4 T E BT ξ S e iλu S u) χ {y>u} F S y) dy du B T S y) dλ = 8π BT T S u) ] ξ ] χ {y>u} F S y) ES dy du 8πDB T. B T S y) T epeating exactly the same arguments one can check that 3 8πD T. Lemma 5 is proved. We choose the kernel-type unction K T ) in the ollowing way ϕ K λ) = ϕ α λ) = 1 λ ) k α, + where α = α T is a positive number. s we will see below, the integrals o the righthand sides in Lemma 5 converge both to zero with the rate T k/k+1). Thereore the choice B T = T gives us the ollowing upper estimate or quadratic risk T K, S ) L T ϕ α,ϕ S ) 1 + o T 1) ), where o T 1) tends to zero uniormly on S and L T ϕ α,ϕ S ) = 1 4 ϕα λ) + T ϕ α λ) 1 ϕ S λ) ) dλ. πt Since the unction S ) is in the ellipsoid k,), its Fourier transorm should belong to the ollowing set { } 1 = ϕ λ k ϕλ) dλ. π eplacing in L T the unction ϕ α by its explicit expression, we obtain L T ϕ α,ϕ S ) = α 1 λ k ) πt α dλ + 1 α λ k ϕ πα k S λ) dλ. α Since ϕ S belongs to, the second term o the right-hand side is less than /α k and the irst term can be calculated explicitly: α ) 1 λ 4αk dλ = k α k + 1)k + 1). α α

18 16 NK S. DLLYN ND YUY. KUTOYNTS It leads us to the ollowing inequality { in sup L T ϕ α,ϕ S ) in α> S α> 8k α πtk + 1)k + 1) + α k } = in α> Gα). The unction Gα) is continuously dierentiable and strictly convex, consequently it attains the minimum at the point α satisying the ollowing equation 8k πtk + 1)k + 1) = k α k+1 which leads to πt k + 1)k + 1) α = 4k, ) 1/k+1) and in Gα) = Gα k + 1) ) = α> α k = Pk,)T k/k+1). This completes the proo o Theorem. 4. emarks 1. One can use exactly the same arguments to ind the Pinsker s constant in the problem o l) S ) estimation when S k + l 1,). The optimal rate o convergence ϕ T and the Pinsker s constant are then ϕ T = T β, ) πk + l 1)k + l 1) β P l k, ) = 1 + β) 1 1+β 4k with β = k/k + l 1).. The condition 5) can be replaced by B ] sup B τ S x) E χ{ξ>x} F S ξ) S dx<, S B S ξ) where τ is positive and less than /k + 1). This condition is a little weaker than 5), but it does not enlarge signiicantly the class o diusion processes. eerences 1. Bosq, D. and Davydov, Y.: Local time and density estimation in continuous time, Math. Meth. Statist. 81) 1999), 45.. Castellana, J. V. and Leadbetter, M..: On smoothed density estimation or stationary processes, Stoch. Proc. ppl ), Durett,.: Stochastic Calculus: Practical Introduction, CC Press, Boca aton, FL, 1996.

19 ESTIMTION OF THE DEIVTIVE Gill,. D. and Levit, B. Ya.: pplication o the van Trees inequality: a Bayesian Cramer-ao bound, Bernoulli ), Gikhman, I. I. and Skorokhod,. V.: Introduction to Theory o andom Processes, Saunders, Philadelphia, P, Golubev, G. K.: LN in problems o non-parametric estimation o unctions and lower bounds or quadratic risks, Theory Probab. ppl. 361) 1991), Golubev, G. K.: Non-parametric estimation o smooth densities in L, Problems Inorm. Transmission 81) 199), Golubev, G. K. and Levit, B. Ya.: On the second order minimax estimation o distribution unctions, Math. Meth. Statist. 51) 1996), Kutoyants, Yu..: Eicient density estimation or ergodic diusion processes, Statist. In. Stoch. Proc. 1) 1998), Leblanc, F.: Density estimation or a class o continuous time process, Math. Meth. Statist. 6) 1997), Lucas,.: Can we estimate the density s derivative with suroptimal rate? Statist. In. Stoch. Proc. 11) 1998), Nguyen, H. T.: Density estimation in a continuous-time Markov processes, nn. Statist ), Nussbaum, M.: Minimax risk: Pinsker s bound. In: S. Kotz ed.), Encyclopedia o Statistical Sciences, Vol. 3, Wiley, New York, 1999, pp Pinsker, M. S.: Optimal iltering o square integrable signals in Gaussian white noise, Problems Inorm. Transmission ), Schipper, M.: Optimal rates and constants in L -minimax estimation o probability density unctions, Math. Meth. Statist. 53) 1996),

Telescoping Decomposition Method for Solving First Order Nonlinear Differential Equations

Telescoping Decomposition Method for Solving First Order Nonlinear Differential Equations Telescoping Decomposition Method or Solving First Order Nonlinear Dierential Equations 1 Mohammed Al-Reai 2 Maysem Abu-Dalu 3 Ahmed Al-Rawashdeh Abstract The Telescoping Decomposition Method TDM is a new

More information

ENSC327 Communications Systems 2: Fourier Representations. School of Engineering Science Simon Fraser University

ENSC327 Communications Systems 2: Fourier Representations. School of Engineering Science Simon Fraser University ENSC37 Communications Systems : Fourier Representations School o Engineering Science Simon Fraser University Outline Chap..5: Signal Classiications Fourier Transorm Dirac Delta Function Unit Impulse Fourier

