Neural Networks: Implementation and Application Prof. Dr. Dietrich Klakow

News

28.03.2019

Final exam grades updated, Re-exam grades online

Hi,

you find the updated final main exam grades for NNIA under Materials -> Exam. -> Final grades

Additionally, you find the final grades for those of you that participated in the re-exam (including project scores) under Materials -> Exam. -> Re-exam final... Read more

Hi,

you find the updated final main exam grades for NNIA under Materials -> Exam. -> Final grades

Additionally, you find the final grades for those of you that participated in the re-exam (including project scores) under Materials -> Exam. -> Re-exam final grades

Best regards,

NNIA team

27.03.2019

Projects

Dear students, all projects have now been graded. You can see the results in the cms. In the past, some students forgot to mark their team members and similar issues. Please make sure as soon as possible that your project grades are correct, so that we can... Read more
Dear students, all projects have now been graded. You can see the results in the cms. In the past, some students forgot to mark their team members and similar issues. Please make sure as soon as possible that your project grades are correct, so that we can publish the final grades. If you encounter any issues, you can contact your tutors.
20.03.2019

Re-exam Registration

Dear students,

since some students reported problems with registering for the re-exam, we have opened an option to register on the CMS ( https://teaching.lsv.uni-saarland.de ). After logging in, you will find a registration option for the re-exam where you also... Read more

Dear students,

since some students reported problems with registering for the re-exam, we have opened an option to register on the CMS ( https://teaching.lsv.uni-saarland.de ). After logging in, you will find a registration option for the re-exam where you also submitted your assignments. To participate in the re-exam you need to register for it. Registering for the re-exam also means that you need to participate in it.

If you encounter any problems with the registration for the re-exam, you can contact Michael ( https://teaching.lsv.uni-saarland.de/nnia/tutors/ ). Please post any general questions about the re-exam in the forum.

 

 

20.03.2019

Reexam

Dear students,

as already announced in the lecture, the reexam will be on Friday, 29.03.2019. The time and location are now also fixed: 8:00-10:00 in building E22 (Günter Hotz Hörsaal). Please be aware that this time all topics will be relevant (including RNNs).... Read more

Dear students,

as already announced in the lecture, the reexam will be on Friday, 29.03.2019. The time and location are now also fixed: 8:00-10:00 in building E22 (Günter Hotz Hörsaal). Please be aware that this time all topics will be relevant (including RNNs). If you want to take part in the reexam, you have to sign up for it!

05.03.2019

Exam inspection

Hi,

the NNIA exam inspection is on Monday March 18th  from 16:00 - 18:00 in the CoLi seminar room (building C7.3).

16:00-16:30 last names starting with A-F

16:30-17:00 last names starting with G-L

17:00-17:30 last names starting with M-Sa

17:30-18:00... Read more

Hi,

the NNIA exam inspection is on Monday March 18th  from 16:00 - 18:00 in the CoLi seminar room (building C7.3).

16:00-16:30 last names starting with A-F

16:30-17:00 last names starting with G-L

17:00-17:30 last names starting with M-Sa

17:30-18:00 last names starting with Sch-Z

In case you cannot come in person, please authorise a fellow student to  represent you. Your representative has to bring her/his student ID and a sheet of paper with a scan of your and your representatives student ID, a sentence

"I <Firstname Examinee> <Lastname Examinee> <Matrikulation Number 
Examinee> authorize <Firstname Representative> <Lastname Representative> 
<Matrikulation Number Representative> to inspect my exam", the date and 
your signature. He/she would have to come when it would be your (the 
examinees) turn.

Best,

NNIA team

21.02.2019

Exam grades online

Hi,

you find the grades for the NNIA exam under Materials -> Exam. For conversion of points to grades we used a linear mapping with the 1.0 being assigned to the best exam (83.5 points achieved) and the 4.0 being 25% of 83.5 points. The resulting distribution of... Read more

Hi,

you find the grades for the NNIA exam under Materials -> Exam. For conversion of points to grades we used a linear mapping with the 1.0 being assigned to the best exam (83.5 points achieved) and the 4.0 being 25% of 83.5 points. The resulting distribution of grades is also attached.

