brain

**Statement.

Let $M(n) = \operatorname{rank}(T(n,n,n))$ denote the rank of the $n \times n$ matrix multiplication tensor.

We restart from the correct structural facts about matrix multiplication tensors and tensor rank, and avoid any assumptions about multiplicativity of rank beyond what is valid: subadditivity under dec...

Let A=(a_{ij})_{1\le i,j\le 3}, \qquad B=(b_{ij})_{1\le i,j\le 3}, and let

Let $V$ be a $2$-dimensional vector space over a field $\mathbb{F}$ with basis ${e_1, e_2}$.

We first restate the structure in a precise way consistent with the exercise.

Let $T(m,n,s)$ denote the trilinear tensor corresponding to the $(m \times n)$ times $(n \times s)$ matrix multiplication problem, defined by $t_{(i,j')(j,k)(i,k)} = 1 \iff i' = i,\, j' = j,\, k' = k,...

The original solution fails because it incorrectly attributes a Karatsuba-style recurrence $M(n)=3M(n/2)$ to the algorithm, which destroys the required bound on multiplications.

We consider Exercise 4.

Let $u(x)=\sum_{i=0}^{n} a_i x^i,\qquad y(x)=\sum_{j=0}^{n} b_j x^j.$ Their product is $z(x)=u(x)y(x)=\sum_{k=0}^{2n} c_k x^k,$ where $c_k=\sum_{i+j=k} a_i b_j.$

**Exercise 4.

In §4.

Let $P$ be an arbitrary $n \times n$ matrix, and consider the tensor defined in equation (74), which is the $n \times n \times n$ tensor $T = \bigl(t_{ijk}\bigr) \quad \text{with} \quad t_{ijk} = \del...

Unusual activity has been detected from your device.

Let $n = n'n''$ with $\gcd(n', n'') = 1$.

We redo the construction cleanly and explicitly, giving full Winograd decompositions and verifying correctness.

The key error in the previous solution is the incorrect step that an arbitrary rank-one term in the flattened matrix decomposition corresponds, after reshaping, to a rank-one $m\times n$ matrix.

Let $V$ be the space of $m\times n$ matrices and let $W$ be the space of $n\times1$ column vectors.

Let $V=F^{m}\otimes F^{n}\otimes F^{s}$, the vector space of all $m\times n\times s$ tensors over a field $F$.

Let ${t_{ijk}}$ be an $m \times n \times s$ tensor of rank $r = \text{rank}(t_{ijk})$, and let ${t'_{ijk}}$ be an $m' \times n' \times s'$ tensor of rank $r' = \text{rank}(t'_{ijk})$.

Let u(x) = x^n + u_{n-1} x^{n-1} + \cdots + u_1 x + u_0 be a **monic polynomial** of degree $n$, with coefficients $u_{n-1}, \dots, u_0$.

Let $\rho(T)$ denote the rank of a tensor $T={t_{ijk}}$, where rank means the least integer $r$ for which t_{ijk}=\sum_{\nu=1}^{r} a_{i\nu} b_{j\nu} c_{k\nu}.

Let the two bilinear forms be z_1(x,y)=x^\top A y,\qquad z_2(x,y)=x^\top B y, where $A,B\in \mathbb{F}^{2\times 2}$.

Let $u(x)$ be a polynomial of degree $n$ over the integers.

Let S_n(x)=1+x+x^2+\cdots+x^n.

Let $M(n)$ denote the minimum number of multiplications needed to evaluate some polynomial of degree $n$, with arbitrary coefficients, when no preliminary adaptation of the coefficients is allowed.

The previous solution fails because it tries to control Euclidean remainders and ignores the integrality constraints.

We wish to compute the real and imaginary parts of the product of two complex numbers $(a + bi)(c + di)$ using only three real multiplications and five real additions, with two of the additions involv...

Let $R(x)=\frac{x^2+10x+29}{x^2+8x+19}.$ Since numerator and denominator have the same degree, divide polynomials: (x^2+10x+29)-(x^2+8x+19)=2x+10.

Let P(x;w_0,\ldots ,w_n) = \sum_{i=0}^{m} \Bigl(a_{i0}w_0+\cdots +a_{in}w_n+b_i\Bigr)x^i,

Exercise 35 established that a general fourth-degree polynomial cannot be computed with three multiplications and fewer than five addition-subtractions.

Assume that a polynomial chain computes a general fourth-degree polynomial with three multiplications and four addition-subtractions.