More information

Mathematical Institute, University of Utrecht. The problem of estimating the mean of an observed Gaussian innite-dimensional vector

Mathematical Institute, University of Utrecht. The problem of estimating the mean of an observed Gaussian innite-dimensional vector On Minimax Filtering over Ellipsoids Eduard N. Belitser and Boris Y. Levit Mathematical Institute, University of Utrecht Budapestlaan 6, 3584 CD Utrecht, The Netherlands The problem of estimating the mean

More information

1 Relative degree and local normal forms

1 Relative degree and local normal forms THE ZERO DYNAMICS OF A NONLINEAR SYSTEM 1 Relative degree and local normal orms The purpose o this Section is to show how single-input single-output nonlinear systems can be locally given, by means o a

More information

Feedback Linearization

Feedback Linearization Feedback Linearization Peter Al Hokayem and Eduardo Gallestey May 14, 2015 1 Introduction Consider a class o single-input-single-output (SISO) nonlinear systems o the orm ẋ = (x) + g(x)u (1) y = h(x) (2)

More information

Scattered Data Approximation of Noisy Data via Iterated Moving Least Squares

Scattered Data Approximation of Noisy Data via Iterated Moving Least Squares Scattered Data Approximation o Noisy Data via Iterated Moving Least Squares Gregory E. Fasshauer and Jack G. Zhang Abstract. In this paper we ocus on two methods or multivariate approximation problems

More information

The Asymptotic Minimax Constant for Sup-Norm. Loss in Nonparametric Density Estimation

The Asymptotic Minimax Constant for Sup-Norm. Loss in Nonparametric Density Estimation The Asymptotic Minimax Constant or Sup-Norm Loss in Nonparametric Density Estimation ALEXANDER KOROSTELEV 1 and MICHAEL NUSSBAUM 2 1 Department o Mathematics, Wayne State University, Detroit, MI 48202,

More information

On the Goodness-of-Fit Tests for Some Continuous Time Processes

On the Goodness-of-Fit Tests for Some Continuous Time Processes On the Goodness-of-Fit Tests for Some Continuous Time Processes Sergueï Dachian and Yury A. Kutoyants Laboratoire de Mathématiques, Université Blaise Pascal Laboratoire de Statistique et Processus, Université

More information

EXISTENCE OF SOLUTIONS TO SYSTEMS OF EQUATIONS MODELLING COMPRESSIBLE FLUID FLOW

EXISTENCE OF SOLUTIONS TO SYSTEMS OF EQUATIONS MODELLING COMPRESSIBLE FLUID FLOW Electronic Journal o Dierential Equations, Vol. 15 (15, No. 16, pp. 1 8. ISSN: 17-6691. URL: http://ejde.math.txstate.e or http://ejde.math.unt.e tp ejde.math.txstate.e EXISTENCE OF SOLUTIONS TO SYSTEMS

More information

RATE EXACT BAYESIAN ADAPTATION WITH MODIFIED BLOCK PRIORS. By Chao Gao and Harrison H. Zhou Yale University

RATE EXACT BAYESIAN ADAPTATION WITH MODIFIED BLOCK PRIORS. By Chao Gao and Harrison H. Zhou Yale University Submitted to the Annals o Statistics RATE EXACT BAYESIAN ADAPTATION WITH MODIFIED BLOCK PRIORS By Chao Gao and Harrison H. Zhou Yale University A novel block prior is proposed or adaptive Bayesian estimation.

More information

( x) f = where P and Q are polynomials.

( x) f = where P and Q are polynomials. 9.8 Graphing Rational Functions Lets begin with a deinition. Deinition: Rational Function A rational unction is a unction o the orm ( ) ( ) ( ) P where P and Q are polynomials. Q An eample o a simple rational

More information

SIO 211B, Rudnick. We start with a definition of the Fourier transform! ĝ f of a time series! ( )

SIO 211B, Rudnick. We start with a definition of the Fourier transform! ĝ f of a time series! ( ) SIO B, Rudnick! XVIII.Wavelets The goal o a wavelet transorm is a description o a time series that is both requency and time selective. The wavelet transorm can be contrasted with the well-known and very

More information

10. Joint Moments and Joint Characteristic Functions

10. Joint Moments and Joint Characteristic Functions 10. Joint Moments and Joint Characteristic Functions Following section 6, in this section we shall introduce various parameters to compactly represent the inormation contained in the joint p.d. o two r.vs.

More information

Outline. Approximate sampling theorem (AST) recall Lecture 1. P. L. Butzer, G. Schmeisser, R. L. Stens

Outline. Approximate sampling theorem (AST) recall Lecture 1. P. L. Butzer, G. Schmeisser, R. L. Stens Outline Basic relations valid or the Bernstein space B and their extensions to unctions rom larger spaces in terms o their distances rom B Part 3: Distance unctional approach o Part applied to undamental

More information

Fluctuationlessness Theorem and its Application to Boundary Value Problems of ODEs

Fluctuationlessness Theorem and its Application to Boundary Value Problems of ODEs Fluctuationlessness Theorem and its Application to Boundary Value Problems o ODEs NEJLA ALTAY İstanbul Technical University Inormatics Institute Maslak, 34469, İstanbul TÜRKİYE TURKEY) nejla@be.itu.edu.tr

More information

2.6 Two-dimensional continuous interpolation 3: Kriging - introduction to geostatistics. References - geostatistics. References geostatistics (cntd.

