Is it possible to get confidence intervals in LSTM forecasting?

More Utkarsh Singh's questions See All

Can someone please redirect me to a tutorial which fully explains what happens inside a LSTM network?

I need to understand the architecture, the mathematical equations, and the processing of time series data when it passes through a LSTM network.

08 September 2018 875 2 View

Why all the explanations on LSTM gates mention 0 and 1 output, whereas sigmoid and tanh can have continuous values in [0,1] and [-1,1] respectively?

I see all the explanations which say that '0' output in forget gate leads to stopping of particular value in past state, whereas '1' allows it. But the sigmoid function is continuous in the range...

08 September 2018 6,249 0 View

How to assess correlation from boxplots of a feature matrix?

Please describe what information does the box plot convey and how to understand correlation from that?

05 June 2018 5,743 6 View

While using Gaussian window for signal truncation, the term 'sqrt(pi)/a' is omitted and simply e^(−π^2t^2 /a^2) is used. Why so?

I have seen such expression in many research publications. Why is the preceding term neglected?

04 May 2018 5,293 4 View

How to compute computational complexity of various signal processing techniques?

I have some idea about the Big O notation. Please direct me to any article or tutorial so that I can derive the computational complexity of any algorithm on my own

03 April 2018 6,983 5 View

Do we have any publications on the applications of deep learning apart from image recognition?

I was going through a brief introduction on deep learning prepared by Mathworks. It gave me an idea that deep learning is mainly useful for image recognition given the fact that there's is only...

03 April 2018 5,174 6 View

What is the significance of different expressions of Gaussian window? How to decide the value of σ?

1) I see so many expressions for gaussian window across different websites (with/without '2' and 'pi'). why there are so many expressions? 2) I am going through an article which expresses the...

03 April 2018 2,644 4 View

Which is the most suitable technique for event forecasting based on real-time data analysis?

Kindly share some publications which have addressed real-life forecasting problems.

01 February 2018 2,316 11 View

What are the commonly asked interview questions in electrical engineering?

I would like to know from the professors what they would ask if they were taking an interview for a teaching position. Academicians who have faced multiple interviews are also welcome to share...

31 December 2017 9,102 1 View

Which model between LSTM and ARMA is better for forecasting time series and why?

1) Consider that the model can have seasonality and can be multivariate in nature. 2) Please also recommend the best suitable variant of LSTM or ARMA

01 January 1970 8,598 0 View

Feedback defines the constitution of an organism?

“Here is a thought experiment. Let's place Rodolpho Llinas's jarred-brain on top of a body (Fig. 1). I bet Llinas would argue that his jarred-brain retains its own consciousness, and the android...

11 August 2024 2,483 1 View

What is the difference between mathematical R^4 space and physical 4D unit space?

We assume that the difference is huge and that it is not possible to compare the two spaces. The R^4 mathematical space considers time as an external controller and the space itself is immobile in...

10 August 2024 6,678 14 View

Self-Organizing Superorganisms—as envisaged by Nenad Sestan (2018)?

The rate of glucose consumption by the neocortex is reduced by over 80% during anesthesia (Sibson et al. 1998), which disables the synapses (Richards 2002) that are inundated by glial tissue (Engl...

08 August 2024 3,118 0 View

Measuring the Intelligence of a Species?

Larger brains, which typically contain more neurons, store and transfer more information (Tehovnik and Chen 2015), but the precise relationship between number of neurons and information has yet to...

05 August 2024 1,238 2 View

How can i do multivariate Time Series forecast using MLP, ANFIS and LSTM?

I need the python code to forecast what crop production will be in the next decade considering climate and crop production variables as seen in the attached.csv file.

05 August 2024 2,977 3 View

The Curse of Evolution and Complexity?

Brain and body mass together are positively correlated with lifespan (Hofman 1993). The duration of neural development is one of the best predictors of brain size, and conception is the best...

05 August 2024 6,247 3 View

Need help with my research project on open source SIEM and machine learning?

Hello everyone, I am currently working on a research project that aims to integrate machine learning techniques into an open source SIEM tool to automate the creation of security use cases from...

04 August 2024 3,196 2 View

Swimming/space travel depends on the proprioceptive muscle spindles?

When the entire neocortex is ablated in rodents, although they are still able to swim, all the limbs move continuously and asynchronously (Vanderwolf 2006; Vanderwolf et al. 1978). Normal animals...

03 August 2024 835 3 View

What are the limitations and challenges of using machine learning for predicting concrete compressive strength in practical applications?

Machine learning (ML) has shown great potential in predicting the compressive strength of concrete, an important property for structural engineering. However, its practical application comes with...

03 August 2024 2,546 2 View

Some new emerging problems on application of RL for scheduling in IoT networks?

I have seen plenty of existing works on applied Reinforcement Learning (RL) policies for optimized scheduling in IoT networks including Q-learning, DQNs, and the newer ones including PPO for...

01 August 2024 8,754 2 View

Joaquim Assunção

https://datascience.stackexchange.com/questions/24403/prediction-interval-around-lstm-time-series-forecast

Guy Mélard

For one-step-ahead forecasts, confidence intervals are based on the distribution of residuals, either assumed (e.g. normal with a mean 0 and an estimated standard deviation, possibly with a transformation), or empirical. The problem is much more complex, to say the least, for several step-ahead forecasts, except in the special case of independence and normality of the true errors. I don't know LSTM (you should provide more details about that) so I cannot help you for the mathematical details.

Wolfgang Hitzl

I think one can use bootstrap or jackknife methods to end up with confidence intervals for predictions made with LSTM