Smart India Hackathon: Horrible experience
Background:- Our task was to do load forecasting for a given area. Hourly energy consumption data for past 5 years was given to us.

One government official asks the following questions:-
1. Why are you using deep learning for the project? Why are you not doing data analysis?
2. Which neural network "algorithm" you are using? He wanted to ask which model we are using, but he didn't have a single clue about Neural Networks.

3. Why are you using libraries? Why not your own code?

Here comes the biggest one,

4. Why haven't you developed your own "algorithm" (again, he meant model)? All you have done is used sone library. Where is "novelty" in your project?

I just want to say that if you don't know anything about ML/AI, then don't comment anything about it. And worst thing was, he was not ready to accept the fact that for capturing temporal dependencies where underlying probability distribution ia unknown, deep learning performs much better than traditional data analysis techniques.
After hearing his first question, second one was not a surprise for us. We were expecting something like that. For a few moments, we were speechless. Then one of us started by showing neural network architecture. But after some time, he rudely repeated the same question, "where is the algorithm". We told him every fucking thing used in the project, ranging from RMSprop optimizer to Backpropagation through time algorithm to mean squared loss error function.

Then very calmly, he asked third question, why are you using libraries? That moron wanted us to write a whole fucking optimized library. We were speechless at this question. Finally, one of us told him the "obvious" answer. We were completely demotivated. But it didnt end here. The real question was waiting. At the end, after listening to all of us, he dropped the final bomb, WHY HAVE YOU USED A NEURAL NETWORK "ALGORITHM" WHICH HAS ALREADY BEEN IMPLEMENTED? WHY DIDN'T YOU MAKE YOU OWN "ALGORITHM"? We again stated the obvious answer that it takes atleast an year or two of continuous hardwork to develop a state of art algorithm, that too when gou build it on top of some existing "algorithm". After listening to this, he left. His final response was "Try to make a new "algorithm"".

Needless to say, we were completely demotivated after this evaluation. We all had worked too hard for this. And we had ability to explain each and every part of the project intuitively and mathematically, but he was not even ready to listen.
Now, all of us are sitting aimlessly, waiting for Hackathon to end.😢😢😢😢😢

Today I'm trying to study how to encode data in idx-ubyte format for my machine learning project.
Professors I'm going to astonish you!

Good day and good coding to all of you! :)

Ok friends let's try to compile Flownet2 with Torch. It's made by NVIDIA themselves so there won't be any problem at all with dependencies right?????? /s

Let's use Deep Learning AMI with a K80 on AWS, totally updated and ready to go super great always works with everything else.

> CUDA error
> CuDNN version mismatch
> CUDA versions overwrite
> Library paths not updated ever
> Torch 0.4.1 doesn't work so have to go back to Torch 0.4
> Flownet doesn't compile, get bunch of CUDA errors piece of shit code
> online forums have lots of questions and 0 answers
> Decide to skip straight to vid2vid
> More cuda errors
> Can't compile the fucking 2d kernel
> Through some act of God reinstalling cuda and CuDNN, manage to finally compile Flownet2
> Try running
> "Kernel image" error
> excusemewhatthefuck.jpg
> Try without a label map because fuck it the instructions and flags they gave are basically guaranteed not to work, it's fucking Nvidia amirite
> Enormous fucking CUDA error and Torch error, makes no sense, online no one agrees and 0 answers again
> Try again but this time on a clean machine
> Still no go
> Last resort, use the docker image they themselves provided of flownet
> Same fucking error
> While in the process of debugging, realize my training image set is also bound to have bad results because "directly concatenating" images together as they claim in the paper actually has horrible results, and the network doesn't accept 6 channel input no matter what, so the only way to get around this is to make 2 images (3 * 2 = 6 quick maths)
> Fix my training data, fuck Nvidia dude who gave me wrong info
> Try again
> Same fucking errors
> Doesn't give nay helpful information, just spits out a bunch of fucking memory addresses and long function names from the CUDA core
> Try reinstalling and then making a basic torch network, works perfectly fine
> FINALLY.png
> Setup vid2vid and flownet again
> SAME FUCKING ERROR

> Try to build the entire network in tensorflow
> CUDA error
> CuDNN version mismatch
> Doesn't work with TF
> HAVE TO FUCKING DOWNGEADE DRIVERS TOO
> TF doesn't support latest cuda because no one in the ML community can be bothered to support anything other than their own machine
> After setting up everything again, realize have no space left on 75gb machine
> Try torch again, hoping that the entire change will fix things

At this point I'll leave a space so you can try to guess what happened next before seeing the result.

