Śreyāśḥ D 🚀
@xegression
Followers
298
Following
107K
Media
602
Statuses
5K
Here to write about Statistics· Quantifying Uncertainty 🎯 Stats · ML · Spatial PhD Fellow · Quantitative Methods @IIMKozhikode Predoc Fellow @IIM_Bangalore
Joined November 2023
someone said: Transformers are just Markov chains with extra steps. .
0
0
0
This is really cool!.Attention matrix as a discrete-time Markov chain. Google used the PageRank algo for searches using Markov chain. Same can be said for transformer
0
2
5
Classical Markov chains assume the Markov property (memoryless), but can convergence and stationarity can be proven for non-Markovian systems?. This can help for MCMC too
0
0
1
memoryless property of markov chain is very elegant mathematically, but in real-world systems does it strictly hold? . We can nullify half of models just because of these simple assumption.
Tried calculating transition probability matrix from Jobhop data from Hugging face. High stability for Professionals (0.51) and Elementary occupations (0.41) and lower stability for Armed Forces (0.06).
0
0
1
Apparently, its practically impossible for army guy to reenter into army profession. Or very hard any other profession into army profession.
Tried calculating transition probability matrix from Jobhop data from Hugging face. High stability for Professionals (0.51) and Elementary occupations (0.41) and lower stability for Armed Forces (0.06).
0
0
1
Tried calculating transition probability matrix from Jobhop data from Hugging face. High stability for Professionals (0.51) and Elementary occupations (0.41) and lower stability for Armed Forces (0.06).
0
0
0
How come Russia is largely absent from the global AI race despite its strong mathematical heritage?.Also russia has great programming minds. Is this due to brain drain or the effects of geopolitical isolation?. Maybe @Golovanov_ammoc Sir, can answer it.
240p quality.Very insightful video on Vladimir Vapnik on SVM.(Though people have stopped using it). Vapnik’s work, especially the invention of the VC-dimension and the Support Vector Machine (SVM), laid the mathematical and practical foundations.
1
0
2
3×3×3 cube that has.(8!×12!×(3^7)×(2^11) / 2) nodes. That's ≈ 43,252,003,274,489,856,000 nodes.its .practically infeasible.
can we model the Rubik’s Cube as a graph?. run a shortest-path algorithm (Dijkstra) on it?.
1
0
1
can we model the Rubik’s Cube as a graph?. run a shortest-path algorithm (Dijkstra) on it?.
0
0
1
240p quality.Very insightful video on Vladimir Vapnik on SVM.(Though people have stopped using it). Vapnik’s work, especially the invention of the VC-dimension and the Support Vector Machine (SVM), laid the mathematical and practical foundations.
0
0
2
Comparing absolute number with rates.
india has the most cheaters in the codeforces contest and the difference in 1st and 2nd rank is huge. if this continues we are getting banned from this platform as well.
0
0
0
ML Statistics. Calibration ≈ Goodness-of-fit + Residual analysis.
0
0
1
I have so many research ideas.so many new things I want to learn,.so much I want to write… .yet I’m doing neither.
3
0
7
I am working on an econometric paper and need to justify the findings ( paper is on Household consumption expenditure indicators and a debt-polynomial regression is applied). and we want to blame the bias-var tradeoff to this function.
0
0
2
Methods for Quantifying Uncertainty:. Prediction intervals;.Bayesian approaches;.Ensembles and Bootstrapping;.Dropout as Bayesian approximation;
0
0
5