2.6 Two-dimensional continuous interpolation 3: Kriging - introduction to geostatistics. References - geostatistics. References geostatistics (cntd. .6 Two-dimensional continuous interpolation 3: Kriging - introduction to geostatistics Spline interpolation was originally developed or image processing. In GIS, it is mainly used in visualization o spatial

More information

Global Weak Solution of Planetary Geostrophic Equations with Inviscid Geostrophic Balance

Global Weak Solution of Planetary Geostrophic Equations with Inviscid Geostrophic Balance Global Weak Solution o Planetary Geostrophic Equations with Inviscid Geostrophic Balance Jian-Guo Liu 1, Roger Samelson 2, Cheng Wang 3 Communicated by R. Temam) Abstract. A reormulation o the planetary

More information

ELEG 3143 Probability & Stochastic Process Ch. 4 Multiple Random Variables

ELEG 3143 Probability & Stochastic Process Ch. 4 Multiple Random Variables Department o Electrical Engineering University o Arkansas ELEG 3143 Probability & Stochastic Process Ch. 4 Multiple Random Variables Dr. Jingxian Wu wuj@uark.edu OUTLINE 2 Two discrete random variables

More information

Goodness of fit test for ergodic diffusion processes

Goodness of fit test for ergodic diffusion processes Ann Inst Stat Math (29) 6:99 928 DOI.7/s463-7-62- Goodness of fit test for ergodic diffusion processes Ilia Negri Yoichi Nishiyama Received: 22 December 26 / Revised: July 27 / Published online: 2 January

More information

Partial Averaging of Fuzzy Differential Equations with Maxima

Partial Averaging of Fuzzy Differential Equations with Maxima Advances in Dynamical Systems and Applications ISSN 973-5321, Volume 6, Number 2, pp. 199 27 211 http://campus.mst.edu/adsa Partial Averaging o Fuzzy Dierential Equations with Maxima Olga Kichmarenko and

More information

Numerical Solution of Ordinary Differential Equations in Fluctuationlessness Theorem Perspective

Numerical Solution of Ordinary Differential Equations in Fluctuationlessness Theorem Perspective Numerical Solution o Ordinary Dierential Equations in Fluctuationlessness Theorem Perspective NEJLA ALTAY Bahçeşehir University Faculty o Arts and Sciences Beşiktaş, İstanbul TÜRKİYE TURKEY METİN DEMİRALP

More information

Introduction to Analog And Digital Communications

Introduction to Analog And Digital Communications Introduction to Analog And Digital Communications Second Edition Simon Haykin, Michael Moher Chapter Fourier Representation o Signals and Systems.1 The Fourier Transorm. Properties o the Fourier Transorm.3

More information

LIKELIHOOD RATIO INEQUALITIES WITH APPLICATIONS TO VARIOUS MIXTURES

LIKELIHOOD RATIO INEQUALITIES WITH APPLICATIONS TO VARIOUS MIXTURES Ann. I. H. Poincaré PR 38, 6 00) 897 906 00 Éditions scientiiques et médicales Elsevier SAS. All rights reserved S046-0030)05-/FLA LIKELIHOOD RATIO INEQUALITIES WITH APPLICATIONS TO VARIOUS MIXTURES Elisabeth

More information

PROBABILITY: LIMIT THEOREMS II, SPRING HOMEWORK PROBLEMS

PROBABILITY: LIMIT THEOREMS II, SPRING HOMEWORK PROBLEMS PROBABILITY: LIMIT THEOREMS II, SPRING 15. HOMEWORK PROBLEMS PROF. YURI BAKHTIN Instructions. You are allowed to work on solutions in groups, but you are required to write up solutions on your own. Please

More information

Theorem Let J and f be as in the previous theorem. Then for any w 0 Int(J), f(z) (z w 0 ) n+1

Theorem Let J and f be as in the previous theorem. Then for any w 0 Int(J), f(z) (z w 0 ) n+1 (w) Second, since lim z w z w z w δ. Thus, i r δ, then z w =r (w) z w = (w), there exist δ, M > 0 such that (w) z w M i dz ML({ z w = r}) = M2πr, which tends to 0 as r 0. This shows that g = 2πi(w), which

More information

BANDELET IMAGE APPROXIMATION AND COMPRESSION

BANDELET IMAGE APPROXIMATION AND COMPRESSION BANDELET IMAGE APPOXIMATION AND COMPESSION E. LE PENNEC AND S. MALLAT Abstract. Finding eicient geometric representations o images is a central issue to improve image compression and noise removal algorithms.

More information

Stolarsky Type Inequality for Sugeno Integrals on Fuzzy Convex Functions

Stolarsky Type Inequality for Sugeno Integrals on Fuzzy Convex Functions International Journal o Mathematical nalysis Vol., 27, no., 2-28 HIKRI Ltd, www.m-hikari.com https://doi.org/.2988/ijma.27.623 Stolarsky Type Inequality or Sugeno Integrals on Fuzzy Convex Functions Dug

More information

2 Frequency-Domain Analysis

2 Frequency-Domain Analysis 2 requency-domain Analysis Electrical engineers live in the two worlds, so to speak, o time and requency. requency-domain analysis is an extremely valuable tool to the communications engineer, more so

More information

Global Weak Solution of Planetary Geostrophic Equations with Inviscid Geostrophic Balance

Global Weak Solution of Planetary Geostrophic Equations with Inviscid Geostrophic Balance Global Weak Solution o Planetary Geostrophic Equations with Inviscid Geostrophic Balance Jian-Guo Liu 1, Roger Samelson 2, Cheng Wang 3 Communicated by R. Temam Abstract. A reormulation o the planetary

More information

Asymptotically Efficient Nonparametric Estimation of Nonlinear Spectral Functionals

Asymptotically Efficient Nonparametric Estimation of Nonlinear Spectral Functionals Acta Applicandae Mathematicae 78: 145 154, 2003. 2003 Kluwer Academic Publishers. Printed in the Netherlands. 145 Asymptotically Efficient Nonparametric Estimation of Nonlinear Spectral Functionals M.

