Глубинное обучение (курс лекций)/2019

Материал из MachineLearning.

(Различия между версиями)
Перейти к: навигация, поиск
(Lectures)
(Lectures)
(6 промежуточных версий не показаны.)
Строка 30: Строка 30:
|-
|-
| 22 Feb. 2019 || align="center"|2 || r.685, 14-35 || Optimization and regularization methods for neural networks ||
| 22 Feb. 2019 || align="center"|2 || r.685, 14-35 || Optimization and regularization methods for neural networks ||
 +
[http://jmlr.org/papers/v15/srivastava14a.html DropOut]<br>
 +
[https://arxiv.org/abs/1502.03167 Batch Normalization]<br>
 +
[http://proceedings.mlr.press/v9/glorot10a/glorot10a.pdf Glorot initialization]<br>
 +
[https://arxiv.org/abs/1412.6980 ADAM optimizer]
|-
|-
| 01&nbsp;Mar.&nbsp;2019 || align="center"|3 || r.526b, 14-35 || Convolutional neural networks for image classification problem ||
| 01&nbsp;Mar.&nbsp;2019 || align="center"|3 || r.526b, 14-35 || Convolutional neural networks for image classification problem ||
 +
[https://drive.google.com/file/d/1DkKyDUvo5JOm1u9Lghfv-Jh1vB3297t7/view?usp=sharing Slides (pptx)]
|-
|-
| 15&nbsp;Mar.&nbsp;2019 || align="center"|4 || r.526b, 16-20 || Convolutional neural networks for image segmentation problem || [https://yadi.sk/i/xkw00-mr-Zk_Kw Slides (pdf)], [https://yadi.sk/i/n9O_1RP3QxGb8A DS Bowl 2018 (pdf)]
| 15&nbsp;Mar.&nbsp;2019 || align="center"|4 || r.526b, 16-20 || Convolutional neural networks for image segmentation problem || [https://yadi.sk/i/xkw00-mr-Zk_Kw Slides (pdf)], [https://yadi.sk/i/n9O_1RP3QxGb8A DS Bowl 2018 (pdf)]
Строка 37: Строка 42:
| 22&nbsp;Mar.&nbsp;2019 || align="center"|5 || r.526b, 14-35 || Object detection and localization on images || [https://yadi.sk/i/_WDz9dcyvwCDIA Slides (pdf)]
| 22&nbsp;Mar.&nbsp;2019 || align="center"|5 || r.526b, 14-35 || Object detection and localization on images || [https://yadi.sk/i/_WDz9dcyvwCDIA Slides (pdf)]
|-
|-
-
| 29&nbsp;Mar.&nbsp;2019 || align="center"|6 || r.523 (instead of r.645), 12-50 || Image style transfer ||
+
| 29&nbsp;Mar.&nbsp;2019 || align="center"|6 || r.523 (instead of r.645), 12-50 || Image style transfer || [https://yadi.sk/i/wPgb4U4XGX2GsQ Main models] <br> [https://yadi.sk/i/UNTi7b4yG4CHeg Enhancements 1] <br> [https://yadi.sk/i/RAiySn6LKdGIyw Enhancements 2] <br> [https://yadi.sk/i/nVgEu5taI52s1w Multi-style online models] <br> [https://yadi.sk/i/seDJTnns2_lzBw Patch-based style transfer]
|-
|-
-
| 05&nbsp;Apr.&nbsp;2019 || align="center"|7 || r.526b, 14-35 || Recurrent neural networks ||
+
| 05&nbsp;Apr.&nbsp;2019 || align="center"|7 || r.526b, 14-35 || Recurrent neural networks || [https://drive.google.com/file/d/1KvSzzctOjRhYwJH_9LJJeZhMp4USTcDV/view?usp=sharing Slides (pdf)]
|-
|-
| 12&nbsp;Apr.&nbsp;2019 || align="center"|8 || r.526b, 14-35 || Attention mechanism ||
| 12&nbsp;Apr.&nbsp;2019 || align="center"|8 || r.526b, 14-35 || Attention mechanism ||
|-
|-
-
| 19&nbsp;Apr.&nbsp;2019 || align="center"|9 || r.526b, 14-35 || Generative adversarial networks ||
+
| 19&nbsp;Apr.&nbsp;2019 || align="center"|9 || r.526b, 14-35 || Generative adversarial networks || [https://yadi.sk/i/SQynrJ3pNrEZLw Slides (pdf)]
|-
|-
| 26&nbsp;Apr.&nbsp;2019 || align="center"|10 || r.526b, 14-35 || Riemannian optimization ||
| 26&nbsp;Apr.&nbsp;2019 || align="center"|10 || r.526b, 14-35 || Riemannian optimization ||

Версия 12:32, 25 апреля 2019

This is an introductory course on deep learning models and their application for solving different problems of image and text analysis.