Ready?

3

2

1

> SAME FUCKING ERROR

In conclusion, NVIDIA is a fucking piece of shit that can't make their own libraries compatible with themselves, and can't be fucked to write instructions that actually work.

If anyone has vid2vid working or has gotten around the kernel image error for AWS K80s please throw me a lifeline, in exchange you can have my soul or what little is left of it

If you don't know how to explain about your software, but you want to be featured in Forbes (or other shitty sites) as quickly as possible, copy this:

I am proud that this software used high-tech technology and algorithms such as blockchain, AI (artificial intelligence), ANN (Artificial Neural Network), ML (machine learning), GAN (Generative Adversarial Network), CNN (Convolutional Neural Network), RNN (Recurrent Neural Network), DNN (Deep Neural Network), TA (text analysis), Adversarial Training, Sentiment Analysis, Entity Analysis, Syntatic Analysis, Entity Sentiment Analysis, Factor Analysis, SSML (Speech Synthesis Markup Language), SMT (Statistical Machine Translation), RBMT (Rule Based Machine Translation), Knowledge Discovery System, Decision Support System, Computational Intelligence, Fuzzy Logic, GA (Genetic Algorithm), EA (Evolutinary Algorithm), and CNTK (Computational Network Toolkit).

🤣 🤣 🤣 🤣 🤣

Why am I such an average ?

It's just a sad realisation. Nobody cares but I wanna send this out there, just to write thoughts.. I am 18 in 3rd year of high school (grammar school so nothing IT related, basically waste of time) and in IT I'm all self taught but I feel like I could be better if I just didn't [something]..

I feel like I wanna learn so many things but when I look at you, it seems like a common problem in the IT sphere so hey, average guy joining the club.

I also feel dumb when programming. I didn't manage to learn C++ in it's entirety because to really accomplish something, you've got so many ways to do it and finding the best one requires deep understanding of the tools you've got at your disposal with the language and I feel like I'm not capable of this(self learn, in school/Uni that's different story).. But many (most) of you are. I've tried many coding challenges and when I got it working, I just saw how someone did it in one line just by layering functions that I've never heard of..

Also, we've got kinda specific national competition here in many fields including IT for high schools.. And the winners always do sometimes like "AI driven Life simulation" or "Self flying drone made from ATMega from scratch with 3D simulation in C# to it" or "Game engine" or whatever shit and it's always from grammar schools and never IT related schools.. They are like me. Maybe someone helped them, I don't know, but they are just so far away from me while I'm here struggling to get the basic level of math for any kind of machine learning..

Yeah I've written Neural Network from scratch in C but meh, honestly it's pretty basic stuff .. I'd rather understand derivatives which we're going to learn next year and I'm too lazy to learn it from khan academy because I always learn something else.. Like processing (actually codetrain started teaching tensorflow so that might be the light for me...) Or VHDL (guys you can create your own chip / CPU from scratch and it's not even hard and OMFG it's so fucking cool , full adder done yay) or RPi or commodore 64 assembly or game development with Godot and just meh..

I mean, this sounds exactly like not knowing what to do and doing nothing in the end. That was me like 6-12 months ago. Now I'm managing to pick 2-3 things and focus them and actually feel the progress.

But I lost track of the original point.. I didn't do anything special, every time I'm programming something, everyone does it better and I feel dumb. I will probably never do anything special, everyone around says "He's still learning he's genius" but they have no idea.

I mean, have you seen one of the newest videos on Google's YouTube channel (I openly hate them, but I will keep that away for now), something like "Sarah story" ? It's about girl that apparently didn't care about IT but self learned tensorflow on high school. I think it may be bullshit (like ALL of their videos ) but it's probably just fancied, not complete lie.

And again, here I am. I now C but I'm incapable of learning to program good which most of you did and are now doing for living. I'm incapable to do anything cool, just understanding what everybody else did and replicating it. I'm incapable of being clever.

Sorry, just misusing devrant to vent a bit

my mind in the last days

"Our company encourages cryptocurrency big data agile machine learning, empowerment diversity, celebrate wellness and synergy, unpack creative cloud real-time front-end bleeding edge cross-platform modular success-driven development of digital signage, powered by an unparalleled REST API backend, driven by a neural network tail recursion AI on our cloud based big data linux servers which output real time data to our Wordpress template interactive dynamic website TypeScript applet, with deep learning tensor flow capabilities.

