我正在创建一个神经网络,并希望使用hash_map来保持每个神经元的输出神经元的权重参考:
class Neuron; //forward declaration was there (sorry I forgot to show it earlier)
typedef double WEIGHT;
typedef stdext::hash_map<boost::shared_ptr<Neuron>,WEIGHT> NeuronWeightMap;
class Neuron
{
private:
NeuronWeightMap m_outputs;
//...
public:
Neuron();
~Neuron();
//...
WEIGHT GetWeight(const boost::shared_ptr<Neuron>& neuron) const
{
NeuronWeightMap::const_iterator itr = m_outputs.find(neuron);
if( itr != m_outputs.end() )
{
return itr->second;
}
return 0.0f;
}
};
我意识到我不能使用boost :: shared_ptr作为stdext :: hash_map的键,那么另一个建议是什么呢?是否有任何解决方法或是使用不同密钥或切换到std :: map的唯一选择? 谢谢!
以下是错误:
1>c:\program files (x86)\microsoft visual studio 8\vc\include\xhash(61) : error C2440: 'type cast' : cannot convert from 'const boost::shared_ptr<T>' to 'size_t'
1> with
1> [
1> T=Neuron
1> ]
1> No user-defined-conversion operator available that can perform this conversion, or the operator cannot be called
1> c:\program files (x86)\microsoft visual studio 8\vc\include\xhash(99) : see reference to function template instantiation 'size_t stdext::hash_value<_Kty>(const _Kty &)' being compiled
1> with
1> [
1> _Kty=boost::shared_ptr<Neuron>
1> ]
1> c:\program files (x86)\microsoft visual studio 8\vc\include\xhash(98) : while compiling class template member function 'size_t stdext::hash_compare<_Kty,_Pr>::operator ()(const _Kty &) const'
1> with
1> [
1> _Kty=boost::shared_ptr<Neuron>,
1> _Pr=std::less<boost::shared_ptr<Neuron>>
1> ]
1> c:\program files (x86)\microsoft visual studio 8\vc\include\hash_map(80) : see reference to class template instantiation 'stdext::hash_compare<_Kty,_Pr>' being compiled
1> with
1> [
1> _Kty=boost::shared_ptr<Neuron>,
1> _Pr=std::less<boost::shared_ptr<Neuron>>
1> ]
1> c:\program files (x86)\microsoft visual studio 8\vc\include\xhash(119) : see reference to class template instantiation 'stdext::_Hmap_traits<_Kty,_Ty,_Tr,_Alloc,_Mfl>' being compiled
1> with
1> [
1> _Kty=boost::shared_ptr<Neuron>,
1> _Ty=common_ns::WEIGHT,
1> _Tr=stdext::hash_compare<boost::shared_ptr<Neuron>,std::less<boost::shared_ptr<Neuron>>>,
1> _Alloc=std::allocator<std::pair<const boost::shared_ptr<Neuron>,common_ns::WEIGHT>>,
1> _Mfl=false
1> ]
1> c:\program files (x86)\microsoft visual studio 8\vc\include\hash_map(90) : see reference to class template instantiation 'stdext::_Hash<_Traits>' being compiled
1> with
1> [
1> _Traits=stdext::_Hmap_traits<boost::shared_ptr<Neuron>,common_ns::WEIGHT,stdext::hash_compare<boost::shared_ptr<Neuron>,std::less<boost::shared_ptr<Neuron>>>,std::allocator<std::pair<const boost::shared_ptr<Neuron>,common_ns::WEIGHT>>,false>
1> ]
1> FILE_PATH_REMOVED\neuralnet.h(21) : see reference to class template instantiation 'stdext::hash_map<_Kty,_Ty>' being compiled
1> with
1> [
1> _Kty=boost::shared_ptr<Neuron>,
1> _Ty=common_ns::WEIGHT
1> ]
1>NeuralNet.cpp
1>c:\program files (x86)\microsoft visual studio 8\vc\include\xhash(61) : error C2440: 'type cast' : cannot convert from 'const boost::shared_ptr<T>' to 'size_t'
1> with
1> [
1> T=Neuron
1> ]
1> No user-defined-conversion operator available that can perform this conversion, or the operator cannot be called
1> c:\program files (x86)\microsoft visual studio 8\vc\include\xhash(99) : see reference to function template instantiation 'size_t stdext::hash_value<_Kty>(const _Kty &)' being compiled
1> with
1> [
1> _Kty=boost::shared_ptr<Neuron>
1> ]
1> c:\program files (x86)\microsoft visual studio 8\vc\include\xhash(98) : while compiling class template member function 'size_t stdext::hash_compare<_Kty,_Pr>::operator ()(const _Kty &) const'
1> with
1> [
1> _Kty=boost::shared_ptr<Neuron>,
1> _Pr=std::less<boost::shared_ptr<Neuron>>
1> ]
1> c:\program files (x86)\microsoft visual studio 8\vc\include\hash_map(80) : see reference to class template instantiation 'stdext::hash_compare<_Kty,_Pr>' being compiled
1> with
1> [
1> _Kty=boost::shared_ptr<Neuron>,
1> _Pr=std::less<boost::shared_ptr<Neuron>>
1> ]
1> c:\program files (x86)\microsoft visual studio 8\vc\include\xhash(119) : see reference to class template instantiation 'stdext::_Hmap_traits<_Kty,_Ty,_Tr,_Alloc,_Mfl>' being compiled
1> with
1> [
1> _Kty=boost::shared_ptr<Neuron>,
1> _Ty=common_ns::WEIGHT,
1> _Tr=stdext::hash_compare<boost::shared_ptr<Neuron>,std::less<boost::shared_ptr<Neuron>>>,
1> _Alloc=std::allocator<std::pair<const boost::shared_ptr<Neuron>,common_ns::WEIGHT>>,
1> _Mfl=false
1> ]
1> c:\program files (x86)\microsoft visual studio 8\vc\include\hash_map(90) : see reference to class template instantiation 'stdext::_Hash<_Traits>' being compiled
1> with
1> [
1> _Traits=stdext::_Hmap_traits<boost::shared_ptr<Neuron>,common_ns::WEIGHT,stdext::hash_compare<boost::shared_ptr<Neuron>,std::less<boost::shared_ptr<Neuron>>>,std::allocator<std::pair<const boost::shared_ptr<Neuron>,common_ns::WEIGHT>>,false>
1> ]
1> FILE_PATH_REMOVED\neuralnet.h(21) : see reference to class template instantiation 'stdext::hash_map<_Kty,_Ty>' being compiled
1> with
1> [
1> _Kty=boost::shared_ptr<Neuron>,
1> _Ty=common_ns::WEIGHT
1> ]
1>Generating Code...
