# Do the order of the features ie channel matter for a 1d convolutional network?

Artificial Intelligence Asked by user289602 on August 24, 2021

Do the test dataset feature order and inference (real world) feature order have to be the same as the training dataset? For example, if features are in the order (a,c,b,e,d) for the training dataset, does that particular order have to match for the inference and test dataset?

Generally, order matters. A (trained) Neural Network (NN) is just a mathematical function trained on taking some given input and producing the corresponding output. So, if you train a certain node on producing large output if (and only if) an animal is present in a picture (for example), but later you give it the numeric evidence for a car being present in the image, it will still produce large output, indicating an animal being present in the image. This is simply because a node doesn't know what it is receiving or supposed to detect. It just follows its standard mathematical procedure.

So, if you train your network on one kind of input data, you must provide the same kind of input data during testing (or at least you cannot simply exchange inputs or outputs of certain nodes without manually correcting for that in the remainder of the network).

In the most simple case, consider whether you could simply swap the inputs to that simple function: $$f(x,y) = x^2 + y$$. If you swap the inputs, the output will be different. The exactly same applies to NNs.

Addition: I think this post explains and especially illustrates the intuition nicely.

Correct answer by Daniel B. on August 24, 2021

## Related Questions

### Is A* with an admissible but inconsistent heuristic optimal?

1  Asked on August 24, 2021 by harry-stuart

### Is there a way to get landmark features automatically learned by a neural network?

1  Asked on August 24, 2021 by user784446

1  Asked on August 24, 2021 by nathan-b

### NEAT can’t solve XOR completely

0  Asked on August 24, 2021 by creepsy

### Why is GPT-3 such a game changer?

1  Asked on August 24, 2021 by parzival

### How can the FCNN reduce the dimensions of the input from $1048 times 100$ to $523 times 100$ with max-pooling?

0  Asked on August 24, 2021

### Are there examples of agents that use a more modest number of parameters on Pendulum (or similar environments)?

1  Asked on August 24, 2021

### How can one be sure that a particular neural network architecture would work?

0  Asked on August 24, 2021 by naveen-reddy-marthala

### How do we make our outputs to have the same size as the true mask?

1  Asked on August 24, 2021 by ravi-teja

### Is it common to have extreme policy’s probabilities?

1  Asked on August 24, 2021 by curiouscat22

### How AlphaGo Zero is learning from $pi_t$ when $z_t = -1$?

1  Asked on August 24, 2021

### Can we use imitation learning for on-policy algorithms?

0  Asked on February 27, 2021 by khush-agrawal

### Why am I getting a difference between training accuracy and accuracy calculated with Keras’ predict_classes on a subset of the training data?

1  Asked on February 23, 2021 by saha

### Can transformer be better than RNN for online speech recognition?

1  Asked on February 23, 2021

### In the case of invalid actions, which output probability matrix should we use in back-propagation?

1  Asked on February 20, 2021 by guineu

### Multi class text classification when having only one sample for classes

1  Asked on February 19, 2021 by fara

### How to identify segment/object that is anomaly using computer vision

0  Asked on February 14, 2021 by tyler-h

### What sort of game problems can neural networks trained/evolved with evolutionary algorithms solve, and how are they typically implemented?

3  Asked on February 13, 2021 by neomerarcana

### How are IOUs for ground truth boxes in YOLO calculated?

1  Asked on February 12, 2021 by nivter

### How to design a good evaluation function for a go-like game?

1  Asked on February 11, 2021 by nae