The float grades will be combined with the project grades and then rounded to the usual levels 1,0, 1.3, etc.

Here the maximum number of points that were achieved for each question

Problem 1 Problem 2 Problem 3 Problem 4 Problem 5 Problem 6 Problem 7 Problem 8 Problem 9 Problem 10 Problem 11 Problem 12 Problem 13 Problem 14
10 10 10 9,75 10 10 10 10 9,5 10 9,5 10 10 7

 

Remark: Problem 14 were the 7 bonus questions.

The date for the exam inspection will follow in the coming days.

Best regards,

NNIA team

18.02.2019

Project 4 - Deadline Extension

Dear students,

we obtained feedback that the project deadline collides with other requirements for many students. Based on this, we will be extending the deadline for project 4 for one week. The new deadline is 07.03.2019.

You can obviously submit earlier.... Read more

Dear students,

we obtained feedback that the project deadline collides with other requirements for many students. Based on this, we will be extending the deadline for project 4 for one week. The new deadline is 07.03.2019.

You can obviously submit earlier. Please also keep in mind that this project is designed to be done over a longer time period. So, if not already done, you should start soon. Since we are dealing with neural networks, please also keep in mind that there is a certain amount of training time involved.

Regards,
NNIA team

 

11.02.2019

NNIA exam (small update)

Dear students,

we have slightly modified the cover sheet of the exam to include bonus points and to make the instructions more specific. The seating plan was also updated and now includes the students who had contacted us due to a missing registration. You find... Read more

Dear students,

we have slightly modified the cover sheet of the exam to include bonus points and to make the instructions more specific. The seating plan was also updated and now includes the students who had contacted us due to a missing registration. You find the updated files on Materials -> Exam.

Regards,

NNIA team

08.02.2019

Important Information for the NNIA exam

Dear students,

please read this information for the exam carefully:

- You need to arrive 15 minutes before the exam starts, i.e. 13:45. We will start with the exam at exactly 14:00.

- The exam will be written in two rooms, E 2.5, HS 1 (Math building) and E... Read more

Dear students,

please read this information for the exam carefully:

- You need to arrive 15 minutes before the exam starts, i.e. 13:45. We will start with the exam at exactly 14:00.

- The exam will be written in two rooms, E 2.5, HS 1 (Math building) and E 2.2 (Günther Hotz Hörsaal). Each student has a fixed seat. We have uploaded the seating plan to Materials -> Exam. Please check the seating plan beforehand to know in which room you are writing and which seat you have. The seating plan is specified by matriculation number. You have to sit at the seat that is assigned to you.

- You are not allowed to have your backpack at your seat. Please leave your backpack, mobile phone (turned off), jacket, etc. at the wall of the room before taking your seat.

- We have published the cover sheet of the exam at Materials -> Exam. Feel free to familiarize yourself already with the format.

- If you have any last questions, ask them in advance on the forum and not on the day of the exam itself. The earlier you ask them, the better.

 

The NNIA team wishes you all the best for the exam!

05.02.2019

Lecture moved today

Dear NNIA participants,

note that today’s lecture is moved to Seminarraum 001 in E2 1 (Bioinformatik)

Best

Dietrich Klakow

29.01.2019

Fourth project is online

We have just released the final project. You find it under Materials. The goal of this project is to get familiar with different flavours of stochastic gradient descent and recurrent neural networks (vanilla RNN and LSTMS). The project consists of an .ipynb jupyter... Read more

We have just released the final project. You find it under Materials. The goal of this project is to get familiar with different flavours of stochastic gradient descent and recurrent neural networks (vanilla RNN and LSTMS). The project consists of an .ipynb jupyter notebook file and an additional train.txt file. As always, the notebook contains all necessary instructions for you. If there are any questions, do not hesitate to ask them in the Forum. The deadline for this project is 28.02.2019

