Contributions
-
What are the common mistakes to avoid during model validation?
One mistake many often make is limiting model validation to local validation, which we do during training. Once you have deployed a model, we must periodically validate in production as well, i.e., when it is serving users. This is because, after deployment, many things could go wrong: - concept drift - covariate shift - feature nonstationarity, etc. Many often limit validation to local validation and hope things will continue to stay consistent as they were during development. But that is rarely the case. A few things I do to validate in deployment: - collect new data (if viable) - log prediction - get user feedback, etc. Next, I use this data as signals to determine the model's reliability, plan model updates, etc.
-
What techniques can you use to improve the performance of statistical models?
One thing I have often seen people overlook in building statistical models is not spending enough time understanding the data generation process. Consider generalized linear models. Every GLM stems from altering the data generation process. When I am building a statistical model, I ask this question: - What information do I get from the label about the data generation process that can help me select an appropriate statistical model? If the data generation process appears like: - Normal dist. → linear reg. - Poisson dist. → Poisson reg. - Bernoulli dist. → logistic reg. Understanding the data generation process gives so much clarity in the modeling stages. Consequently, we get to know which algorithm to use and, most importantly, why.
Activity
-
Where Did the Regularization Term Come From? Regularization is an important idea in machine learning that helps models perform better by preventing…
Where Did the Regularization Term Come From? Regularization is an important idea in machine learning that helps models perform better by preventing…
Liked by Avi Chawla
-
Where did the regularization term in loss function originate from? Experimentally, it’s pretty easy to verify the effectiveness of Regularization…
Where did the regularization term in loss function originate from? Experimentally, it’s pretty easy to verify the effectiveness of Regularization…
Shared by Avi Chawla
-
Here's a cool way to visualize skewed geographical data. Skewness can potentially distort everything you do with data: - data analysis, - data…
Here's a cool way to visualize skewed geographical data. Skewness can potentially distort everything you do with data: - data analysis, - data…
Shared by Avi Chawla
Experience & Education
-
Daily Dose of Data Science
View Avi’s full experience
See their title, tenure and more.
or
Licenses & Certifications
Volunteer Experience
-
-
-
-
CSE Department Representative
Institute day'18
- Present 6 years 6 months
Education
Had been the CSE Department Representative at the Institute Day’18 organised in collaboration with Technex’18.
-
Volunteer and Member of Event Management Team
IIT(BHU) Convocation 2017
- Present 6 years 10 months
Education
-
Web Committee
IIT BHU Students' Parliament
- 1 year 1 month
Education
Publications
-
IIT (BHU) Varanasi at MSR-SRST 2018: A Language Model Based Approach for Natural Language Generation
Association for Computational Linguistics
Courses
-
Artificial Intelligence
-
-
Computer Programming
-
-
Computer System Organisation
-
-
Data Structures
-
-
Discrete Mathematics
-
-
Engineering Mathematics-1
-
-
Engineering Mathematics-2
-
-
Information Technology Workshop-1
-
-
Information Technology Workshop-2
-
-
Machine Learning
On Udacity
-
Machine Learning
By Andrew Ng
-
Mathematical Methods
-
-
Operating Systems
-
-
Operating Systems
-
-
Probability and Statistics
-
Projects
-
Surface Realisation Shared Task'18
-
See projectShared task on Surface Realisation organised by Workshop on Multilingual Surface Realization at ACL'18, Melbourne.
Honors & Awards
-
Simplify Badge
Mastercard
In recognition of educating the Merchant Team at Mastercard with the proposed state-of-the-art AI solution to Merchant Aggregation and transitioning the pipeline with simplified explanations.
-
Sense-of-Urgency Badge
Mastercard
In recognition of the assistance provided in the delivery of Network Scores to Issuers.
-
Launch for Social Impact
Mastercard
In Recognition of Best Pro Bono Team Challenge submission & Overall Outstanding Pro Bono Contributions
Test Scores
-
JEE ADVANCED
Score: AIR 720
-
JEE Mains
Score: AIR 4497(GEN)
-
CBSE Class XII
Score: 93.6%
-
CBSE Class X
Score: CGPA 9.60
Languages
-
English
Full professional proficiency
-
Hindi
Full professional proficiency
More activity by Avi
-
One underrated SQL command: 𝐐𝐔𝐀𝐋𝐈𝐅𝐘. With 𝐐𝐔𝐀𝐋𝐈𝐅𝐘, you can filter the results of a window function like RANK() without needing another…
One underrated SQL command: 𝐐𝐔𝐀𝐋𝐈𝐅𝐘. With 𝐐𝐔𝐀𝐋𝐈𝐅𝐘, you can filter the results of a window function like RANK() without needing another…
Liked by Avi Chawla
-
Quantization: Run ML Models on Tiny Hardware Quantization is a game-changer for running machine learning models on tiny hardware. As ML engineers…
Quantization: Run ML Models on Tiny Hardware Quantization is a game-changer for running machine learning models on tiny hardware. As ML engineers…
Liked by Avi Chawla
-
If you need some reading material on improving tree-based modeling skills... ...my recent book has 7 chapters on it. Assuming you are somewhat…
If you need some reading material on improving tree-based modeling skills... ...my recent book has 7 chapters on it. Assuming you are somewhat…
Shared by Avi Chawla
-
𝐓𝐮𝐫𝐧 𝐉𝐮𝐩𝐲𝐭𝐞𝐫 𝐍𝐨𝐭𝐞𝐛𝐨𝐨𝐤𝐬 𝐢𝐧𝐭𝐨 𝐖𝐞𝐛 𝐀𝐩𝐩𝐬 𝐰𝐢𝐭𝐡 𝐌𝐞𝐫𝐜𝐮𝐫𝐲 With 𝐌𝐞𝐫𝐜𝐮𝐫𝐲, you can add interactive widgets to…
𝐓𝐮𝐫𝐧 𝐉𝐮𝐩𝐲𝐭𝐞𝐫 𝐍𝐨𝐭𝐞𝐛𝐨𝐨𝐤𝐬 𝐢𝐧𝐭𝐨 𝐖𝐞𝐛 𝐀𝐩𝐩𝐬 𝐰𝐢𝐭𝐡 𝐌𝐞𝐫𝐜𝐮𝐫𝐲 With 𝐌𝐞𝐫𝐜𝐮𝐫𝐲, you can add interactive widgets to…
Liked by Avi Chawla
Other similar profiles
Explore collaborative articles
We’re unlocking community knowledge in a new way. Experts add insights directly into each article, started with the help of AI.
Explore MoreOthers named Avi Chawla in India
-
Avi Chawla
Associate People Operations- |US Benefits|
-
Avi Chawla
Software Engineer at American Express
-
Avi Chawla
Software Engineer at Ericsson
-
Yash Avi Chawla
“Deputy Manager at ICICI Bank | Specialist in SME Growth Strategies | Empowering Small and Medium Enterprises”
18 others named Avi Chawla in India are on LinkedIn
See others named Avi Chawla