More information

Mathematical Notation Math Calculus & Analytic Geometry III

Mathematical Notation Math Calculus & Analytic Geometry III Name : Mathematical Notation Math 221 - alculus & Analytic Geometry III Use Word or WordPerect to recreate the ollowing documents. Each article is worth 10 points and can e printed and given to the instructor

More information

GENERALIZED ABSTRACT NONSENSE: CATEGORY THEORY AND ADJUNCTIONS

GENERALIZED ABSTRACT NONSENSE: CATEGORY THEORY AND ADJUNCTIONS GENERALIZED ABSTRACT NONSENSE: CATEGORY THEORY AND ADJUNCTIONS CHRIS HENDERSON Abstract. This paper will move through the basics o category theory, eventually deining natural transormations and adjunctions

More information

STAT 801: Mathematical Statistics. Hypothesis Testing

STAT 801: Mathematical Statistics. Hypothesis Testing STAT 801: Mathematical Statistics Hypothesis Testing Hypothesis testing: a statistical problem where you must choose, on the basis o data X, between two alternatives. We ormalize this as the problem o

More information

On the Efficiency of Maximum-Likelihood Estimators of Misspecified Models

On the Efficiency of Maximum-Likelihood Estimators of Misspecified Models 217 25th European Signal Processing Conerence EUSIPCO On the Eiciency o Maximum-ikelihood Estimators o Misspeciied Models Mahamadou amine Diong Eric Chaumette and François Vincent University o Toulouse

More information

Universidad de Chile, Casilla 653, Santiago, Chile. Universidad de Santiago de Chile. Casilla 307 Correo 2, Santiago, Chile. May 10, 2017.

Universidad de Chile, Casilla 653, Santiago, Chile. Universidad de Santiago de Chile. Casilla 307 Correo 2, Santiago, Chile. May 10, 2017. A Laplace transorm approach to linear equations with ininitely many derivatives and zeta-nonlocal ield equations arxiv:1705.01525v2 [math-ph] 9 May 2017 A. Chávez 1, H. Prado 2, and E.G. Reyes 3 1 Departamento

More information

Solution of the Synthesis Problem in Hilbert Spaces

Solution of the Synthesis Problem in Hilbert Spaces Solution o the Synthesis Problem in Hilbert Spaces Valery I. Korobov, Grigory M. Sklyar, Vasily A. Skoryk Kharkov National University 4, sqr. Svoboda 677 Kharkov, Ukraine Szczecin University 5, str. Wielkopolska

More information

On Convexity of Reachable Sets for Nonlinear Control Systems

On Convexity of Reachable Sets for Nonlinear Control Systems Proceedings o the European Control Conerence 27 Kos, Greece, July 2-5, 27 WeC5.2 On Convexity o Reachable Sets or Nonlinear Control Systems Vadim Azhmyakov, Dietrich Flockerzi and Jörg Raisch Abstract

More information

Rigorous pointwise approximations for invariant densities of non-uniformly expanding maps

Rigorous pointwise approximations for invariant densities of non-uniformly expanding maps Ergod. Th. & Dynam. Sys. 5, 35, 8 44 c Cambridge University Press, 4 doi:.7/etds.3.9 Rigorous pointwise approximations or invariant densities o non-uniormly expanding maps WAEL BAHSOUN, CHRISTOPHER BOSE

More information

TLT-5200/5206 COMMUNICATION THEORY, Exercise 3, Fall TLT-5200/5206 COMMUNICATION THEORY, Exercise 3, Fall Problem 1.

TLT-5200/5206 COMMUNICATION THEORY, Exercise 3, Fall TLT-5200/5206 COMMUNICATION THEORY, Exercise 3, Fall Problem 1. TLT-5/56 COMMUNICATION THEORY, Exercise 3, Fall Problem. The "random walk" was modelled as a random sequence [ n] where W[i] are binary i.i.d. random variables with P[W[i] = s] = p (orward step with probability

More information

PROBABILITY: LIMIT THEOREMS II, SPRING HOMEWORK PROBLEMS

PROBABILITY: LIMIT THEOREMS II, SPRING HOMEWORK PROBLEMS PROBABILITY: LIMIT THEOREMS II, SPRING 218. HOMEWORK PROBLEMS PROF. YURI BAKHTIN Instructions. You are allowed to work on solutions in groups, but you are required to write up solutions on your own. Please

More information

NON-AUTONOMOUS INHOMOGENEOUS BOUNDARY CAUCHY PROBLEMS AND RETARDED EQUATIONS. M. Filali and M. Moussi

NON-AUTONOMOUS INHOMOGENEOUS BOUNDARY CAUCHY PROBLEMS AND RETARDED EQUATIONS. M. Filali and M. Moussi Electronic Journal: Southwest Journal o Pure and Applied Mathematics Internet: http://rattler.cameron.edu/swjpam.html ISSN 83-464 Issue 2, December, 23, pp. 26 35. Submitted: December 24, 22. Published:

More information

In many diverse fields physical data is collected or analysed as Fourier components.

