What I did last year and what I should improve on.

I made an application that takes an MMS message and uses Clarafai to identify what is in the image (mms2concept). It was deployed on AWS but I took it down since no one was using it. I have the source code for that on GitHub with the Apache license. [1]

I wrote a blog post on medium about how to think about type one and type two errors and how they related to ML applications . [2]

I wrote another blog post on how to use Quilt (a data package library) to classify flowers using a SVM.[3]

I competed in an Alexa App challenge by Capital One where me and a few others created a concept for an Alexa app that would allow you to donate your reward points to charity.

I analyzed a dataset about HackerNews using BigQuery on Google Cloud Platform. [5]

I made the first steps to make a system synthesize smart contracts using an SMT solver. I used Rosette which compiles to a system called pyramid which happens to be a lisp . [4]

I made further progress on using Knot Theory to create a useful Proof of Work since someone tried to answer my question at https://cstheory.stackexchange.com/questions/32292/knot-recognition-as-a-proof-of-work-system . If it turns out to be not successful it should be a good knot enumerator which I have been trying to design for five years.

I lost a bunch of weight going from 320 LBS to approximately 240 LBS. I think this helped me be a more productive person this year or at least one that makes me a quarter of the man I used to be.

I successfully quit using google on my laptop and phone for searching. Now I use duckduckgo. This took me at least two years.

I took a bunch of cool pictures and so did my brother. The best one that my brother took this year (in my opinion) is at the top.

I want to post more photos on my instagram.

I closed my LLC for my startup (ncategory) which I should not have made an LLC in the first place. This took a year to figure out.

I terminated a project that would allow you to clarify whether a food had peanuts or not. The reason was I failed to categorize type two and type one errors and their outcome. This formed the basis of the blogpost on item [2]

I should have invested more money into the S&P and not be complacent in my investments.

I should try to finish projects more often and get ideas validated faster.

I should follow up on people I meet at Meetups and go to more mixers but they are tiring to do all the time.

[1] https://github.com/zitterbewegung/mms2concept

[2] https://towardsdatascience.com/making-computer-vision-systems-that-dont-kill-your-users-2c98228d9032

[3] https://codeburst.io/how-to-use-quilt-to-train-an-ml-model-39f661f79cf0

[4] https://github.com/zitterbewegung/Belmod

[5] https://blog.1ppm.club/measuring-point-inflation-c88c325bf090

EDIT: Fixed a link for item 4 since it is in a new github repo.

Like what you read? Give Joshua Herman a round of applause.

From a quick cheer to a standing ovation, clap to show how much you enjoyed this story.