A Framework for Multiple-Instance Learning的复现与思考

准确率:0.6956521739130435。

思考

1.任何都是基于假设去处理数据。本文的假设就是出现一点在正包的交点中。假设是正包交于一点且该点距离所有的包距离最短。

遇到的问题:

1.我想找一点尽可能降低该点与正包的距离。实现。只是采用了简单的欧式距离。正包为欧式距离，负包是欧式距离加负号，

2.如果对musk数据未做任何处理，会导致e的x次幂直接为0。

复现代码:

from enum import auto

from scipy.io import loadmat

import numpy as np

import torch

import torch.utils.data as data_utils

from torch import nn

import torch.optim as optim

m = loadmat(r"musk_2_original.mat")

daaa=m.keys()

from torch import nn

x=torch.ones(1, 166, requires_grad=True)

x=x*100

x=torch.tensor(x,requires_grad=True)

optimizer = optim.Adam({x}, lr=4)

scheduler = torch.optim.lr_scheduler.StepLR(optimizer, step_size=500, gamma=0.1)

m1 = loadmat(r"musk_1_original.mat")

val_bag=[]

val_bag_lable=[]

train_bag=[]

train_bag_lable=[]

for i in m1['pos_bag']:

    for j in i:

        val_bag.append(torch.Tensor(j.T))

        val_bag_lable.append(1)

for i in m1['neg_bag']:

    for j in i:

        val_bag.append(torch.Tensor(j.T))

        val_bag_lable.append(0)

for i in m['pos_bag']:

    for j in i:

        train_bag.append(torch.Tensor(j.T))

        train_bag_lable.append(1)

for i in m['neg_bag']:

    for j in i:

        train_bag.append(torch.Tensor(j.T))

        train_bag_lable.append(0)

for j in range(5000):

    l=0

    optimizer.zero_grad()

    for i in range(len(train_bag)):

        p=0

        li=[]

        if train_bag_lable[i]==1:

            for m in range(len(train_bag[i])):

                w=torch.sqrt(torch.sum((train_bag[i][m]-x)**2))

                li.append(w)

                p=p+w

            #print("min",min(li))

        if train_bag_lable[i]==0:

            for m in range(len(train_bag[i])):

                w=torch.sqrt(torch.sum((train_bag[i][m]-x)**2))

                p=p+w

        l=l+p

    l.backward()

    optimizer.step()

    test_acc=0

    dist=0

    pos=0

    neg=0

    for i in range(len(val_bag)):

        for m in range(len(val_bag[i])):

            k1=torch.sqrt(torch.sum((val_bag[i][m]-x)**2))

            if m==0:

                dist=k1

            else:

                if dist>k1:

                    dist=k1

        if dist<1000:

            if val_bag_lable[i]==1:

                test_acc=test_acc+1

                pos=pos+1

        elif val_bag_lable[i]==0:

                test_acc=test_acc+1

                neg=neg+1  

#        if val_bag_lable[i]==1:    

#            print("dist",dist)      

    print("epoch",j,"loss",l,"acc",test_acc/len(val_bag),"pos",pos,"neg",neg)