In many diverse fields physical data is collected or analysed as Fourier components. 1. Fourier Methods In many diverse ields physical data is collected or analysed as Fourier components. In this section we briely discuss the mathematics o Fourier series and Fourier transorms. 1. Fourier

More information

International Journal of Mathematical Archive-8(6), 2017, 1-7 Available online through ISSN

International Journal of Mathematical Archive-8(6), 2017, 1-7 Available online through   ISSN nternational Journal o Mathematical Archive-8(6), 07, -7 Available online through www.ijma.ino SSN 9 5046 DETERMNATON OF ENTROPY FUNCTONAL FOR DHS DSTRBUTONS S. A. EL-SHEHAWY* Department o Mathematics,

More information

P i [B k ] = lim. n=1 p(n) ii <. n=1. V i :=

P i [B k ] = lim. n=1 p(n) ii <. n=1. V i := 2.7. Recurrence and transience Consider a Markov chain {X n : n N 0 } on state space E with transition matrix P. Definition 2.7.1. A state i E is called recurrent if P i [X n = i for infinitely many n]

More information

Minimax Risk: Pinsker Bound

Minimax Risk: Pinsker Bound Minimax Risk: Pinsker Bound Michael Nussbaum Cornell University From: Encyclopedia of Statistical Sciences, Update Volume (S. Kotz, Ed.), 1999. Wiley, New York. Abstract We give an account of the Pinsker

More information

A Simple Explanation of the Sobolev Gradient Method

A Simple Explanation of the Sobolev Gradient Method A Simple Explanation o the Sobolev Gradient Method R. J. Renka July 3, 2006 Abstract We have observed that the term Sobolev gradient is used more oten than it is understood. Also, the term is oten used

More information

Scattering of Solitons of Modified KdV Equation with Self-consistent Sources

Scattering of Solitons of Modified KdV Equation with Self-consistent Sources Commun. Theor. Phys. Beijing, China 49 8 pp. 89 84 c Chinese Physical Society Vol. 49, No. 4, April 5, 8 Scattering o Solitons o Modiied KdV Equation with Sel-consistent Sources ZHANG Da-Jun and WU Hua

More information

On Cramér-von Mises test based on local time of switching diffusion process

On Cramér-von Mises test based on local time of switching diffusion process On Cramér-von Mises test based on local time of switching diffusion process Anis Gassem Laboratoire de Statistique et Processus, Université du Maine, 7285 Le Mans Cedex 9, France e-mail: Anis.Gassem@univ-lemans.fr

More information

Additional exercises in Stationary Stochastic Processes

Additional exercises in Stationary Stochastic Processes Mathematical Statistics, Centre or Mathematical Sciences Lund University Additional exercises 8 * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *

More information

Received: 30 July 2017; Accepted: 29 September 2017; Published: 8 October 2017

Received: 30 July 2017; Accepted: 29 September 2017; Published: 8 October 2017 mathematics Article Least-Squares Solution o Linear Dierential Equations Daniele Mortari ID Aerospace Engineering, Texas A&M University, College Station, TX 77843, USA; mortari@tamu.edu; Tel.: +1-979-845-734

More information

THE SNAIL LEMMA ENRICO M. VITALE

THE SNAIL LEMMA ENRICO M. VITALE THE SNIL LEMM ENRICO M. VITLE STRCT. The classical snake lemma produces a six terms exact sequence starting rom a commutative square with one o the edge being a regular epimorphism. We establish a new

More information

Math Review and Lessons in Calculus

Math Review and Lessons in Calculus Math Review and Lessons in Calculus Agenda Rules o Eponents Functions Inverses Limits Calculus Rules o Eponents 0 Zero Eponent Rule a * b ab Product Rule * 3 5 a / b a-b Quotient Rule 5 / 3 -a / a Negative

More information

Fourier Series. 1. Review of Linear Algebra

Fourier Series. 1. Review of Linear Algebra Fourier Series In this section we give a short introduction to Fourier Analysis. If you are interested in Fourier analysis and would like to know more detail, I highly recommend the following book: Fourier

More information

Problem Set. Problems on Unordered Summation. Math 5323, Fall Februray 15, 2001 ANSWERS

Problem Set. Problems on Unordered Summation. Math 5323, Fall Februray 15, 2001 ANSWERS Problem Set Problems on Unordered Summation Math 5323, Fall 2001 Februray 15, 2001 ANSWERS i 1 Unordered Sums o Real Terms In calculus and real analysis, one deines the convergence o an ininite series

More information

The Poisson summation formula, the sampling theorem, and Dirac combs

The Poisson summation formula, the sampling theorem, and Dirac combs The Poisson summation ormula, the sampling theorem, and Dirac combs Jordan Bell jordanbell@gmailcom Department o Mathematics, University o Toronto April 3, 24 Poisson summation ormula et S be the set o

More information

1. Definition: Order Statistics of a sample.

1. Definition: Order Statistics of a sample. AMS570 Order Statistics 1. Deinition: Order Statistics o a sample. Let X1, X2,, be a random sample rom a population with p.d.. (x). Then, 2. p.d.. s or W.L.O.G.(W thout Loss o Ge er l ty), let s ssu e

More information

Mathematical Notation Math Calculus & Analytic Geometry III

Mathematical Notation Math Calculus & Analytic Geometry III Mathematical Notation Math 221 - alculus & Analytic Geometry III Use Word or WordPerect to recreate the ollowing documents. Each article is worth 10 points and should be emailed to the instructor at james@richland.edu.

