ekinakyurek

DAY 1

1. The State of Deep Learning : Overview Talk (I) Amnon Shashua

• Deep Learning w. overparametrized networks enabled: Training Error ↓ Test Error ↓

• Quantum Entanglement related with Deep Nets https://journals.aps.org/prl/abstract/10.1103/PhysRevLett.122.065301

• Self Driving Cars, Diriving Policy, Ethics in AI

  • Failure allowed only due to perception error and it should be less then human's error

ekinakyurek

[ec2-user@ip-172-31-14-239 ~]$ sh benchmark.sh
++ mktemp -d
+ cd /tmp/tmp.WbX6402Bko
+ git clone https://github.com/ekinakyurek/Mac-Network .
Cloning into '.'...
remote: Enumerating objects: 68, done.
remote: Counting objects: 100% (68/68), done.
remote: Compressing objects: 100% (52/52), done.
remote: Total 328 (delta 30), reused 47 (delta 15), pack-reused 260
Receiving objects: 100% (328/328), 29.89 MiB | 50.42 MiB/s, done.

ekinakyurek

#!/usr/bin/env bash
set -uxe

cd $(mktemp -d)

git clone https://github.com/ekinakyurek/Mac-Network .
sed -i '1d' benchmark.jl  # remove pkg import/activate

export JULIA_DEPOT_PATH=$(mktemp -d)
julia --project                 -e 'using Pkg; Pkg.instantiate(); pkg"free Knet"; pkg"add Knet#cuarrays CuArrays#master CUDAnative#master CUDAdrv#master"'

ekinakyurek

import Base: get, length, getindex, push!, append!

struct IndexedDict{T}
    toIndex::Dict{T,Int};
    toElement::Vector{T};
  
    IndexedDict{T}(toIndex,toElement) where T = new(toIndex,toElement)
    IndexedDict{T}(toIndex,toElement) where T<:Integer = error("Cannot Create IndexedDict of Integers")
    IndexedDict{T}(toIndex,toElement) where T<:AbstractArray = error("Cannot Create IndexedDict of Arrays")
end

ekinakyurek

#model url: https://github.com/FluxML/Metalhead.jl/releases/download/v0.1.1/resnet.bson
#install 
using Knet, KnetLayers, BSON, Images

struct ResidualBlock
  layers
  shortcut
end

function ResidualBlock(filters, kernels::Array{Tuple{Int,Int}}, pads::Array{Tuple{Int,Int}}, strides::Array{Tuple{Int,Int}}, shortcut = identity)

ekinakyurek