Let P(x) = u_1 x^3 + u_2 x^2 + u_0, where $u_0, u_1, u_2$ are independent parameters.

Let $\lambda_0, \lambda_1, \ldots, \lambda_r$ be a polynomial chain in which all addition and subtraction steps are **parameter steps**, and suppose there is at least one **parameter multiplication**.

The previous solution fails because it never defines a correct model of computation and therefore cannot justify any dimension or “independent parameter” count.

A polynomial chain computes expressions from the variable $x$ and parameters using additions, subtractions, and multiplications.

Let a **polynomial chain** be defined as in Section 4.

Let $f_0, \ldots, f_r$ be multivariate polynomials with integer coefficients in the variables $\alpha_1, \ldots, \alpha_s$.

Let $R_1, \dots, R_m \subset \mathbb{R}^{n+1}$, and assume each $R_i$ has at most $t$ degrees of freedom.

Let $R$ be the set of all $(n+1)$-tuples $(q_n,\ldots,q_0)$ of real numbers with $q_n \ne 0$.

The construction in Theorem M is a straight-line program for a polynomial, and the associated coefficients $\beta_i$ are obtained by the reverse propagation rule for the final value $\lambda_{10}$.

Let u(x) = u_3 x^3 + u_2 x^2 + u_1 x + u_0 be a cubic polynomial with real coefficients.

Restart from the goal: construct a Pan-style evaluation scheme (16), meaning a straight-line program that minimizes multiplications by first generating needed powers of $x$ via a short addition chain,...

We restart from the structural requirement of Theorem E.

Let f(z) = a_n z^n + a_{n-1} z^{n-1} + \cdots + a_0 be a polynomial of degree $n$ with real coefficients, having at least $n-1$ roots with nonnegative real part.

We are asked to write a MIX program that evaluates a fifth-degree polynomial according to scheme (11) in Section 4.

Solution to TAOCP 4.6.4 Exercise 21.

The proof of Ryser's identity is correct and complete.

Let the scheme (11) represent the nested evaluation form for a fifth-degree polynomial $u(x)=u_5x^5+u_4x^4+u_3x^3+u_2x^2+u_1x+u_0,$ rewritten in adapted Horner form with coefficients $a_0,\ldots,a_5$...

Let $x_0, x_1, \ldots, x_n$ be distinct real numbers.

Yes, the inequality $\nu(n) \le 2^{l(n) - \lambda(n)}$ holds for all positive integers $n$.

We are asked to show that the interpolation formula (45) reduces to a simple expression involving binomial coefficients when the nodes are in an arithmetic progression, namely $x_k = x_0 + kh, \qquad...

Let $N = 2^n$ and write $\omega = e^{2\pi i/N}$.

**Solution to Exercise 4.

Let $A,B \in F^{n\times n}$ and let $C=AB$.

Let $X = (x_{ij})$ be an $n \times n$ matrix.

Let $X = (x_{ij})$ be an $n \times n$ matrix.

Let R(X)=(-1)^n\sum_{\epsilon\in\{0,1\}^n} (-1)^{\epsilon_1+\cdots+\epsilon_n} \prod_{i=1}^n\sum_{j=1}^n \epsilon_jx_{ij}.

**Exercise 4.

We are asked to improve steps S1, .

The clean way to remove the confusion is to derive the evaluation recurrence directly from the structure of the falling factorial basis, and then count operations in a single unified loop.

Let $u(z) = u_n z^n + u_{n-1} z^{n-1} + \cdots + u_1 z + u_0$ be a polynomial of degree $n$, where each coefficient $u_k$ is complex and $z = x + iy$ is a complex variable.

Let $n$ be given and write $u(x)=u_n x^n+u_{n-1}x^{n-1}+\cdots+u_1x+u_0.$ Define the even and odd parts with respect to $x^2$: $E(x)=\sum_{k\ge 0} u_{2k} x^{2k}, \qquad O(x)=\sum_{k\ge 0} u_{2k+1} x^{...

Let $u(x,y)=\sum_{i+j\le n} u_{ij} x^i y^j$ be a bivariate polynomial of total degree $n$.

Let $u(x) = u_n x^n + u_{n-1} x^{n-1} + \cdots + u_1 x + u_0$ be a polynomial with coefficients in a ring $\mathcal{R}$, and suppose we wish to evaluate $u(x)$ when $x$ itself is a polynomial over $\m...

