Conflict Resolution

 Posted on November 28, 2019  |  1 minutes  |  138 words  |  John Chen

Living in a crowded, house with different people has changed my world view, and introduced me to conflict! This is good, and expands my perspectives. These are also good exmaples to talk about in an interview. Conflict can occur in many situations; most often it occurs in daily life when we have disagreements or different perspectives. I encountered and overcame several conflicts during my time at 95 Lippincott. As a general rule, the most important rule to conflict resolution is discussion, and empathy: understanding the other person. [Read More]

Transformers_explained

 Posted on November 28, 2019  |  3 minutes  |  609 words  |  John Chen

Transformers are one of the crucial building blocks of modern day NLP. There has been much confusion on them, even though thousands of blog posts, animations exist. I finally clarified my understanding after a short 5 minute conversation! So here is my brief, opinionated explanation of a transformer! Transfomer: Is a parallelizable, and efficient way of performing neural language modelling Dispenses with RNNs; uses only attention Is non-autoregressive, and autoregressive Because of those last two points, the architecture deals with inputs in a very special way! [Read More]

Reproduction Headaches

 Posted on November 26, 2019  |  1 minutes  |  25 words  |  John Chen

I absolutely detest trying to reproduce research code! Well, as it turns out, I am just trying to reproduce working code that involves multiple dependencies.

On Travel

 Posted on November 22, 2019  |  1 minutes  |  213 words  |  John Chen

Currently writing this on a train, from Toronto to Montreal. Travel embiggens the soul. It is an activity that we can all do, and all have fun! It is a chance to put our best foot forward, and live life as nicely as possible! Here are my thoughts and tips, that I have accumulated: Train Generally, don’t really enjoy this mode! My biggest concerns are the speed, the slow wifi, and the relatively expensive price. [Read More]

Research_plans

 Posted on November 21, 2019  |  2 minutes  |  346 words  |  John Chen

Once you start looking at a problem, there are a wide variety things you want to try. Everything from related fields, to technical details might call for new analysis! Here is a public listing. Technical details: character level vs word level, the specific architecture, and whether code exists Style Transfer High and Low: Character-level and Multi-dataset Thematic Bonus This research project looks at performing style transfer using character level embeddings/language model, and then also using multiple datasets. [Read More]

Recruiting Season End

 Posted on November 21, 2019  |  1 minutes  |  65 words  |  John Chen

It’s officially done.

Recruiting season is over for me.

One thing I learned is that recruiting-mode (recruiting-me) is a very UNnatural state of affairs for me.

And I just can’t keep it up.

Even though I have follow-ups from big tier companies (ones that I didn’t know were so lucrative), I don’t feel the need to keep going.

I have Twilio, Akuna and Google left.

Divisibility

 Posted on November 20, 2019  |  1 minutes  |  56 words  |  John Chen

Divisibility in terms of sampling from a population.

This is closely related to sufficiency/calbiration in machine learning!

Quote of the day: When R has only two values we recognize this condition as requiring a parity of positive/negative predictive values across all groups.

8 different quantities for (recall) rates (recall) and predictive values (precision) (from FairML Book)

Choose Your Own Adv

 Posted on November 19, 2019  |  3 minutes  |  535 words  |  John Chen

You do training on only 5k examples of a dataset. (the first 5k indices) You notice an improvement in training. Do you follow this? Or do you ignore this? Generally, I would ignore it. It could be (and is very like to be) noise. If I really think this is a promising area, I would need a theory (curriculum learning), and then I would explore as a contribution to this field. [Read More]

Slurm

 Posted on November 18, 2019  |  1 minutes  |  85 words  |  John Chen

SLURM is a job scheduler for cluster computing. You can run commands via srun (interactive mode) and sbatch (batch processing). There are tons of flags you can set. Some of these will change whether or not you get output (specifically, if you get it in a file, or piped to the shell)! Importantly, when you submit a job, you get access to variables relating to it! You get access to SLURM environment variables [Read More]

Bi Directional Bfs Research

 Posted on November 17, 2019  |  1 minutes  |  126 words  |  John Chen

When you do research, there are two main approaches: Start with a grand idea, intuition, or otherwise behaviour you want to enforce. Perform steps in order to achieve this vision. Start off of an existing idea. Take tiny steps that change the existing work, until it is your own. I would also place addressing “what’s wrong” with the paper under this category. These two approaches are fundamentally different, and essentially represent working backwards from a solution vs starting from a problem, and finding a new contribution. [Read More]