More information

A tail inequality for suprema of unbounded empirical processes with applications to Markov chains

A tail inequality for suprema of unbounded empirical processes with applications to Markov chains E l e c t r o n i c J o u r n a l o P r o b a b i l i t y Vol. 13 2008, Paper no. 34, pages 1000 1034. Journal URL http://www.math.washington.edu/~ejpecp/ A tail inequality or suprema o unbounded empirical

More information

Journal of Mathematical Analysis and Applications

Journal of Mathematical Analysis and Applications J. Math. Anal. Appl. 377 20 44 449 Contents lists available at ScienceDirect Journal o Mathematical Analysis and Applications www.elsevier.com/locate/jmaa Value sharing results or shits o meromorphic unctions

More information

Discrete Mathematics. On the number of graphs with a given endomorphism monoid

Discrete Mathematics. On the number of graphs with a given endomorphism monoid Discrete Mathematics 30 00 376 384 Contents lists available at ScienceDirect Discrete Mathematics journal homepage: www.elsevier.com/locate/disc On the number o graphs with a given endomorphism monoid

More information

A note on functional inequalities for some Lévy processes

A note on functional inequalities for some Lévy processes A note on unctional inequalities or some Lévy processes D Chaaï & F Malrieu University o Toulouse March 00, compiled September 7, 00 Abstract By using a Γ approach like in [AL00], we establish a modiied

More information

Journal of Mathematical Analysis and Applications

Journal of Mathematical Analysis and Applications J. Math. Anal. Appl. 352 2009) 739 748 Contents lists available at ScienceDirect Journal o Mathematical Analysis Applications www.elsevier.com/locate/jmaa The growth, oscillation ixed points o solutions

More information

On a Closed Formula for the Derivatives of e f(x) and Related Financial Applications

On a Closed Formula for the Derivatives of e f(x) and Related Financial Applications International Mathematical Forum, 4, 9, no. 9, 41-47 On a Closed Formula or the Derivatives o e x) and Related Financial Applications Konstantinos Draais 1 UCD CASL, University College Dublin, Ireland

More information

Optimal robust estimates using the Hellinger distance

Optimal robust estimates using the Hellinger distance Adv Data Anal Classi DOI 10.1007/s11634-010-0061-8 REGULAR ARTICLE Optimal robust estimates using the Hellinger distance Alio Marazzi Victor J. Yohai Received: 23 November 2009 / Revised: 25 February 2010

More information

Roberto s Notes on Differential Calculus Chapter 8: Graphical analysis Section 1. Extreme points

Roberto s Notes on Differential Calculus Chapter 8: Graphical analysis Section 1. Extreme points Roberto s Notes on Dierential Calculus Chapter 8: Graphical analysis Section 1 Extreme points What you need to know already: How to solve basic algebraic and trigonometric equations. All basic techniques

More information

Figure 3.1 Effect on frequency spectrum of increasing period T 0. Consider the amplitude spectrum of a periodic waveform as shown in Figure 3.2.

Figure 3.1 Effect on frequency spectrum of increasing period T 0. Consider the amplitude spectrum of a periodic waveform as shown in Figure 3.2. 3. Fourier ransorm From Fourier Series to Fourier ransorm [, 2] In communication systems, we oten deal with non-periodic signals. An extension o the time-requency relationship to a non-periodic signal

More information

In the index (pages ), reduce all page numbers by 2.

In the index (pages ), reduce all page numbers by 2. Errata or Nilpotence and periodicity in stable homotopy theory (Annals O Mathematics Study No. 28, Princeton University Press, 992) by Douglas C. Ravenel, July 2, 997, edition. Most o these were ound by

More information

Root Arrangements of Hyperbolic Polynomial-like Functions

Root Arrangements of Hyperbolic Polynomial-like Functions Root Arrangements o Hyperbolic Polynomial-like Functions Vladimir Petrov KOSTOV Université de Nice Laboratoire de Mathématiques Parc Valrose 06108 Nice Cedex France kostov@mathunicer Received: March, 005

More information

Supplementary material for Continuous-action planning for discounted infinite-horizon nonlinear optimal control with Lipschitz values

Supplementary material for Continuous-action planning for discounted infinite-horizon nonlinear optimal control with Lipschitz values Supplementary material or Continuous-action planning or discounted ininite-horizon nonlinear optimal control with Lipschitz values List o main notations x, X, u, U state, state space, action, action space,

More information

Lower Tail Probabilities and Related Problems

Lower Tail Probabilities and Related Problems Lower Tail Probabilities and Related Problems Qi-Man Shao National University of Singapore and University of Oregon qmshao@darkwing.uoregon.edu . Lower Tail Probabilities Let {X t, t T } be a real valued

More information

Comptes rendus de l Academie bulgare des Sciences, Tome 59, 4, 2006, p POSITIVE DEFINITE RANDOM MATRICES. Evelina Veleva

Comptes rendus de l Academie bulgare des Sciences, Tome 59, 4, 2006, p POSITIVE DEFINITE RANDOM MATRICES. Evelina Veleva Comtes rendus de l Academie bulgare des ciences Tome 59 4 6 353 36 POITIVE DEFINITE RANDOM MATRICE Evelina Veleva Abstract: The aer begins with necessary and suicient conditions or ositive deiniteness

