Aryan

@aryan1113

Knows a thing or two about Data Science Learning about black-boxes, one at a time ๐Ÿ˜„ IIIT'25

Joined on Feb 2, 2023

HI

  • Table of Contents :::warning Remember, Standard Deviation = โˆšVariance Variance : ฯƒยฒ Standard deviation : ฯƒ ::: What not to do
     Like  Bookmark
  • While Iโ€™m still early in my journey with e-commerce search systems, this note is a repository of search functionalities I could comprehend; which for the most of us, is beautifully abstracted as a black box. Search and Reco go hand in hand, so to draw a line we'll keep recommendations as the listings that pop up separately either on the home page, or under a listing currently being viewed by the user. Search is user-initiated, while recommendations are system-initiated. Reco has to be Near Real Time (NRT), while Search can be a bit relaxed (okay if we don't show up items listed just moments ago). Table of Contents Avenues for DS in Search We can make 'search smarter' by targeting multiple avenues, discussed below :
     Like  Bookmark
  • If you're the author, please scroll directly to the Suggestions section. :::info Here 'we' refers to the authors, it just fits my writing style, I have in no part contributed to the paper and full credits to the authors. ::: This week as I was travelling to campus for the last time, for PBI presentations, I had a ton of time to read some interesting papers. One of them is from IITGn, exploring media bias, paper link in regards to coverage on air pollution. We limit our study to only air-pollution and newspapers being our media source, due to easily available transcripts over a large time period. The study can be extended for other media sources as well, with the only pain-point being availability of transcripts.
     Like  Bookmark
  • Read this quite a while ago, while travelling back to campus for the very last time for PBI presentations (Project Based Internships, basially I swapped a semester with an internship). Arxiv Link to the very detailed paper by The Sustainability Lab from IIT Gandhinagar, talking about identifying Brick Kilns from satellite imagery. Why ? Well there isn't much information available outside of government surveys on how many kilns, and of what type operate around the country. This paper, by analyzing kilns across the Gangetic plain proposes an extendable methodology to monitor brick kiln type over time, and compliance with rules / guidelines. Surveys are expensive and can be gamed (bribing to avoid detection, too tiring to drive out and survey these kilns) What are Brick Kilns ? Basically where soft clay/earth is heated for extended periods of time to make bricks, which are then used by the construction industry.
     Like  Bookmark
  • Well unlike other notes that have an appendix at the end, I'll give some context at the very beginning so the entirety of the paper makes sense to you. Context Smart meters are basically digital electricity meters that transmit consumption data to energy providers in real time, through IoT or some wired connection, this is to help them plan better and get energy consumption data in near real time. Disaggregation is used to get an idea of power consumption by a particular device or appliance within a building or metered unit. NILM : Non Intrusive Load Monitoring Historically, NILM has had fragmented/inconsistent metrics
     Like  Bookmark
  • The 'Geographical under-representation' paper I read recently, got me intrigued with "Erasure" and read about Geographical Erasure in Language Generation Models (this paper) recently. Another paper from KIAC, IISc (wow, Kotak funds stuff other than finance-driven initiatives). Table of Contents : Why do we care about Erasure From the previous paper Due to inherent biases of the way we collect data, we can over-represent certain nations and result in an hegemony (dominance of few groups). Adding to this, User Experience would be unpleasent for users residing in these places.
     Like 1 Bookmark
  • username : root password : ROOT100 Structure of mosfet Gate control is only in 1 direction MOS region How to manufacture MOSFET Mode of operation
     Like  Bookmark
  • Basics of CMOS Technology NAND : pMOS in parallel { remember by APP, A : (N)AND, P : pMOS, P : parallel} Types of analysis Transient : calculates a circuit's response over a period of time defined by the user To calculate Rise Time Fall Time
     Like  Bookmark
  • This focuses on Introduction To Deep Learning course, that is offered by ECE dept. Class recordings, two lectures from April 2023 Batch2021 covered only the first 5chapters of the recommended book and we barely had classes post midsem. Entire April was for Project Discussion and classes moved online, so we could not cover any topic in detail. Projects were mainly stupid simple tutorial ones, try to have some uniqueness and present results comparing different learning rates and optimizers. Book to follow : Deep Learning by Ian Goodfellow, first 6chapters Playlist if you are bored, by Chai Time Data Science
     Like 1 Bookmark
  • Introduction Vivado HLx Fullform of verilog is Verify Logic HDL : Hardware Description Language Only for digital circuits Cadence is used to analyze and make circuits. Verilog is a Programming language, to make a circuit accordingly.
     Like  Bookmark
  • :beginner: Flow of session :small_blue_diamond:Kisine kuch suna hai kya:small_blue_diamond:Why this domain Background :::success What is DS, via Venn Diagram What problems do we solve Real life use cases (churn prediction, spam detection) From Data Engineer to Data Science
     Like 2 Bookmark
  • Topics to go around : Basic Probability Distributions z-score Expectation, Variance, and Co-variance Conditional Probability Independance of variables Bias and Variance Central Limit Theorem
     Like  Bookmark
  • Initial Instructions Open terminal in work folder Create new library with desired name, under technology file click "Attach to a existing technology library". Select "gpdk180" under the technology library to attach. Create new cellview from the File section (top left of window). Give name to cellview and wait a few seconds to get a grid workarea. We Integrate all instances here.
     Like  Bookmark
  • Table of content Implementing Match Cutting Link to article The term match cutting is explained in the terminology section. Items in the Netflix catalogue (series/movies/shows) have millions of frames and to create manually match cuts, one has to label cuts and match them based on memory. This method misses out on lot of possible combinations and is very time consuming. To automate selecting similar shots for transitions, we make use of neural networks.
     Like 1 Bookmark
  • Table of Contents L1 Introductory Lecture How to set learning rates : Try out exhaustive set, which works fine Design an adaptive learning rate that adapts to landscape Adaptive Learning Rates Factors : How large the gradient is
     Like  Bookmark
  • Link to Final Project Presentation Link to python code Table of contents Instructions to run file Download the predictor.py file Check if xgboost is installed locally, if not type "pip install xgboost" on command window.
     Like  Bookmark