Aesir

you too

I'll leave the question open for a bit to see if anyone else has any suggestions as well but really appreciate the time you have spent helping me

thanks for your time

Is there another way to test it?

so I want to make sure that my tuning here is as good as it can be before I move on to the test evaluation.

then run it against the test sets, I can't "undo" that

for example if I decide to take the parameters shown as the most accurate for the validation set, re-train my whole model on all of the data I set aside for training and validation

I am just hesitant to proceed from this point onwards as I will then expose my model to the test data

which made me think something must be wrong. and indeed when I use any kind of sensible configuration i.e. using more than 5 units in the LSTM layer I seem to get kind of good performance

I am just surprised that the ones I picked seemed to work kind of ok first time without realy thinking about it

Yes very true I think the hyperparameters do have some effect on the validation acurracy of the model as you said and the plots show

I agree they're hard to interpret probably because of the lack of data as you suggest. But they do seem to suggest that at least the model generalises well and as time progresses more data would become available to train and validate with. True I will also think on what else you suggested.

So as the data set is small are the hyper parameters largely irrelevant? And the the model will typically always perform well against the validation data set?

But I thought it would be better instead of jumping straight into that more complicated problem to see which methods work well on a single time series first, to see if I can learn anything that might help me design a more complicated model. Do you think the approach is wrong?

Yes I am inclined to agree with you, but I need to evaluate the feasibility of such models with these kind of (unfortunately small) data sets. I therefore want to make sure before proceeding that I have done everything I can correctly in this tuning phase, regardless of whether I use the final model or a simpler one. Ideally after this step I will work on a model which instead of forecasting one time series at a time, will produce many forecasts for all the time series at once, hopefully learning the correlations between the individual time series.

I also think it is weird my validation accuracy typically is better than my training accuracy... I have built the model using keras and it is quite simple. To my knowledge it doesn't use any dropout (or at least I have not configured any)

The data is a real time series not artificial it is unfortunately limited in size to the number of points that I have. I don't think that it is the best solution to the problem but right now I am trying to pick a suitable model from many (ARIMA, ETS, LSTMs etc) and at the moment the validation accuracy is comparable to that of simpler models. So I'd like to just confirm the model is working correctly before I evaluate it on the test set, so that I can correctly compare how good it is w.r.t the other models I have tried.

Anyway I have to work, hope you have a nice day / night!

but I see now that it is a can of worms

As for the initial start of the topic. I still feel nutrition should be a valid topic, however and I hope it remains.

sure no worries. I understand now much better. I hope my suggestion is considered about being made to be more prominent. I must say it has been extremely frustrating trying to participate here to this point.

anyway thanks for taking the time to explain that to me. It has helped me and I am sure the site will be more useful to me because of it

and I feel this will lose users

and not take it to the meta site to discuss

*up

I understand it now you have explained it too me, but many people would just give u[

but the format that you want to have here as how the knowledge base should and shouldn't be used is different to other SE communities and this needs to be made clearer

so I get that you want to save the community and website and I think there are obviously knowledgeable people here who can help and I think it is worth having as a source of information

saying we can't do personal diagnosis is fair enough I get that. but saying that you cannot provide an answer to me based on just personal experience is not clear. I have only just understood that now. and I have tried and it hasn't been enjoyable as a newcomer to use this website at all.

but there is nothing to say that you won't answer questions based on your experience in the medical field

There is a lot of stuff about how to ask a question, that you cannot give personal medical adive

A question can only be answered using knowledge shown to be demonstrably true in the form of a study clinical trial etc. We are unable to answer questions that can only be answered based on professional experience and/or personal understanding of a field.

You may be familiar with how other SE websites work. These are communities which are largely run by experienced individuals providing answers based on experience and or proven fact. Here at Medical SE we have a slightly different format due to the impact health related advice can have on an individual.

yes hang on, I'm just writing something I would have found useful

Something like

Well saying something to that affect.

It isn't really clear to me as a newcomer that this site cannot be used like other stack exchange websites

It comes across as much too difficult to use to be useful to anybody at the moment, I must say. It is probably bad timing on my part

I think that should be made clearer, it is very confusing for me at the moment what I can and can't ask here and how to.

Yes I see that now, still I think it is a shame

I came here to gain from that insight of knowledge and experience

but I want to benefit from the experience of knowledgable people here. I am able to use google myself or read a study. I don't have the insight to interpret results that an educated individual in the field would have.

I agree it needs to be strictly monitored and therefore must be incredibly difficult

Of course but I go to my doctor for their experience in the field built over time, not really because I couldn't look up something to understand my illness online

it seems very different here

If you look at other stack exchange websites they have users who answer based on mostly experience however

I think that is where a website like this is beneficial to a user.

Perhaps I do have a different understanding of how this site should be used, but I can use google. If I came here to ask a question it is because there is not sufficient information available online that I was able to find

I understand your point of view and as someone who is not responsible for moderating the site I can't comment to the increased workload allowing nutrition to continue would have. All I can say is that I am of the opinion ignoring a topic which has an unarguable link to medical health (however misunderstood) is a wrong course of action and will only serve to damage the potential growth of the community. Again I can only say this as a new user however who came here to gain a better understanding of predominantly nutrition based topics, so I am biased.

, but this shouldn't be seen as a limitation. This site has people, like you which could offer an insight into a topic. A good answer doesn't always have to be backed up with 1000 sources and an extensive medical trial. But if you as an individual have significant personal experience and can provide some insight into why never eating vegetables might have a correlation with zinc related deficiencies in the body then that question can be answered (for example). For a nutrition based question I came here to benefit from personal insight and experience of individuals not a study.