My output prediction is always the same

ayacayu · September 20, 2017, 3:41pm

Hi everyone, I have a problem. Im trying to do a model(Linear Regression Multivariable) but always it predicts the same result. My model works for one variable with a toy Data set.But When I put the Data of 3 variables It breaks.

I really need your help. My code is here. Don’t pay attention to the first part because the only thing that they do is prepare the data. I think the problem is in the model.

github.com

tratohecho3/Machine-Learning/blob/master/question.py

import torch
import torch.nn as nn
import numpy as np
import matplotlib.pyplot as plt
from torch.autograd import Variable
import pyanitools as pya
import math
import torch.nn.functional as F
import torch.nn.init as init

def ten_percent(data,output):
	ten_percent = int(len(data) * 0.1)	
	training_set = data[ten_percent:]
	test_set  = data[:ten_percent]
	output_training = output[ten_percent:]
	output_test = output[:ten_percent]

	return training_set,test_set,output_training,output_test

def changeHA(x):

This file has been truncated. show original

Thank you

FuriouslyCurious · September 20, 2017, 8:14pm

If it is predicting the same results, then your training phase is not learning anything. Make sure your training hyperparameters are suitable, and no bugs present in data being fed to regression model.

ayacayu · September 21, 2017, 1:12pm

Hi, the problem is that my model works for a input of one parameter… I change the input size and doesn’t learn anything…so what is the problem?

FuriouslyCurious · September 22, 2017, 1:57pm

You can start with the example regression code here:

Once you get official example to work, then start modifying it for your project. It will be hard to get someone else to debug your program.