Regards,

NNIA team
22.01.2019

Assignment 10 online + updated slides + exam registration

The assignment sheet 10 has been released in the CMS. As always, you find it under Materials. The slides of chapter 8 (optimization) where changed shortly before the lecture. We have uploaded now the new version to the CMS. If you downloaded the slides before the... Read more

The assignment sheet 10 has been released in the CMS. As always, you find it under Materials. The slides of chapter 8 (optimization) where changed shortly before the lecture. We have uploaded now the new version to the CMS. If you downloaded the slides before the lecture, please make sure to update them because the slide numbers in the assignment sheet refer to the new version of the chapter.

 

Please do not forget to register for the exam in HIS-POS/LSF. This is a requirement to participate in the exam.
 

Regards,
NNIA team

15.01.2019

Assignment 9 online

The assignment sheet 9 has been released in the CMS. As always, you find it under Materials.

Regards,
NNIA team

09.01.2019

Comment on Project 3

Given recent feedback, we would like to point out that project 3 does contain on purpose topics that have not been covered in the lecture, yet. We choose this format of self-study among other reasons due to the lecture that got cancelled. The exercises expect more... Read more

Given recent feedback, we would like to point out that project 3 does contain on purpose topics that have not been covered in the lecture, yet. We choose this format of self-study among other reasons due to the lecture that got cancelled. The exercises expect more self-study than usual but we have provided links to resources in the exercises. The lecture will cover some of the topics but only shortly. Obviously, if you have any questions you can ask as always in the tutorials, after the lecture, in the office hour or in the forum.

05.01.2019

Third project is online

We have just released the third project. You find it under Materials. The goal of this project is to familiarize yourself with the concepts of dropout, early stopping, and data augmentation. The project consists of an .ipynb jupyter notebook file. The notebook... Read more

We have just released the third project. You find it under Materials. The goal of this project is to familiarize yourself with the concepts of dropout, early stopping, and data augmentation. The project consists of an .ipynb jupyter notebook file. The notebook contains all necessary instructions for you. If there are any questions, do not hesitate to ask them in the Forum. The deadline for this project is 26.01.2019. As always we recommend that you start early.

Regards,
NNIA team
04.01.2019

No lecture 08.01. and new deadline exercise sheet 8

Dear students,

 

there will be no lecture on Tuesday, 08.01.2019. The next lecture will be in the following week (15.01.2019). The tutorials will continue as usual in the week 07.01. - 11.01. For details, please check the calendar on... Read more

Dear students,

 

there will be no lecture on Tuesday, 08.01.2019. The next lecture will be in the following week (15.01.2019). The tutorials will continue as usual in the week 07.01. - 11.01. For details, please check the calendar on https://teaching.lsv.uni-saarland.de

 

Since there will be no lecture, the deadline of the current exercise sheet 8 will be postponed from 07.01. to 14.01. You can obviously also submit earlier, if you like. The project's deadline stays unchanged.

 

Regards,

NNIA team

18.12.2018

Eigth assignment online + feedback

The 8th assignment sheet has been released in the CMS. As always, you find it under Materials.

We got the feedback that certain parts of project 2 might benefit from additional information. If you are missing information to solve an exercise, please ask in the... Read more

The 8th assignment sheet has been released in the CMS. As always, you find it under Materials.

We got the feedback that certain parts of project 2 might benefit from additional information. If you are missing information to solve an exercise, please ask in the forum mentioning the specific (!) part of the exercise where you are stuck. We'll then try to help you there. If we update a project or exercise sheet based on such questions, we'll add a news item on the webpage (teaching.lsv.uni-saarland.de). To not spam you with e-mails, we'll not send around a notification in these cases.

