วีดีโอ

In the example, at 39:09, there is a mistake regarding the last interval [t3, t4] of h(tj|x) because the discrete hazard should not be greater than 1, as it represents a probability.

Best introduction video that I have found!

Excellent

This was a great start man❤

Thanks so much for this clear explanation 🎉

thks for sharing

Good to see an unbiased and down to earth analysis of current solutions. Sadly the paper is paywalled.

Great video with beautiful explanation. Thank you for your valuable time sir.

Thank you for a good explanation, but other sources say that Zx - the Page Matrix - must be an Lx[T-L] Hankel matrix constructed from the original sequence rather than just the same sequence reshaped into Lx[T/L]. In other words, it must have the i'th column consisting of elements from X1(i) to X1(i + L) - not from X1(i*L) to X1(i*L+L). So, which is correct?

Very good.

Great

Thank you

How does this compare to Empirical Dynamic Analysis?

What to do with high censored rates/data in a "narrowed" down subset?

Another thing from third parties: Titanic is a hoax join da resistance

😜 𝚙𝚛𝚘𝚖𝚘𝚜𝚖

No wonder this guy is from CMU, great content, and clear explanation.

thanks!

Great presentation, thank you!

Thank you for your research .. I am conducting research into counterfeit cryptocurrency an this video helped me immensely

h28 vum.fyi

tremendous and helpful .great job!

great job ... helpful for us .sir keep it up......

Good Job

Perspicuous!

What a crisp and clear presentation. Nicely done!

Nice lucid presentation, thank you. When you "leave one out" for cross-validation, are you filling in gaps (like predicting Q3 numbers if you know Q1,Q2,Q4), or exclusively predicting into the future (like predicting Q4 if you know Q1,Q2,Q3) ?

Great work!

This presentation summarizes a research study on Edge Intelligence conducted at the Vienna University of Technology and the University of Ottawa. It is accepted for publication at the ACM SIGMETRICS 2020 conference as well as Proceedings of the ACM on Measurement and Analysis of Computing Systems. Table of Contents: 00:25 Use case scenario 04:19 Edge data analytics 05:36 Research goal 06:18 Staleness control problem 09:04 SCEDA algorithm 12:53 Evaluation 18:57 Key points ACM Reference Format: Atakan Aral, Melike Erol-Kantarci, and Ivona Brandić. 2020. Staleness Control for Edge Data Analytics. Proc. ACM Meas. Anal. Comput. Syst. 4, 2, Article 38 (June 2020), 24 pages. doi.org/10.1145/3392156 Atakan Aral, Melike Erol-Kantarci, and Ivona Brandić. 2020. Staleness Control for Edge Data Analytics. In ACM SIGMETRICS / International Conference on Measurement and Modeling of Computer Systems (SIGMETRICS ’20 Abstracts), June 8-12, 2020, Boston, MA, USA. ACM, New York, NY, USA, 2 pages. doi.org/10.1145/3393691.3394219 Abstract: A new generation of cyber-physical systems has emerged with a large number of devices that continuously generate and consume massive amounts of data in a distributed and mobile manner. Accurate and near real-time decisions based on such streaming data are in high demand in many areas of optimization for such systems. Edge data analytics bring processing power in the proximity of data sources, reduce the network delay for data transmission, allow large-scale distributed training, and consequently help meeting real-time requirements. Nevertheless, the multiplicity of data sources leads to multiple distributed machine learning models that may suffer from sub-optimal performance due to the inconsistency in their states. In this work, we tackle the insularity, concept drift, and connectivity issues in edge data analytics to minimize its accuracy handicap without losing its timeliness benefits. To this end, we propose an efficient model synchronization mechanism for distributed and stateful data analytics. Staleness Control for Edge Data Analytics (SCEDA) ensures the high adaptability of synchronization frequency in the face of an unpredictable environment by addressing the trade-off between the generality and timeliness of the model. Making use of online reinforcement learning, SCEDA has low computational overhead, automatically adapts to changes, and does not require additional data monitoring. Contact Author: Dr. Atakan Aral Vienna University of Technology (TU Wien) Institute of Information Systems Engineering atakan.aral@tuwien.ac.at | +43-1-58801-188734 | rucon.ec.tuwien.ac.at