More information

Strong Lyapunov Functions for Systems Satisfying the Conditions of La Salle

Strong Lyapunov Functions for Systems Satisfying the Conditions of La Salle 06 IEEE TRANSACTIONS ON AUTOMATIC CONTROL, VOL. 49, NO. 6, JUNE 004 Strong Lyapunov Functions or Systems Satisying the Conditions o La Salle Frédéric Mazenc and Dragan Ne sić Abstract We present a construction

More information

Chapter 6 Reliability-based design and code developments

Chapter 6 Reliability-based design and code developments Chapter 6 Reliability-based design and code developments 6. General Reliability technology has become a powerul tool or the design engineer and is widely employed in practice. Structural reliability analysis

More information

Mixed Signal IC Design Notes set 6: Mathematics of Electrical Noise

Mixed Signal IC Design Notes set 6: Mathematics of Electrical Noise ECE45C /8C notes, M. odwell, copyrighted 007 Mied Signal IC Design Notes set 6: Mathematics o Electrical Noise Mark odwell University o Caliornia, Santa Barbara rodwell@ece.ucsb.edu 805-893-344, 805-893-36

More information

2. ETA EVALUATIONS USING WEBER FUNCTIONS. Introduction

2. ETA EVALUATIONS USING WEBER FUNCTIONS. Introduction . ETA EVALUATIONS USING WEBER FUNCTIONS Introduction So ar we have seen some o the methods or providing eta evaluations that appear in the literature and we have seen some o the interesting properties

More information

Ruelle Operator for Continuous Potentials and DLR-Gibbs Measures

Ruelle Operator for Continuous Potentials and DLR-Gibbs Measures Ruelle Operator or Continuous Potentials and DLR-Gibbs Measures arxiv:1608.03881v4 [math.ds] 22 Apr 2018 Leandro Cioletti Departamento de Matemática - UnB 70910-900, Brasília, Brazil cioletti@mat.unb.br

More information

EXISTENCE OF ISOPERIMETRIC SETS WITH DENSITIES CONVERGING FROM BELOW ON R N. 1. Introduction

EXISTENCE OF ISOPERIMETRIC SETS WITH DENSITIES CONVERGING FROM BELOW ON R N. 1. Introduction EXISTECE OF ISOPERIMETRIC SETS WITH DESITIES COVERGIG FROM BELOW O R GUIDO DE PHILIPPIS, GIOVAI FRAZIA, AD ALDO PRATELLI Abstract. In this paper, we consider the isoperimetric problem in the space R with

More information

arxiv: v2 [math.co] 29 Mar 2017

arxiv: v2 [math.co] 29 Mar 2017 COMBINATORIAL IDENTITIES FOR GENERALIZED STIRLING NUMBERS EXPANDING -FACTORIAL FUNCTIONS AND THE -HARMONIC NUMBERS MAXIE D. SCHMIDT arxiv:6.04708v2 [math.co 29 Mar 207 SCHOOL OF MATHEMATICS GEORGIA INSTITUTE

More information

Optimal Control. with. Aerospace Applications. James M. Longuski. Jose J. Guzman. John E. Prussing

Optimal Control. with. Aerospace Applications. James M. Longuski. Jose J. Guzman. John E. Prussing Optimal Control with Aerospace Applications by James M. Longuski Jose J. Guzman John E. Prussing Published jointly by Microcosm Press and Springer 2014 Copyright Springer Science+Business Media New York

More information

Outline of Fourier Series: Math 201B

Outline of Fourier Series: Math 201B Outline of Fourier Series: Math 201B February 24, 2011 1 Functions and convolutions 1.1 Periodic functions Periodic functions. Let = R/(2πZ) denote the circle, or onedimensional torus. A function f : C

More information

Analytic continuation in several complex variables

Analytic continuation in several complex variables Analytic continuation in several complex variables An M.S. Thesis Submitted, in partial ulillment o the requirements or the award o the degree o Master o Science in the Faculty o Science, by Chandan Biswas

More information

THE GAMMA FUNCTION THU NGỌC DƯƠNG

THE GAMMA FUNCTION THU NGỌC DƯƠNG THE GAMMA FUNCTION THU NGỌC DƯƠNG The Gamma unction was discovered during the search or a actorial analog deined on real numbers. This paper will explore the properties o the actorial unction and use them

More information

Products and Convolutions of Gaussian Probability Density Functions

Products and Convolutions of Gaussian Probability Density Functions Tina Memo No. 003-003 Internal Report Products and Convolutions o Gaussian Probability Density Functions P.A. Bromiley Last updated / 9 / 03 Imaging Science and Biomedical Engineering Division, Medical

More information

ELEMENTS OF PROBABILITY THEORY

ELEMENTS OF PROBABILITY THEORY ELEMENTS OF PROBABILITY THEORY Elements of Probability Theory A collection of subsets of a set Ω is called a σ algebra if it contains Ω and is closed under the operations of taking complements and countable

More information

EEO 401 Digital Signal Processing Prof. Mark Fowler

EEO 401 Digital Signal Processing Prof. Mark Fowler EEO 401 Digital Signal Processing Pro. Mark Fowler Note Set #10 Fourier Analysis or DT Signals eading Assignment: Sect. 4.2 & 4.4 o Proakis & Manolakis Much o Ch. 4 should be review so you are expected

More information

Maximum Likelihood Drift Estimation for Gaussian Process with Stationary Increments