We were also asked to provide an explanation for the general backpropagation on the slides. Please note that the slides in chapter 6 in "Example: Backward Propagation" also mention the formulas for a general activation function. Additionally, we uploaded the notes of the general derivation performed on the blackboard. You find it under Materials -> Miscellaneous.

 

The NNIA team wishes you a nice lecture-free break and a good start into the new year!

18.12.2018

Second project update 2

Based on feedback from the forum, we changed two minor issues in project 2:

- X_train.shape[1] was replaced with X_trainval.shape[1] in task 2 (see forum)

- The loss was specified as cross-entropy loss and an epsilon was added for stable computations in task 2... Read more

Based on feedback from the forum, we changed two minor issues in project 2:

- X_train.shape[1] was replaced with X_trainval.shape[1] in task 2 (see forum)

- The loss was specified as cross-entropy loss and an epsilon was added for stable computations in task 2 (see forum)

If you already started with the project, you do not need to download the new ipynb and can directly change it in your existing notebook.

11.12.2018

Seventh assignment is online

The 7th assignment sheet has been released in the CMS. As always, you find it under Materials. Please, also keep in mind the instructions for the submission.

Best,
NNIA team

10.12.2018

Second project update

A few hours after releasing the second project on Sunday, we uploaded a revision. It does not change the assignments but it makes the code a bit better readable and easier to use. Please make sure that you are using the updated version of the project. You can easily... Read more

A few hours after releasing the second project on Sunday, we uploaded a revision. It does not change the assignments but it makes the code a bit better readable and easier to use. Please make sure that you are using the updated version of the project. You can easily check that you have the new version: If you manually downloaded the additional FashionMNIST dataset files, you have the old version. If the code automatically obtains the data for you, you have the revision.

09.12.2018

Second project is online

​We have just released the second project. You find it under Materials. The goal of the second project is to familiarize yourself with Tensorflow and to implement your first neural networks. The project consists of an .ipynb jupyter notebook file. The notebook... Read more

​We have just released the second project. You find it under Materials. The goal of the second project is to familiarize yourself with Tensorflow and to implement your first neural networks. The project consists of an .ipynb jupyter notebook file. The notebook contains all necessary instructions for you. If there are any questions, do not hesitate to ask them in the Forum.

 
The deadline for this project is 04.01.2019, but we recommend that you start early. Please be aware that the first week of January is course free ("vorlesungsfrei") [1], so there won't be any tutorial or office hour right before the deadline.
 
Regards,
NNIA team
 
04.12.2018

Sixth assignment is online + Forum

The 6th assignment sheet has been released in the CMS. As always, you find it under Materials.

Recently, several students told us that they are not able to log into the forum. Please read the instructions here carefully. If you have a UdS university e-mail... Read more

The 6th assignment sheet has been released in the CMS. As always, you find it under Materials.

Recently, several students told us that they are not able to log into the forum. Please read the instructions here carefully. If you have a UdS university e-mail address that starts with s8 or s9, you should be able to access the forum. If you have never accessed the forum or if you have trouble login in, you need to request a password for your account here. Your LDAP forum account is different from the account you use on teaching.lsv.uni-saarland.de to submit your assignments! If you still have problems with the access, feel free to contact Michael.

Best,

NNIA team

 

 

27.11.2018

Fifth assignment is online

The 5th assignment sheet has been released in the CMS. As always, you find it under Materials.

Best,

NNIA team

20.11.2018

Fourth assignment is online

The 4th assignment sheet has been released in the CMS. As always, you find it under Materials.

Best,

NNIA team

19.11.2018

Exercise 3.2 postponed to 26.11.2018

Based on the fact that the softmax function was not covered it the lecture yet we decided to postpone the deadline of exercise 3.2 on assignment 3 to next week.

That is, when handing in assignment 4, please attach a solution of exercise 3.2 to the .pdf... Read more

Based on the fact that the softmax function was not covered it the lecture yet we decided to postpone the deadline of exercise 3.2 on assignment 3 to next week.