答案 0 :(得分:11)
您的.h文件中可能已声明Neuron
。
所以下面的内容可行:
struct hasher {
size_t operator()(const boost::shared_ptr<Neuron>& n) { return (size_t)n.get(); }
};
typedef stdext::hash_map<boost::shared_ptr<Neuron>,WEIGHT,hasher> NeuronWeightMap;
问题可能是boost:shared_ptr
没有默认的散列函数(看起来像什么?),所以你只需提供一个。
答案 1 :(得分:8)
您还可以为哈希提供模板专业化:
#include <functional>
#include <boost/shared_ptr.hpp>
template<class T>
class std::tr1::hash<boost::shared_ptr<T>> {
public:
size_t operator()(const boost::shared_ptr<T>& key) const {
return (size_t)key.get();
}
};
与unordered_set完美配合:
class Foo;
typedef boost::shared_ptr<Foo> FooPtr;
typedef std::tr1::unordered_set<FooPtr> FooSet;
答案 2 :(得分:3)
“普通”Neuron*
应该是hash_map
:
typedef stdext::hash_map<Neuron*, WEIGHT> NeuronWeightMap;
如果要查找shared_ptr
- 神经元,可以使用shared_ptr
s get()
方法访问原始指针:
NeuronWeightMap weights;
boost::shared_ptr<Neuron> n;
weights.find(n.get());
答案 3 :(得分:1)
由于您已经使用了Boost,为什么不将boost::unordered_map
用于哈希表?虽然您最好检查一下如何正确地执行此操作,但它可以使用shared_ptr
开箱即用(Boost.Unordered)
答案 4 :(得分:0)
从上面的代码编译错误是什么?
你有没有尝试过宣布神经元课程?
class Neuron; // class Neuron must be forward declared to be used
typedef double WEIGHT;
typedef stdext::hash_map<boost::shared_ptr<Neuron>,WEIGHT> NeuronWeightMap;
class Neuron
{
private:
NeuronWeightMap m_outputs;
//...
public:
Neuron();
~Neuron();
//...
};
在这种情况下,标准地图也应该没问题,你使用hash_map的具体原因是什么?
答案 5 :(得分:0)
通过添加hash_value()函数,我可以使用shared_ptr作为stdext :: hash_map中的键。 下面我使用boost :: unordered_map与std :: hash_map
进行比较#include <boost/smart_ptr.hpp>
#include <boost/unordered_map.hpp>
#include <hash_map>
#include <assert.h>
struct MyItem
{
int i;
};
typedef boost::shared_ptr<const MyItem> MyItem_csptr;
struct MyExtra
{
double d;
};
//Boost's unordered_map already knows how to hash shared_ptr
typedef boost::unordered_map<MyItem_csptr,MyExtra> MapMyItemToExtra;
//But for stdext::hash_map, we need to write this:
template<class T>
size_t hash_value(const boost::shared_ptr<T>& aSptr)
{
return reinterpret_cast<size_t>( aSptr.get() );
};
typedef stdext::hash_map<MyItem_csptr,MyExtra> MapMyItemToExtra2;
template<class MAP>
void TryMapMyItemToExtra_T()
{
MAP tMap;
MyItem_csptr tItem1(new MyItem);
MyItem_csptr tItem2(new MyItem);
tMap[tItem1].d=1.0;
tMap[tItem1].d=1.1;
tMap[tItem2].d=2.0;
assert( tMap[tItem1].d == 1.1 );
assert( tMap[tItem2].d == 2.0 );
}
void TryMapMyItemToExtra()
{
TryMapMyItemToExtra_T<MapMyItemToExtra>();
TryMapMyItemToExtra_T<MapMyItemToExtra2>();
}
答案 6 :(得分:0)
哈希函数的问题也出现在stdext::hash_set
中。声明hash_value()
的解决方案是最好的,因为您只能在主头文件中声明它一次。