General Information
I am part of a group of students from Charles University which includes
Todor Antić,
Ben Benčík,
Adam Džavoronok,
Guillermo Gamboa,
Jelena Glisic,
Sofiia Kotsiubynska,
Júlia Križanová,
Volodymyr Kuznietsov,
Tymofii Reizin,
Jakub Šošovička,
Filip Úradník,
and
Patrik Zavoral.
Project Description
The goal of this project is to understand possible motivation behind a platform's entry into its marketplace.
For example: What leads Amazon to launch Amazon Basics products?
To do this we will use data provided by Keepa and unsurpervised learning.
See this presentation for additional information.
Research Log
Week 1 (5/28-5/31)
- I met with my mentor and started familiarizing myself with my project (and Rutgers in general).
- I read a paper[1] about principal component analysis (PCA) and
started experimenting with Keepa and its API.
Week 2 (6/3-6/7)
- I had my first presentation where I introduced the goal of our project.
- I used the API to get data about sales rank changes of third party products after Amazon's entry.
- I read a paper[2] related to the project.
- I experimented with word embeddings and PCA - for visualization.
Week 3 (6/10-6/14)
- I fixed an error in my embedding visualization code.
- I visualized the embeddings of different Amazon subcategories using BERT[3].
- I started analyzing the sales rank behavior after the product launch.
- I read a paper[4] about interpretable embeddings.
Week 4 (6/17-6/21)
- I rewrote a lot of my code to make it better.
- I used the embedding visualization to show which categories Amazon has entered into.
- I changed the BERT model to distilBERT[5] for better performance.
- I rewrote my caching to use sqlitedict python library.
Week 5 (6/24-6/28)
- I fixed how an average is calculated from Keepa-provided timestamped data.
- For each category Amazon has entered into, I obtained the release date of the oldest Amazon product.
- For each category Amazon has entered into, I calculated the percentage of Amazon products.
- I figured out how the obtain features we want about the products.
- I rewrote my code for calculating the embedding of categories.
- I experimented a bit with using different methods to get category embeddings.
Week 6 (7/1-7/5)
- I removed weird categories from visualization.
- I added a 3D visualization.
- I calculated the influence of Amazon entry on third party products.
- I finished the code to obtain the features we want.
Week 7 (7/8-7/12)
- I analyzed the development of amazon products over the years.
- I started obtaining features of products of a chosen root category.
- I started on the prediction of future Amazon entries (so far retroactively using historical data).
Week 8 (7/15-7/19)
- I analyzed average features of products in Amazon-entered categories when compared to other products.
- I gave my final presentation summarizing all the progress we made.
- I returned to Prague.
Week 9 (7/22-7/26)
- I attended talks hosted at Charles University.
- I worked on my final paper.
References
- [1] J. Shlens. "A Tutorial on Principal Component Analysis" 2005
- [2] Zhu, Feng, and Qihong Liu. "Competing with Complementors: An Empirical Look at Amazon.com." Strategic Management Journal 39, no. 10 (October 2018): 2618-2642.
- [3] Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. "BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding" 2019
- [4] Vinamra Benara, Chandan Singh, John X. Morris, Richard Antonello, Ion Stoica, Alexander G. Huth, and Jianfeng Gao. "Crafting Interpretable Embeddings by Asking LLMs Questions" 2024
- [5] Victor Sanh, Lysandre Debut, Julien Chaumond, Thomas Wolf. "DistilBERT, a distilled version of BERT: smaller, faster, cheaper and lighter" 2020
Acknowledgements
This work was carried out while the author, Robert Jaworski, was a participant in the 2024 DIMACS REU program at Rutgers University,
CNS-2150186. The author was supported by CoSP, a project funded by European Union’s Horizon 2020 research and innovation programme,
grant agreement No. 823748.
The duck infestation
Here, we investigate the duck infestation of DIMACS, fourth floor of the CoRE building.
View log entries
June 10
During the two weeks I've spent here so far, I've been noticing small rubber ducks located around the place.
I guess they are not actually rubber ducks, because they are made of hard plastic and not rubber.
However, since they are shaped like tiny rubber ducks, I've decided to call them rubber ducks.
