Practical 4.1 – RNN forward and backward
Recurrent Neural Networks – Forward and backward
Full project: https://github.com/Atcold/torch-Video-Tutorials
Notes:
13:22 – x[t] is concatenated with h[t−1]; at least it is written in green...
21:28 – Not quite. The unrolling number T represents the hierarchy you want to use for processing your input and does not necessary need to be equal or greater to the max length of your matching sequence in order to capture those relationships, given that the state is preserved across sequences chunks (h[3].new_sequence = h[3].previous_sequence), and not zeroed.
Видео Practical 4.1 – RNN forward and backward канала Alfredo Canziani
Full project: https://github.com/Atcold/torch-Video-Tutorials
Notes:
13:22 – x[t] is concatenated with h[t−1]; at least it is written in green...
21:28 – Not quite. The unrolling number T represents the hierarchy you want to use for processing your input and does not necessary need to be equal or greater to the max length of your matching sequence in order to capture those relationships, given that the state is preserved across sequences chunks (h[3].new_sequence = h[3].previous_sequence), and not zeroed.
Видео Practical 4.1 – RNN forward and backward канала Alfredo Canziani
Показать
Комментарии отсутствуют
Информация о видео
Другие видео канала
14L – Lagrangian backpropagation, final project winners, and Q&A session12 – Planning and controlWeek 15 – Practicum part B: Training latent variable energy based models (EBMs)Behind the scenesTeraDeep Image Parser02 – Supervised learning / ClassificationPerson detectorPractical 3.2 – CNN modelsWeek 9 – Practicum: (Energy-based) Generative adversarial networks[LIVE] Free energy gentle introductionPurdue theme08 – From LV-EBM to target prop to (vanilla, denoising, contractive, variational) autoencoder06L – Latent variable EBMs for structured predictionWhy not?Matrix multiplication, signals, and convolutionsPractical 3.3 – CNN trainingGoodbye to DL20,21,22,23 apartmentModel-Predictive Policy Learning with Uncertainty Regularization for Driving in Dense Traffic4 febbraio 2012 Molo AudaceStreaming test