这是一篇民科的直接从生成函数角度进行推导的题解。
首先，根据 Prufer 序列，易知相当于计数长度为 \(n-2\) 的，元素在 \([1,n]\) 内且所有元素出现次数的最大值恰为 \(m-1\) 的序列个数。
但是最大值恰为 \(m-1\) 并不好求，考虑差分转化为出现次数全都不超过 \(m-1\) 的，减去出现次数全都不超过 \(m-2\) 的。
于是考虑如何计数长度为 \(n-2\)，元素出现次数不超过 \(m-1\) 的序列个数。
序列当然是用 EGF，因为要考虑顺序。由于每种元素的出现次数都不能超过 \(m-1\)，所以可以考虑对每种元素构造 EGF。
则设 \[F(x) = \sum\limits_{i=0}^{m-1} \frac{x^i}{i!}\]
因为同元素之间换顺序也只算一种方案。
由于有 \([1,n]\) 的元素，所以答案就是 \[(n-2)![x^{n-2}]F^n(x)\] 多项式倍增快速幂或者 ln exp 快速幂解决。
复杂度 \(O(n \log^2 n)\) 或 \(O(n \log n)\)。
代码： 1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
using namespace std;
const int N = 1 << 18;
const int mod = 998244353;
const int G = 3;
inline int fpow(int a,int b)
{
int ret = 1;
for(;b;b >>= 1)
(b & 1) && (ret = (long long)ret * a % mod),a = (long long)a * a % mod;
return ret;
}
struct poly
{
int a[N + 5];
inline const int &operator[](int x) const
{
return a[x];
}
inline int &operator[](int x)
{
return a[x];
}
inline void clear(int x = 0)
{
memset(a + x,0,(N - x + 1) << 2);
}
};
int m,len,k,n,lg2[N + 5];
int cnt[N + 5];
int rev[N + 5],fac[N + 5],ifac[N + 5],inv[N + 5];
int rt[N + 5],irt[N + 5];
inline void init(int len)
{
for(n = 1;n < len;n <<= 1);
for(register int i = 2;i <= n;++i)
lg2[i] = lg2[i >> 1] + 1;
int w = fpow(G,(mod - 1) / n);
rt[n >> 1] = 1;
for(register int i = (n >> 1) + 1;i <= n;++i)
rt[i] = (long long)rt[i - 1] * w % mod;
for(register int i = (n >> 1) - 1;i;--i)
rt[i] = rt[i << 1];
fac[0] = 1;
for(register int i = 1;i <= n;++i)
fac[i] = (long long)fac[i - 1] * i % mod;
ifac[n] = fpow(fac[n],mod - 2);
for(register int i = n;i;--i)
ifac[i - 1] = (long long)ifac[i] * i % mod;
for(register int i = 1;i <= n;++i)
inv[i] = (long long)ifac[i] * fac[i - 1] % mod;
}
inline void ntt(poly &a,int type,int n)
{
type == -1 && (reverse(a.a + 1,a.a + n),1);
int lg = lg2[n] - 1;
for(register int i = 0;i < n;++i)
rev[i] = (rev[i >> 1] >> 1) | ((i & 1) << lg),
i < rev[i] && (swap(a[i],a[rev[i]]),1);
for(register int w = 2,m = 1;w <= n;w <<= 1,m <<= 1)
for(register int i = 0;i < n;i += w)
for(register int j = 0;j < m;++j)
{
int t = (long long)rt[m | j] * a[i | j | m] % mod;
a[i | j | m] = dec(a[i | j],t),a[i | j] = add(a[i | j],t);
}
if(type == -1)
for(register int i = 0;i < n;++i)
a[i] = (long long)a[i] * inv[n] % mod;
}
inline void mul(poly &a,const poly &b,int n)
{
static poly x,y;
int lim = 1;
x.clear(),y.clear();
for(;lim < (n << 1);lim <<= 1);
x = a,y = b;
x.clear(n),y.clear(n);
ntt(x,1,lim),ntt(y,1,lim);
for(register int i = 0;i < lim;++i)
x[i] = (long long)x[i] * y[i] % mod;
ntt(x,-1,lim);
x.clear(n),a = x;
}
inline poly inverse(const poly &f,int n)
{
static int s[30];
static poly g,h,q;
int lim = 1,top = 0;
g.clear();
for(;n > 1;s[++top] = n,n = (n + 1) >> 1);
