jaidevd

Talk Details

Title

Optimizing PyTorch Models

Elevator Pitch - 300 chars

Most AI models never leave a Jupyter notebook. This workshop shows how to turn

jaidevd

#!/usr/bin/env python3
"""
Generate an HTML page that embeds an image as a data URL.

Usage:
    python embed_image.py path/to/image.png output.html
"""

import argparse
import base64

jaidevd

Hypothesis:

For a given package ABC, if conda installs it without any additional arguments, pip install the same package without any additional arguments, they are not coming from the same source. They will have a slightly different version, patch or minor.

Procedure:

jaidevd

let g:python3_host_prog="~/conda/bin/python"
let g:python2_host_prog="~/conda/bin/python"


call plug#begin('~/.local/share/nvim/plugged')

" Autocomplete stuff

if has('nvim')
  Plug 'Shougo/deoplete.nvim', { 'do': ':UpdateRemotePlugins' }

jaidevd

X <- numeric(length=1000)
params.a <- sqrt(30) / 4
params.c <- 0.811

sqrt.fx <- function(x) {
  f <- (x ^ 4) - 2 * (x ^ 3) + (x ^ 2)
  return(sqrt(30 * f))
}

sample <- function() {

jaidevd

Suppose you have $n$ random variables, $\mathbf{x}_1, \mathbf{x}_2, x_3, \ldots x_n $.

Now, between any two random variables $x_i, x_j$ the covariance is defined as:

$$ \sigma_{\mathbf{x}_i, \mathbf{x}_j} = E[(\mathbf{x}_i - \mu_{\mathbf{x}_i})(\mathbf{x}_j - \mu_{\mathbf{x}_j})] $$

(where $\mu_{\mathbf{x}_i}$ is the mean of $\mathbf{x}_i$).

jaidevd

from flask import request, Flask

app = Flask(__name__)


@app.route("/", methods=["POST"])
def index():
    if request.content_type == 'application/pdf':
        with open('request.pdf', 'wb') as fout:
            fout.write(request.data)

jaidevd

# coding: utf-8

from keras import layers as L
from keras import backend as K
from keras.models import Model
import tensorflow as tf
import numpy as np

_epsilon = tf.convert_to_tensor(K.epsilon(), np.float32)

jaidevd

Generative vs discriminative classifiers

Discriminative classifiers:

Discriminative classifiers divide the feature space into regions that separate the data belonging to different classes such that every separate region contains samples belonging to a single class. The decision boundary is determined by constructing and solving equations of the form

$$ \mathbf{y} = \mathbf{w}^T\mathbf{\phi(X)}+b $$

jaidevd

[
  {
    "text": "#BREAKINGNEWS MALAYSIA AIRLINES FLIGHT #MH17 CONFIRMED SHOT DOWN OVER #DONETSK OBLAST, SHORTLY BEFORE REACHING RUSSIAN AIR SPACE",
    "labels": [
      {
        "start": 14,
        "end": 31,
        "label": "ORG"
      },
      {