Instructors: Dmitry Kropotov, Victor Kitov, Nadezhda Chirkova, Oleg Ivanov and Evgeny Nizhibitsky.

E-mail for questions: bayesml@gmail.com. Please include in subject the tag [CMC DL19].

The timetable in Spring 2019: Fridays, most lectures begin at 14-35, seminars begin at 16-20. Exact place and time are given in tables below.

Link to a chat

Announcements

Rules and grades

We have 5 practical assignments during the course. For each assignment, a student may get up to 5 points. A student is allowed to upload his fulfilled assignment during one week after deadline with grade reduction of 0.1 points per day. All assignments are prepared in English.

Also each student may give a small 10-minutes talk in English on some recent DL paper. For this talk a student may get up to 5 points.

The final grade for the course is calculated as follows: Round-up (0.3*<Exam_grade> + 0.7*<Semester_grade>), where <Semester_grade> = <Assignments_total_grade> + <Talk_grade>. For the final grade 5 it is necessary to fulfill all practical assignments, get >= 4 exam grade and make an oral talk. For the final grade 4 it is necessary to fulfill at least 4 practical assignments, get >= 3 exam grade and make an oral talk. For the final grade 3 it is necessary to fulfill at least 3 practical assignments and get >=3 exam grade.

Practical assignments

Practical assignments are provided on course page in anytask.org. Invite code: bgvpqVE

Lectures

Date No. Place and time Topic Materials
15 Feb. 2019 1 r.685, 14-35 Introduction. Automatic differentiation.
22 Feb. 2019 2 r.685, 14-35 Optimization and regularization methods for neural networks

DropOut
Batch Normalization
Glorot initialization
ADAM optimizer

01 Mar. 2019 3 r.526b, 14-35 Convolutional neural networks for image classification problem

Slides (pptx)

15 Mar. 2019 4 r.526b, 16-20 Convolutional neural networks for image segmentation problem Slides (pdf), DS Bowl 2018 (pdf)
22 Mar. 2019 5 r.526b, 14-35 Object detection and localization on images Slides (pdf)
29 Mar. 2019 6 r.523 (instead of r.645), 12-50 Image style transfer Main models
Enhancements 1
Enhancements 2
Multi-style online models
Patch-based style transfer
05 Apr. 2019 7 r.526b, 14-35 Recurrent neural networks Slides (pdf)
12 Apr. 2019 8 r.526b, 14-35 Attention mechanism
19 Apr. 2019 9 r.526b, 14-35 Generative adversarial networks Slides (pdf)
26 Apr. 2019 10 r.526b, 14-35 Riemannian optimization
17 May 2019 11 r.526b, 14-35 Students' presentations

Seminars

Date No. Place and time Topic Need laptops Materials
15 Feb. 2019 1 r.685, 16-20 Automatic differentiation. No Notes on backprop
22 Feb. 2019 2 r.685, 16-20 Introduction to Azure and Pytorch Yes Notebook
01 Mar. 2019 3 r.526b, 16-20 Convolutional neural networks for MNIST Yes
22 Mar. 2019 4 r.526b, 16-20 Semantic segmentation applications No Notebook, Portrait Demo
29 Mar. 2019 5 r.526b, 14-35 Image style transfer No
05 Apr. 2019 6 r.526b, 16-20 Recurrent neural networks Yes
12 Apr. 2019 7 r.526b, 16-20 Attention mechanism Yes
19 Apr. 2019 8 r.526b, 16-20 Generative adversarial networks No
26 Apr. 2019 9 r.526b, 16-20 Riemannian optimization No
17 May 2019 10 r.526b, 16-20 Students' presentations No

Arxiv

2017

2016

Личные инструменты