That is, when handing in assignment 4, please attach a solution of exercise 3.2 to the .pdf file. 

In case you already have a solution to the exercise, you can of course hand it in already. 

Best,

NNIA team

16.11.2018

Update to first project

We have updated the first exercise (PCA) of the project. We changed the feature normalization in Task 1 and added a new question (Question 4).

Please make sure to download the new version of the project. You find it under Materials

Best,

NNIA team

15.11.2018

First project is online

We have just released the first project. You find it under Materials. The goal of the first project is to familiarize yourself with PCA and multiple linear regression. The project consists out of a single .ipynb jupyter notebook file. The notebook contains all... Read more

We have just released the first project. You find it under Materials. The goal of the first project is to familiarize yourself with PCA and multiple linear regression. The project consists out of a single .ipynb jupyter notebook file. The notebook contains all necessary instructions for you. If there are any questions, do not hesitate to ask them in the Forum

All necessary libraries are contained in the NNIA_conda_env.yml file. Make sure to install and activate it using conda. 

Best,

NNIA team

13.11.2018

Third assignment is online

The 3rd assignment sheet has been released in the CMS. As always, you find it under Materials.

Exercise 3.2 covers the softmax function. While it is on the slides of today's lecture (chap4), it was not yet discussed in class.

If you have troubles with this... Read more

The 3rd assignment sheet has been released in the CMS. As always, you find it under Materials.

Exercise 3.2 covers the softmax function. While it is on the slides of today's lecture (chap4), it was not yet discussed in class.

If you have troubles with this exercise, please check out chapter 4 of the Deep Learning Book (which you also find under materials).

In general, we recommend reading this book in addition to attending the lectures and tutorials. 

Further, we would like to encourage you again to use the Forum for asking questions related to the assignment sheets. 

 

Best,

NNIA team

10.11.2018

Office Hours

Starting from next week Friday (16.11.2018), we will offer office hours. This is an additional chance for you to discuss the assignments or any other NNIA related questions with the tutor team. Office hours take place in building C7 1 in room 0.03 (students room)... Read more

Starting from next week Friday (16.11.2018), we will offer office hours. This is an additional chance for you to discuss the assignments or any other NNIA related questions with the tutor team. Office hours take place in building C7 1 in room 0.03 (students room) from 14:00 – 15:00.

29.10.2018

Tutorials

You have been automatically assigned to a tutorial slot based on your preferences. 

Note that the tutorials start from Monday (29.10.2018) and there will be no tutorial on Thursday (01.11.2018).

However, all students of the Thursday tutorial are free to attend... Read more

You have been automatically assigned to a tutorial slot based on your preferences. 

Note that the tutorials start from Monday (29.10.2018) and there will be no tutorial on Thursday (01.11.2018).

However, all students of the Thursday tutorial are free to attend any other tutorial this week. 

The first tutorial will cover an introduction to NumPy, so please make sure to bring a laptop if possible.

Finally, since this question came up many times, it is possible to form groups with students from other tutorials. 

Show all
 

Neural Networks: Implementation and Application

 

Audience

This advanced lecture addresses Bachelor and Master students in Computerlinguistics, Computer Science, CuK, Mechatronics, or VC.

Contents

We plan to cover the following topics:

  1. Linear Algebra and Principal Component Analysis (PCA)
  2. Numerical Computation
  3. Machine Learning Basics
  4. Deep Feedforwad Neural Networks
  5. Regularization for Deep Learning
  6. Optimization for Deep Learning
  7. Convolutional Neural Networks
  8. Sequence Modelling: Recurrent and Recursive Neural Networks

Organisation

Lectures take place in HS II in building E1 3.

Time: Tuesday, 14:15-15:45 

Starts: October 23th

The final exam will take place on February 12th, from 14:00-16:00 in Günter-Hotz-Hörsaal building E2 2 and HS I in building E2 5. 



Privacy Policy | Legal Notice
If you encounter technical problems, please contact the administrators