You know what they say - if it looks like a duck...
Anyway. I asked about the ducks and received assurances that the duck situation is just a mild infestation. Completely harmless.
Understanding that it's a joke I didn't inquire further. But after noticing the ducks in random places
(even places I would swear held no ducks just a couple minutes ago) for two weeks I decided to join in on the fun
and start this log to document the locations of the various ducks.
Next entry
June 14
This is a bit embarassing. I start a log to document the movement of ducks and then immediately forget about it.
I guess it's been a busy week. But it's over now and I can turn my attention to the important stuff - ducks.
I'm hundred percent sure the ducks are spreading. It's either this, or I'm completely blind which I refuse to accept.
One duck sits right above the elevator button which I use every day several times. (Stairs? Never heard of them.)
I don't think the duck was there when we arrived. Obviously, there is no way for me to check this so I cannot be completely sure...
Which is exactly the reason I started this log for - so that I would know where the ducks are and which types are located where.
I didn't mention the types yet? My bad. There are two types of ducks (that I know of - there could be more, after all):
plain yellow ones and glittery transparent ones. I'm not sure if there is any significance behind the choice of the duck colors,
but from what I noticed, the transparent ducks are usually solitary, whereas the yellow ones often (but not always)
appear in groups. The largest such group I discovered can be found on one of the window sills in room 401.
I only managed to search the elevator hallway and the most important room on the whole floor - room 401B (the kitchen).
Here is the list of the ducks I found:
-
A group of six yellow ducks having a meeting under the fire extinguisher in the kitchen.
I tried not to disturb them.
- A single yellow duck near the microwave.
- A single yellow duck in the corner of the right side seating area.
- A group of five yellow ducks on the sill of the whiteboard to right side of the elevator.
- A yellow duck on top of the card reader in front of the left corridor.
- A sparkly duck on top of the elevator control panel.
I have been unable to gain access to room 401, so the large duck group I mentioned is not in the list, because
I only want to include rooms I checked thoroughly.
Also, when you hover over a description, it changes the picture.
Next entry
June 15
This is just a quick update: I found a sparkly duck in the elevator. It sits on top of the hand sanitizer dispenser.
Next entry
June 18
Today I checked the corridors on the fourth floor. I have not only found quite a few ducks around,
but also an entirely new type of duck: a transparent duck. This makes me wonder how many different
types are there in total. Here is a list of ducks I found:
- A transparent duck by room 404.
- A yellow duck by room 446.
- A yellow duck on the notice board.
- A sparkly duck by room 427.
Next entry
June 19
If it weren't for this journal I would be very confused right now. Why? Because two duck locations I already knew about
got new ducks. Why? I don't know. I tried asking the ducks, but they didn't tell me anything. On the other hand, they seemed
to be conversing among themselves.
-
Another sparkly duck on top of the hand-sanitizer dispenser in the elevator.
(This is getting out of hand. Now there's two of them!)
-
A new sparkly duck on top of the card reader in front of the left corridor.
Now there's one yellow and one sparkly duck on the card reader.
Anyone know how to speak Duck? Might be interesting to learn what the ducks are talking about.
Next entry
June 20
A new type of duck today: A pinkish transparent duck by room 415.
Next entry
June 21
The ducks in the elevator are gone! I didn't even know this was possible.
I wonder how that happened. Did they move? Did someone move them? Are the ducks okay?
- The empty and sad elevator.
- A yellow duck on one of the noticeboards.
Next entry
June 25
Today's development is a bit concerning - a lot of ducks have gone missing without a trace.
Is there some kind of predator which hunts small plastic ducks?
On a brighter note, a group of four ducks has seized control of the microwave,
undoubtedly making sure everyone uses it correctly.
To fight the alarming decrease in duck numbers, I decided to bring my own sparkly duck into my office.
Hopefully, it won't get bullied by the other ducks.
- Two ducks disappeared from the card reader.
- One duck disappeared from the elevator control panel.
- Another duck disappeared from the seating area.
- The duck near the microwave is also gone.
-
However, there is now one yellow and three sparkly ducks on top of the microwave.
- My office now has a duck in it as well.