Don't get what the fuck I just said? Udemy offers countless courses on python based buzzwords. Be the first out of 13 people to sell your soul and private information, and you'll get the first three minutes of the course free!"

QA personal voice assistant that runs locally without cloud, it’s like never ending project. I look at it from time to time and time pass by. Chat bots arrived, some decent voice algorithms appeared. There is less and less stuff to code since people progress in that area a lot.

I want to save notes using voice, search trough them, hear them, find some stuff in public data sources like wikipedia and also hear that stuff without using hands, read news articles and stuff like that.

I want to spend, more time for math and core algorithms related to machine learning and deep learning.

Problem is once I remember how basic network layers, error correction algorithms work or how particular deep learning algorithm is constructed and why is that, it’s already a week passed and I don’t remember where I started.

I did it couple of times already and every time I remember more then before but understanding core requires me sitting down with pen and paper and math problems and I don’t have time for that.

Now when I’m thinking about it - maybe I should write it somewhere in organized way. Get back to blogging and write articles about what I learned. This would require two times the time but maybe it would help to not forget.

I’m mostly interested in nlp, tts, stt. Wavenet, tacotron, bert, roberta, sentiment analysis, graphs and qa stuff. And now crystallography cause crystals are just organized graphs in 3d.

Well maybe if I’m lucky I retire in the next decade or at least take a year or two years off to have plenty of time to finish this project.

!dev
At my current work (sports wear bla bla bla) we recently had couple of brands come by (136people) and had a presentation. One of the market lead peps stood and talked about future plans and projects in the following:

👨🏻‍💼MarketGuy: "we want to improve the e-shop service and direct booking system. Think about it, AI, machine learning and deep network, these are all out there and we should consider working with it!"

👨🏻Me: ... *Thinking* "buzzwords, buzzwords everywhere.. dude you don't even know how to excel..

👨🏻‍💼: *Continues babbling about website, Blockchain and AI together with sportswear and the future of working together*

MAH GAAAD (┛◉Д◉)┛彡┻━┻
I need a coffee.. ☕

What I have learned from neutral networks for my life.

It's already a year that I'm familiar with NNs. I did not write anything serious and did not learn it that deep. But, actually, the basic knowledge gave me an interesting view to my life. I just want to share one fact with you.

There is a learning speed in NNs, which specifies how fast does the network learn. If it is too high, any new information will be accepted very easily but will wipe the past of the network's knowledge and if it is too low, the network will hardly accept new info but remember everything. When people born, they learn everything very fast and by the age they become more hard-learners Here, I've learned that you should not live in the past, and not for the current day. You just have to keep the balance.

I was once working on a deep neural network project(few years back when deep learning was just gaining momentum) and my project guide(alloted by the college) told me that this technology is useless and will be obsolete in near future. I don't know why he said that. Till this day I think what was the reason behind him saying that.

Now watching so much research done in this field, he might be realizing how much wrong he was.

Does gradient descent in artificial neural networks apply the most changes closest to the input layer?

Hi everyone. I'm sorry to take up some of your time. I've recently moved out and am now living alone (broke up with my gf of 3 years). After all the work that I put into moving out, I'm out of energy and I can't find it in myself to do what I want to do. I feel a bit trapped and need some help. If anyone knows a way out of this shithole I put myself into, I'd greatly appreciate it.

I'm also having network issues and, on top of that, I can't install CentOS 7 on a smart array... Not so smart after all, apparently.

I'm generally feeling like I've made a bad choice, but, deep inside, I know I want to focus on work and learning.

Any tips appreciated. Thanks!

The most stressful day of month.
I need to put hours into hour counting programs so computer can analyze those hours using deep learning algorithms and pay me a wage I don't deserve.

Each program work differently.
One of it works inside the local company network.
Other one I need to connect outside from company network.
In all of them I can't make mistake or I need to write to someone to fix my mistakes.
One of this programs use java applet, other is simple php website.
One of them blocks row in calendar when I click so when I login again and click I can't edit this row because it's locked by me who is editing this row.
One of them is requesting me to provide my work in minutes.
I need to follow strict procedures to report any holidays or national holidays that I need first figure out when they happen.
Wish me luck.

Looking to get a good understanding of the fundamental ideology and math behind neural networks and support vector machines. I am well versed with math so I can deal with heavier stuff if needed, I would like to see formulas but an explanation to their conception would be nice. Does anyone have any resources like this? Practical hands on practice exercises would be a plus