Maximum Likelihood Drift Estimation for Gaussian Process with Stationary Increments Austrian Journal of Statistics April 27, Volume 46, 67 78. AJS http://www.ajs.or.at/ doi:.773/ajs.v46i3-4.672 Maximum Likelihood Drift Estimation for Gaussian Process with Stationary Increments Yuliya

More information

Classification of effective GKM graphs with combinatorial type K 4

Classification of effective GKM graphs with combinatorial type K 4 Classiication o eective GKM graphs with combinatorial type K 4 Shintarô Kuroki Department o Applied Mathematics, Faculty o Science, Okayama Uniervsity o Science, 1-1 Ridai-cho Kita-ku, Okayama 700-0005,

More information

Probability and Measure

Probability and Measure Part II Year 2018 2017 2016 2015 2014 2013 2012 2011 2010 2009 2008 2007 2006 2005 2018 84 Paper 4, Section II 26J Let (X, A) be a measurable space. Let T : X X be a measurable map, and µ a probability

More information

y2 = 0. Show that u = e2xsin(2y) satisfies Laplace's equation.

y2 = 0. Show that u = e2xsin(2y) satisfies Laplace's equation. Review 1 1) State the largest possible domain o deinition or the unction (, ) = 3 - ) Determine the largest set o points in the -plane on which (, ) = sin-1( - ) deines a continuous unction 3) Find the

More information

CISE-301: Numerical Methods Topic 1:

CISE-301: Numerical Methods Topic 1: CISE-3: Numerical Methods Topic : Introduction to Numerical Methods and Taylor Series Lectures -4: KFUPM Term 9 Section 8 CISE3_Topic KFUPM - T9 - Section 8 Lecture Introduction to Numerical Methods What

More information

MATH 205C: STATIONARY PHASE LEMMA

MATH 205C: STATIONARY PHASE LEMMA MATH 205C: STATIONARY PHASE LEMMA For ω, consider an integral of the form I(ω) = e iωf(x) u(x) dx, where u Cc (R n ) complex valued, with support in a compact set K, and f C (R n ) real valued. Thus, I(ω)

More information

Logarithm of a Function, a Well-Posed Inverse Problem

Logarithm of a Function, a Well-Posed Inverse Problem American Journal o Computational Mathematics, 4, 4, -5 Published Online February 4 (http://www.scirp.org/journal/ajcm http://dx.doi.org/.436/ajcm.4.4 Logarithm o a Function, a Well-Posed Inverse Problem

More information

Lower Tail Probabilities and Normal Comparison Inequalities. In Memory of Wenbo V. Li s Contributions

Lower Tail Probabilities and Normal Comparison Inequalities. In Memory of Wenbo V. Li s Contributions Lower Tail Probabilities and Normal Comparison Inequalities In Memory of Wenbo V. Li s Contributions Qi-Man Shao The Chinese University of Hong Kong Lower Tail Probabilities and Normal Comparison Inequalities

More information

Entropy jumps in the presence of a spectral gap

Entropy jumps in the presence of a spectral gap Entropy jumps in the presence o a spectral gap Keith Ball, Franck Barthe and Assa Naor September 4, 00 Abstract It is shown that i X is a random variable whose density satisies a Poincaré inequality, and

More information

Basic mathematics of economic models. 3. Maximization

Basic mathematics of economic models. 3. Maximization John Riley 1 January 16 Basic mathematics o economic models 3 Maimization 31 Single variable maimization 1 3 Multi variable maimization 6 33 Concave unctions 9 34 Maimization with non-negativity constraints

More information

Laplace s Equation. Chapter Mean Value Formulas

Laplace s Equation. Chapter Mean Value Formulas Chapter 1 Laplace s Equation Let be an open set in R n. A function u C 2 () is called harmonic in if it satisfies Laplace s equation n (1.1) u := D ii u = 0 in. i=1 A function u C 2 () is called subharmonic

More information

YURI LEVIN AND ADI BEN-ISRAEL

YURI LEVIN AND ADI BEN-ISRAEL Pp. 1447-1457 in Progress in Analysis, Vol. Heinrich G W Begehr. Robert P Gilbert and Man Wah Wong, Editors, World Scientiic, Singapore, 003, ISBN 981-38-967-9 AN INVERSE-FREE DIRECTIONAL NEWTON METHOD

More information

Finite-dimensional spaces. C n is the space of n-tuples x = (x 1,..., x n ) of complex numbers. It is a Hilbert space with the inner product

Finite-dimensional spaces. C n is the space of n-tuples x = (x 1,..., x n ) of complex numbers. It is a Hilbert space with the inner product Chapter 4 Hilbert Spaces 4.1 Inner Product Spaces Inner Product Space. A complex vector space E is called an inner product space (or a pre-hilbert space, or a unitary space) if there is a mapping (, )

More information

Conditional distribution approximation and Stein s method

Conditional distribution approximation and Stein s method Conditional distribution approximation and Stein s method Han Liang Gan Submitted in total ulilment o the requirements o the degree o Doctor o Philosophy July 214 Department o Mathematics and Statistics

More information

Functional Limit theorems for the quadratic variation of a continuous time random walk and for certain stochastic integrals

Functional Limit theorems for the quadratic variation of a continuous time random walk and for certain stochastic integrals Functional Limit theorems for the quadratic variation of a continuous time random walk and for certain stochastic integrals Noèlia Viles Cuadros BCAM- Basque Center of Applied Mathematics with Prof. Enrico

More information