This is how this year began:
I spent the new year camping in a wild hot spring (梵梵溫泉) near Yilan with a couple of friends.
This year I decided to keep track of developments in different spheres of my life. Here I am sharing just two -
- (Technical) Things I learnt
- Fitness
Fitness
January
- 3 muscle-ups in a row
February and March
- None
April
- 4 muscle-ups in a row
May
- 200 burpees in 35 mins
- 5 pull-ups, 1 muscle-up, 4 pull-ups in a row
- (1 pull-up, 1 muscle-up, 1 pull-up, 1 muscle-up, 1 pull-up, 1 muscle-up) in a row
June
- 5 muscle-ups in a row
- Began doing 9 min HIIT every morning
July
- Swam 1500m while breathing on both sides
August
- Went downhill due to falling sick
September
- 3300m swim across the Sun Moon Lake
October
- 100 burpees in 9 min 34 sec
- 50 burpees in 4 min 25 sec -> 4 min -> 3 min 36 sec -> 3 min 35 sec
November
- Hanging from bar for 3 min 15 sec (switched hands in between a couple of times)
December
- Hanging from bar for 4 min 30 sec (while switching hands in between)
- 24 dead hang pull-ups in a row
Things I learnt
January
- Contrastive Predictive Coding
February
- Self-supervised learning using contrastive comparison
- Transformers - architecture and implementation
- Some progress in Julia
March
- Neural ODEs and the adjoint method (still shaky on the details)
- Some more progress in Julia
April
- Finished Differential Equations (MIT 18.03)
- Reviewed multivariable calculus on Khan academy to brush up on concepts like flux, Green's theorem etc
May
- Finished "Fourier Transform and its Applications" by Stanford.
- Began Group Theory (MIT OCW)
June
- Normalizing flows in detail (part of Deep Unsupervised Learning by UCB)
- More Group Theory (MIT OCW)
July
- After umpteenth time, finally got a firm grasp on variational inference (ELBO or VLB), again as a part of Deep Unsupervised Learning by UCB
- Deep learning on graphs: Starting looking into source code of Deep Graph Library
- decided to take a bottom-up approach to learning DL on graphs
August
- More Group theory
- (didn't get much done as I was bedridden for two weeks)
September
- Finished first two courses of Algorithms Specialization on Coursera
October
- Finished course three of Algorithms Specialization on Coursera
- Pell's equations and how they are related to square-triangular numbers
- Euclidean algorithm and its connection to continued fractions (and its elegant visual representation)
- Smooth AP: How a team at Oxford managed to make MAP differentiable
- Brushed up on transformer layer again
November
- Finished Algorithms Specialization on Coursera
- [Revision] Variational inference and its connection with importance sampling
- Began reading Axler's Linear Algebra Done Right and Spivak's Calculus on Manifolds
- Spivak's book is ridden with typos and errors - discontinued reading it after about two weeks
December
- Halfway through Deep RL course by UCB
- Only possess theoretical knowledge currently. Will spend some time creating some projects before moving on with this course.
- NeurIPS papers:
- Deep graph learning
- Learned message passing using DGL
- Re-implemented GraphConv and GraphAttentionConv using DGL
Posts
0 comments:
Post a Comment