Next entry
June 26
My fears about the mass duck disappearance were completely unnecessary - today I found a record-breaking number of ducks.
Beside the large increase in population, this day also marks a significant rise in the number of known duck species.
I saw blue, orange, pink, purple and white ducks, most of them a bit translucent.
Three new offices have gained ducks.
So far, I have been unable to determine the pattern of behind which offices have ducks.
And I completely gave up on my endeavor to find the meaning behind the duck types - there's simply too many now.
- A blue duck on top of a noticeboard.
- A pinkish-orange duck on a different noticeboard.
- A purple duck on another noticeboard.
- A blue duck on top of a hand-sanitizer dispenser.
- A translucent yellow duck on a window sill on the west side of the floor.
- A yellow duck on top of another hand-sanitizer dispenser.
- A yellow duck on a window sill on the east side of the floor (the elevator side)
- A yellow duck on one of the coat racks.
- A white duck near room 426.
- A pinkish duck near room 430.
- A purple duck near room 440.
Next entry
June 27
I checked the other coat rack and wasn't disappointed - there was another yellow duck there.
The new large amount ducks almost makes me uncomfortable.
They are everywhere. Always watching. Moving only when noone is around.
Having secret meetings about duck-knows-what. I mean who-knows-what.
Next entry
July 2
Today I finally managed to get access to room 401 and discovered a veritable army of ducks sitting on the window.
Counting them was tedious, but I persevered. I can confidently say that there approximately 96 ducks sitting on the window.
Judging from the fact that the ducks are facing the room and are not looking out of window,
we can guess that the purpose of the duck army isn't to protect us against an outside threat.
Their orientation suggests that their assignment is to watch the inhabitants of the fourth floor (human, duck, or both).
Perhaps, they have to make sure none of the ducks escape?
Or are they making sure we (the humans) don't harm the other ducks?
But how could they if they are locked inside the least used room?
As with all things surrounding the ducks, the existence of the duck army makes no apparent sense and is a bit disturbing.
Perhaps, we will learn more in the future. However, I am not very optimistic.
-
There is a new yellow duck on the sill of the whiteboard.
I'm not sure if it's sick or just taking a nap.
-
There are 96 yellow ducks in room 401.
This means that there definitely more duck inhabitants of the fourth floor than human inhabitants.
Next entry
July 3
Today, two offices got new ducks. The new ducks by the offices are yellow.
The people inhabiting the office showed no increased distress due to the ducks.
They were, however, a bit distressed by my not so subtle surveillance.
Perhaps, I'm overreacting when it comes to the ducks. Everyone acts as if the situation was completely harmless.
Just a mild infestation. That's what we were told at the beginning of the program.
But to me, the presence of the ducks is a bit unsettling.
- An orange on a window sill on the west part of the floor.
-
Room 426 lost its duck. And it had been a white duck. Not an ordinary yellow one.
- Two yellow ducks by room 444.
- There are now three yellow ducks by room 446.
Next entry
July 7
I made a terrible mistake today.
I needed to print something and decided to visit my office after 9PM, completely forgetting about the ducks.
However, I was quickly reminded of their existence when I interrupted something...
something like a cult meeting in the printer room.
Do you know the feeling when you enter a room, everyone is silent and wearing an all too innocent expression on their face,
as if they were trying to convince you they were NOT just talking you and didn't just stop because you came in?
That they weren't just having a conversation they don't want you to know?
That's what it felt like... but... there were just the ducks. Unmoving. Colorful. Innocent. Everywhere.
I didn't know what to do. I quickly picked up my printed forms, fumbled with my phone to take a few pictures
(sorry for the poor quality) and fled. Er. Made a strategic retreat.
But that's not the worst part. When I returned to my office to turn off the computer, I noticed my duck,
which I placed on the table a couple weeks ago, had moved.
I don't remember moving it. But I must have done it. How else could it have changed its position?
- A group of six ducks outside room 406.
- A yellow duck on the had sanitizer dispenser inside the room 432 (the printer room).
- Ducks interfering with the post inside room 432 (it's also the mail room).
- More ducks interfering with the post.
- The duck cult.
- The duck in my office.