Beskjeder
We did not manage to solve the problem with Devilry. The feedback on the mini-project is sent through e-mail.
We would publish the grade for the mini-project together with some comments on Devilry. Unfortunately, currently we are experiencing some technical problems. We will make the grades and comments available when these problems are solved.
Please note that the grades for the mini-project are given from the following scale: F, F-E, E, E-D, D, D-C, C, C-B, B, B-A, A.
Suggested answers for the mock exam can be found here.
The curriculum in general consists of what we have gone through in class, as well as slides, with the following exceptions:
- The slides on value-based control (subsection 2) from the reinforcement learning lecture will not be part of the curriculum.
- The slides on restricted Boltzmann machines and their applications (from Bayesian Deep Learning and Restricted Boltzmann Machines) will not be part of the curriculum.
Some of the lecture notes will also be part of the curriculum:
- Recurrent neural networks
- Recurrent neural network extensions (except section 4 on recursive neural networks)...
The mock exam can now be found here, we recommend you to give it a try before friday, when we will go through it in class.
You may now hand in your project on devilry. You may deliver as many times as you like, we will only look at the most recent one. It is only required that one person on the team hand in the project. Recall that the deadline is this Friday at 23.59h.
- That your project shows that you understand concepts, ideas and algorithms from the lectures.
- If you are able to define and express the solution for a practical problem with the help of deep learning libraries like TensorFlow
Regarding the project report: It should not need to be more than 3-4 pages (without figures). You should not write it like a scientific paper, more of a working report. Write about what you have done, things you have tried, things that didn't work, things that did work, and perhaps some ideas regarding the failures and successes (if any!). Write dow...
Last ordinary lecture of the course will be on the 23rd November. There will be no lecture on the 30th November as announced earlier. But we release a mock exam in week 49 and plan to offer a discussion related to the mock examination on the 7th December.
Due to the late availability of computing resources and other factors, we extend the deadline for final mini project submission until 30. November 23:59 Hrs.
If you need GPU computing resources for the mini project, please let us know!
We make the group session (1300-1400) on Wednesdays truely "on-demand". If you plan to participate in the session, please send a message to n.d.warakagoda@its.uio.no or eilifso@student.matnat.uio.no, at least half an hour before the start time.
If you have projects in other similar courses these may not be "reused" in this course. You may possibly do similar things, but then it should be clear what is done as part of another course. You should write this in your proposals if you find yourself in such a situation.
Assignment 3 is ready with the code and data. We have made it optional considering the workload of the final project. That means that you are not required to submit the solutions. But we strongly encourage you to perform the tasks in it.
We have opened a devilry delivery for you project proposals. The deadline is saturday 20. Oct.
Project proposal (Max. 1 page):
- Short description of the goal for your project
- Short description of what methods you intend to use
- Links or references to projects or articles you will base your work on, if any.
- Please deliver in PDF format
To give you more time for your project, we decided to make Assignment 3 optional. In other words, we will release an assignment on RNNs and text processings, but you won't have to deliver it.
We will do an extra group session on Monday the 8th at 4PM in Oslo at Ole Johan Dahls hus, the room Pascal. As usual there will be one on Wednesday at 1pm at Kjeller as well.
Ideas for the final mini project are available here.
Please note that you can pick an idea from this list or be based on your own idea.
- Depending on your TensorFlow version you may get an error about 'truncated_normal', change this to ''normal' in that case.
- If you don't have matplotlib or have issues with this, you may use PIL instead. from PIL import Image; img = Image.fromarray(numpy_array); buffer = BytesIO(); img.save(buffer, 'PNG)';
- Make sure the output of 'next_action' is of dtype 'int' with value between 0 and 3 (including). (Numpy ints cause issue, for this to work you need to change the 'not isinstance(move, int)' to 'isinstance(move, str)' the two places it occurs in simulator.py)
Deadline is Tuesday October 16th at 23.59 PM. In addition to the group session on Wednesday at 1PM there will be a session at Ole Johan Dahls hus in Oslo at 4PM on Tuesday. Note that the deadline is in just over two weeks, so you should really get started right away.
Deadline for the project will be Monday 26 of November at 23.59 pm. This is two weeks earlier than originally scheduled so that you don't have to work with your project in the exam period.
Assignment 1 deadline (27.09 23:59), is coming up!
You can deliver your solution at devilry.ifi.uio.no.
If you have any questions regarding the delivery, please email us!
There is a permanent change in room for the group sessions, they will be at room 402 from now on. The group session has now also been added to the "Schedule tab".
If you have any questions regarding the assignment, lectures or need help with e.g. installing TensorFlow you may come to "Auditoriet" at ITS between 13-14 Wednesday (first time is tomorrow)
Deadline: Thursday, September 27th, 23.59h.
PS.
For f-measure you can use 𝛽 = 1.
Remember to make the summaries, before tf.summary.merge_all() is called.
Follow lectures live at webcast-edge.uio.no/liveedge/its-301/playlist.m3u8. If it doesn't work you may want to e.g. try a different browser.