Manu Joseph @manujosephv

Problem Solver | Machine Learning Practitioner | Non-Academic Researcher deep-and-shallow.com Bengaluru, India Joined December 2008

Tweets

241
Followers

275
Following

244
Likes

460

Packt Publishing @PacktPublishing

3 years ago

"An end-to-end guide for designing and developing industry-ready forecasting systems with practical examples and lots of code. A book for beginners & professionals" - Manu Joseph, Author Get it here: packt.link/r6JxI #MachineLearning #DeepLearning #TimeSeries #ARIMA

0 7 15 2K 0

Download Image

Manu Joseph @manujosephv

3 years ago

Literally the OG!! It's come to a point that i go to PyTorch forums and directly look for ptrblck's answer.. :-D

Kai @kaikim29

3 years ago

Literally the OG!! It's come to a point that i go to PyTorch forums and directly look for ptrblck's answer.. :-D

55 193 4K 645K 137

Download Image

0 0 3 396 1

Gael Varoquaux 🦋 @GaelVaroquaux

3 years ago

Our benchmark of tree-based models vs deep learning for tabular data: final version. TL:DR: from small compute budget, @scikit_learn's HistGradientBoosting is best. With finer tuning of hyperparams, XGBoost brings a gain (here n ranges from 3,000 to 10,000)

12 115 569 89K 208

Download Image

Suman Dey @datawithsuman

3 years ago

Data Scientist has to work on different type of projects as part of their workflow. 🚀 One of the most common one is Forecasting. If you are new to forecasting, and want to learn various tips and tricks required to build better forecasting projects, get a good book 👇🏻

4 34 133 22K 62

Sairam Sundaresan @DSaience

3 years ago

Trying out something new :) Poorly drawn machine learning #1 - How momentum helps gradient descent

20 94 912 0 150

Download Image

arXiv Daily @Arxiv_Daily

3 years ago

AskYourDB: An end-to-end system for querying and visualizing relational databases using natural language deepai.org/publication/as… by @manujosephv et al. #Computation #Language

0 1 2 0 0

Stable Diffusion @StableDiffusion

3 years ago

Over the 7 weeks since Stable Diffusion's release, we've seen many amazing open-source contributions from the community. A lot of them have come in the form of awesome Google Colab notebooks! 🔥 Here is a thread of 14 awesome notebooks we've seen from the community ↓

54 680 4K 0 2K

arXiv Daily @Arxiv_Daily

3 years ago

Ensemble Creation via Anchored Regularization for Unsupervised Aspect Extraction deepai.org/publication/en… by Pulah Dhandekar and @manujosephv #DeepLearning #Computation

0 1 1 0 0

Luca Moschella @moschella_luca

3 years ago

Welcome Relative Representations, enabling zero-shot communication between latent spaces without any training! arxiv.org/abs/2209.15430 It turns out that distinct neural networks learn intrinsically equivalent latent spaces [1/6]

22 229 1K 0 508

Download Image

Manu Joseph @manujosephv

3 years ago

This is a brilliant framework to understand interpretability methods..

Christoph Molnar 🦋 christophmolnar.bsky.social @ChristophMolnar

3 years ago

This is a brilliant framework to understand interpretability methods..

3 17 115 0 77

0 0 1 0 0

Ethan Caballero @ethanCaballero

3 years ago

Adept.ai has unveiled ACT-1, a large Transformer trained to use digital tools such as a web browser. It’s hooked up to a Chrome extension that allows ACT-1 to observe the browser & take actions, like clicking, typing, & scrolling, etc‍: adept.ai/act

20 105 677 0 170

Download Video

meowbooks ✡✝ @untitled01ipynb

3 years ago

If it was physically possible for me to feel sorry about a meme it would be just about now.

7 66 551 0 33

Download Image

Pratik Bhavsar @ptkbhv

3 years ago

Is anyone tracking improvements to train large language models? Some I saw - 2D positional encodings (GLM > GPT3) - allowing arbitrary order to predict spans - Normformer - P-tuning v2 - MoE (mixture of experts) - Scaling data is as important as scaling model size (Chinchilla)

5 10 76 0 59

Jungo Kasai 笠井淳吾 @jungokasai

3 years ago

#NLProc: Stop annotating randomly-sampled data! Select-then-annotate to make language models better few-shot in-context learners (ICL)! 12% gain with GPT-3 etc over 10 tasks. 10x annotation efficiency. Different from the finetuning (FT) paradigm. arxiv.org/abs/2209.01975 1/5

8 65 364 0 116

Download Image

Manu Joseph @manujosephv

3 years ago

Me: There are 12% mislabeled images (wild animals) in ImageNet? Also Me: ImageNet SOTA is 91%

Davis Blalock @davisblalock

3 years ago

Me: There are 12% mislabeled images (wild animals) in ImageNet? Also Me: ImageNet SOTA is 91% https://t.co/XVZqW1YMka

4 21 78 0 19

Download Image

0 0 0 0 0

Download Gif

ICML Conference @icmlconf

3 years ago

🎤#icml2022 video presentations are now available. 🎊🎆 slideslive.com/icml-2022

2 71 239 0 57

Davis Blalock @davisblalock

3 years ago

"Few-shot Learning with Retrieval Augmented Language Models" New SotA on NaturalQuestions, TriviaQA, FEVER, and 5 KILT tasks with an 11B-parameter retrieval-augmented model. These results include beating models like PaLM that used up to 50x more pretraining compute. [1/9]

1 33 174 0 64

Download Image

Daniel Vila Suero @dvilasuero

3 years ago

Parsr: Transforms PDF, Documents and Images into Enriched Structured Data Nice Python toolchain for cleaning, parsing and data extraction from documents. github.com/axa-group/Parsr #python #opensource #ocr

8 67 440 0 261

Sara Hooker @sarahookr

3 years ago

Efficient Methods for Natural Language Processing: A Survey A survey that started in wonderful @dagstuhl and now includes the input of many institutions and authors. A great starting point for a practitioner or new researcher trying to navigate the NLP efficiency landscape.

4 132 622 0 272

Download Image

Adam Roberts @ada_rob

3 years ago

I've found myself referring to this table in the appendix of our recent paper (arxiv.org/pdf/2204.05832…) A LOT so I thought I'd point it out for those who might benefit from it. Basic upshot is that ED:MLM works really well, especially for classification tasks, even in 0shot.