Let the chains in Exercise 34 be written in the standard form determined by the exponents \[ e_0>e_1>\cdots, \] and recall that two addition chains are regarded as equivalent when they have the same p...

We are asked to consider two addition chains for an integer n = 2^{e_0} + 2^{e_1} + \cdots + 2^{e_t}, \quad e_0 > e_1 > \cdots > e_t \ge 0, and to determine whether the **S-and-X chain** and the **Alg...

Let $u(x)=u_{2n+1}x^{2n+1}+u_{2n-1}x^{2n-1}+\cdots+u_1x.$ Factor out $x$: $u(x)=x\left(u_{2n+1}x^{2n}+u_{2n-1}x^{2n-2}+\cdots+u_1\right).$ Introduce the substitution $y=x^2$.

We are asked: > How many addition chains of length $9$ have (52) as their reduced directed graph (RDG)?

Let the addition chain be 1=a_1<a_2<\cdots<a_m=n,\qquad a_i=a_j+a_k\ (j,k<i), and assign cost $a_j a_k$ to step $a_i=a_j+a_k$.

We are asked to **explore the problem of minimizing** f = cq + (r - q) for an addition chain

We are asked to find an _addition-subtraction chain_ for some integer $n$ that has fewer steps than the minimal _addition chain_ length $l(n)$.

Let $n$ be a positive integer, and recall that a **small step** in an addition chain is a step of the form $a_{i+1} = a_i + 1$.

Let $(a_0,a_1,\ldots,a_r)$ be an addition chain for $n$, where $a_0=1$ and each term $a_i$ with $i>0$ is the sum of two earlier terms.

Yes, the inequality $\nu(n) \le 2^{l(n) - \lambda(n)}$ holds for all positive integers $n$.

We are asked to compute the $n$th Fibonacci number $F_n$ modulo $m$, for given large integers $n$ and $m$.

Let $y = (.d_1 d_2 \ldots d_k)_2$ be a binary fraction, where $0 < y < 1$ and each $d_j \in {0,1}$.

Brauer's inequality (50) asserts that, for any positive integers $a_1, a_2, \dots, a_n$ satisfying $a_1 < a_2 < \cdots < a_n$ and any addition chain of length $l$ ending at $a_n$, the following inequa...

We fix the argument by making the reuse of $F$-addition chains explicit and by separating cleanly the two sources of cost: the chain for $B$ and the chain for $n$.

Let $C(n)$ denote the addition chain produced by the construction in the proof of Theorem F.

Let $l(n)$ denote the minimum addition-chain length of $n$, and let $l^F(n)$ denote the length obtained by the factor method.

The reviewer's objections are fatal.

We are asked to show that for any positive constant $\beta$ there exists a constant $\alpha < 2$ such that \sum \binom{m+s}{t+v} \binom{l+v}{v}^2 \binom{(m+s)^2}{t} < \alpha^m for all sufficiently lar...

Let the multiplicity of an element $x$ in a multiset $A$ be denoted by $m_A(x)$.

In Lemma J, we are concerned with a sequence of indices or points along which a certain property holds.

Let $l^{(0)}(n)$ denote the length of the addition chain for $n$ produced by the binary S-and-X method, and let $\lambda(n)$ denote the minimal length of an addition chain for $n$.

Let s(n)=l(n)-\lambda(n),\qquad s^*(n)=l^*(n)-\lambda(n), where $\lambda(n)=\lfloor \log_2 n\rfloor$.

The proposed solution does **not** answer the exercise that was asked.

**Solution to Exercise 4.

The reviewer is correct.

Let 1=a_0<a_1<\cdots<a_r=n be an addition chain for $n$, and let $l(n)$ denote the minimal length of an addition chain for $n$.

Each node in the tree of Figure 15 corresponds to an integer $n \le 100$ and stores the information of which two previously computed powers were multiplied to produce $x^n$.

Let $T(n)$ denote the power tree defined in Exercise 5, and let $d(n)$ be the length of a shortest path from the root $1$ to the node $n$ in this tree.

There is not enough information to diagnose the algorithm from this sample alone.

We consider each part separately.

**Exercise 4.

**Exercise 4.

We are asked to construct the first $r+1$ levels of the "power tree" as defined in Figure 14 and in Exercise 4.

We are asked to compute $2^{375}$ by various exponentiation methods.

Let u(x)=u_nx^n+\cdots+u_0 be reducible over the integers, and let