g[0] = fpow(f[0],mod - 2);
for(;top;--top)
{
n = s[top];
for(;lim < (n << 1);lim <<= 1);
q = g,h = f,h.clear(n);
ntt(g,1,lim),ntt(h,1,lim);
for(register int i = 0;i < lim;++i)
g[i] = (long long)g[i] * g[i] % mod * h[i] % mod;
ntt(g,-1,lim);
for(register int i = 0;i < n;++i)
g[i] = dec(add(q[i],q[i]),g[i]);
g.clear(n);
}
return g;
}
inline void derivative(poly &f,int n)
{
for(register int i = 1;i < n;++i)
f[i - 1] = (long long)f[i] * i % mod;
f[n - 1] = 0;
}
inline void integral(poly &f,int n)
{
for(register int i = n - 1;~i;--i)
f[i + 1] = (long long)f[i] * inv[i + 1] % mod;
f[0] = 0;
}
inline poly ln(const poly &f,int n)
{
static poly g;
g = f,derivative(g,n),mul(g,inverse(f,n),n),integral(g,n);
return g;
}
inline poly exp(const poly &f,int n)
{
static int s[30];
static poly g,h;
int lim = 1,top = 0;
g.clear();
for(;n > 1;s[++top] = n,n = (n + 1) >> 1);
g[0] = 1;
for(;top;--top)
{
n = s[top];
for(;lim < (n << 1);lim <<= 1);
h = g,g = ln(g,n);
for(register int i = 0;i < n;++i)
g[i] = dec(f[i],g[i]);
g[0] = add(g[0],1);
ntt(g,1,lim),ntt(h,1,lim);
for(register int i = 0;i < lim;++i)
g[i] = (long long)g[i] * h[i] % mod;
ntt(g,-1,lim);
g.clear(n);
}
return g;
}
inline poly power(const poly &f,int k,int n)
{
static poly g;
g = ln(f,n);
for(register int i = 0;i < n;++i)
g[i] = (long long)g[i] * k % mod;
g = exp(g,n);
return g;
}
namespace Mod_sqrt
{
typedef pair<int,int> cp;
int w;
inline cp operator*(const cp &a,const cp &b)
{
return cp(((long long)a.first * b.first % mod + (long long)a.second * b.second % mod * w % mod) % mod,((long long)a.first * b.second % mod + (long long)a.second * b.first % mod) % mod);
}
inline cp pow(cp a,int b)
{
cp ret(1,0);
for(;b;b >>= 1)
(b & 1) && (ret = ret * a,1),a = a * a;
return ret;
}
inline int mod_sqrt(int x)
{
int y = rand() % mod;
for(;fpow(w = ((long long)y * y % mod - x + mod) % mod,mod - 1 >> 1) <= 1;y = rand() % mod);
cp ret = pow(cp(y,1),mod + 1 >> 1);
return min(ret.first,mod - ret.first);
}
}
using Mod_sqrt::mod_sqrt;
inline poly sqrt(const poly &f,int n)
{
static int s[30];
static poly g,h;
int top = 0;
g.clear();
for(;n > 1;s[++top] = n,n = (n + 1) >> 1);
g[0] = mod_sqrt(f[0]);
for(;top;--top)
{
n = s[top];
for(register int i = 0;i < n;++i)
h[i] = add(g[i],g[i]);
h = inverse(h,n),mul(g,g,n);
for(register int i = 0;i < n;++i)
g[i] = add(g[i],f[i]);
mul(g,h,n);
}
return g;
}
inline int calc(int n,int m)
{
if(m <= 0)
return 0;
static poly f;
for(register int i = 0;i <= m;++i)
f[i] = ifac[i];
f.clear(m + 1),f = power(f,n,n - 1);
return (long long)f[n - 2] * fac[n - 2] % mod;
}
int main()
{
scanf("%d%d",&len,&m),init((len - 1) << 1);
printf("%d\n",(calc(len,m - 1) - calc(len,m - 2) + mod) % mod);
}