#pragma region template
#pragma GCC optimize("Ofast")
#include <bits/stdc++.h>
using namespace std;
#ifdef __LOCAL
 #include <debug>
#else
 #define debug(...) void(0)
#endif

using ll=long long;
using ld=long double;
using vi=vector<int>;
using vll=vector<ll>;
using pi=pair<int,int>;
using pll=pair<ll,ll>;

#define overload2(a,b,c,...) c
#define overload3(a,b,c,d,...) d
#define overload4(a,b,c,d,e,...) e
#define overload5(a,b,c,d,e,f,...) f

#define TYPE1(T) template<typename T>
#define TYPE2(T,U) template<typename T,typename U>
#define TYPE(...) overload2(__VA_ARGS__,TYPE2,TYPE1)(__VA_ARGS__)
#define TYPES1(T) template<typename... T>
#define TYPES2(H,T) template<typename H,typename... T>
#define TYPES(...) overload2(__VA_ARGS__,TYPES2,TYPES1)(__VA_ARGS__)

#define REP4(i,s,n,d) for(int i=(s);i<(n);i+=(d))
#define REP3(i,s,n) REP4(i,s,n,1)
#define REP2(i,n) REP3(i,0,n)
#define REP1(n) REP2(tomato,n)
#define REP(...) overload4(__VA_ARGS__,REP4,REP3,REP2,REP1)(__VA_ARGS__)

#define RREP4(i,s,n,d) for(int i=(n)-1;i>=s;i-=d)
#define RREP3(i,s,n) RREP4(i,s,n,1)
#define RREP2(i,n) RREP3(i,0,n)
#define RREP1(n) RREP2(tomato,n)
#define RREP(...) overload4(__VA_ARGS__,RREP4,RREP3,RREP2,RREP1)(__VA_ARGS__)

#define ALL(v) v.begin(),v.end()
#define RALL(v) v.rbegin(),v.rend()
#define SORT(v) sort(ALL(v))
#define RSORT(v) sort(RALL(v))
#define REVERSE(v) reverse(ALL(v))
#define UNIQUE(v) SORT(v),v.erase(unique(ALL(v)),v.end())

TYPES(T) void input(T&... a){ (cin>>...>>a); }
#define DECLARE(T,...) T __VA_ARGS__;input(__VA_ARGS__);
#define INT(...) DECLARE(int,__VA_ARGS__)
#define STR(...) DECLARE(string,__VA_ARGS__)
#define LL(...) DECLARE(long long,__VA_ARGS__)
#define CHR(...) DECLARE(char,__VA_ARGS__)
#define DBL(...) DECLARE(double,__VA_ARGS__)
#define VI(n,v) vi v(n);cin>>v;
#define VLL(n,v) vll v(n);cin>>v;

TYPE(T) istream&operator>>(istream&is,vector<T>&v){
  for(auto&a:v)cin>>a;
  return is;
}
TYPE(T) ostream&operator<<(ostream&os,const vector<T>&v){
  if(&os==&cerr)os<<"[";
  REP(i,v.size()){
    os<<v[i];
    if(i+1<v.size())os<<(&os==&cerr?",":" ");
  }
  if(&os==&cerr)os<<"]";
  return os;
}
TYPE(T,S) istream&operator>>(istream&is,pair<T,S>&p){
  cin>>p.first>>p.second;
  return is;
}

TYPE(T) using pq=priority_queue<T>;
TYPE(T) using pqg=priority_queue<T,vector<T>,greater<T>>;
TYPE(T) T pick(queue<T>& que){assert(que.size()); T a=que.front();que.pop();return a;}
TYPE(T) T pick(pq<T>& que){assert(que.size()); T a=que.top();que.pop();return a;}
TYPE(T) T pick(pqg<T>& que){assert(que.size()); T a=que.top();que.pop();return a;}
TYPE(T) T pick(stack<T>& sta){assert(sta.size()); T a=sta.top();sta.pop();return a;}

string YES(bool f=true){return (f?"YES":"NO");}
string Yes(bool f=true){return (f?"Yes":"No");}
string yes(bool f=true){return (f?"yes":"no");}

constexpr int INF=1e9+7;
constexpr ll LINF=ll(1e18)+7;
constexpr ld EPS=1e-10;

vi iota(int n){vi a(n);iota(ALL(a),0);return a;}
TYPE(T) void fin(T a){cout<<a<<endl;exit(0);}
TYPE(T) bool chmax(T &a,T b){return (a<b&&(a=b,true));}
TYPE(T) bool chmin(T &a,T b){return (a>b&&(a=b,true));}
TYPES(T,Ns)
auto make_vector(T x,int n,Ns ...ns){
  if constexpr(sizeof...(ns)==0)return vector<T>(n,x);
  else return vector(n,make_vector<T>(x,ns...));
}
bool in(const ll S,const int a){return (S>>a)&1;}
int popcount(const ll S){return __builtin_popcountll(S);}
#pragma endregion template

#include <atcoder/modint>
#include <atcoder/convolution>
using namespace atcoder;
using mint=modint998244353;
ostream& operator<<(ostream &os,mint a){os<<a.val();return os;}
istream& operator>>(istream &is,mint &a){
  long long b;is>>b;a=b;
  return is;
}
constexpr int SIZE=5000000;
mint Factorial[SIZE],Finverse[SIZE];
void Cinit(){
  Factorial[0]=mint(1);
  for(int i=1;i<SIZE;i++)Factorial[i]=Factorial[i-1]*mint(i);
  Finverse[SIZE-1]=Factorial[SIZE-1].inv();
  for(int i=SIZE-2;i>=0;i--)Finverse[i]=mint(i+1)*Finverse[i+1];
}
mint nCk(int n,int k){
  if(Factorial[0]==0)Cinit();
  if(n<k)return 0;
  return Factorial[n]*Finverse[n-k]*Finverse[k];
}

int main(){
  ios::sync_with_stdio(false);
  cin.tie(nullptr);

  INT(n,k);
  mint ans=0;
  REP(x,1,n)ans+=nCk(n-1,x)*mint(x).pow(k);
  fin(ans);
}