Skip to content

Instantly share code, notes, and snippets.

View SamKirkiles's full-sized avatar

Sam Kirkiles SamKirkiles

22 followers · 2 following
View GitHub Profile
@SamKirkiles
SamKirkiles / Validation Data
Created June 18, 2024 17:42
Frog: [[[0.23137254901960785, 0.24313725490196078, 0.24705882352941178], [0.16862745098039217, 0.18039215686274507, 0.1764705882352941], [0.19607843137254902, 0.18823529411764706, 0.16862745098039217], [0.26666666666666666, 0.21176470588235294, 0.1647058823529412], [0.3843137254901961, 0.28627450980392155, 0.20392156862745098], [0.4666666666666667, 0.3568627450980392, 0.24705882352941178], [0.5450980392156862, 0.4196078431372549, 0.29411764705882354], [0.5686274509803921, 0.43137254901960786, 0.3137254901960784], [0.5843137254901961, 0.4588235294117647, 0.34901960784313724], [0.5843137254901961, 0.47058823529411764, 0.36470588235294116], [0.5137254901960784, 0.403921568627451, 0.30196078431372547], [0.49019607843137253, 0.38823529411764707, 0.2980392156862745], [0.5568627450980392, 0.45098039215686275, 0.3568627450980392], [0.5647058823529412, 0.4392156862745098, 0.33725490196078434], [0.5372549019607843, 0.4117647058823529, 0.30980392156862746], [0.5058823529411764, 0.3803921568627451, 0.2784313725490196], [
@SamKirkiles
SamKirkiles / weights
Created June 18, 2024 02:23
B1:[-1.526787794440281e-12, 2.5593275423978253e-11, -7.933900005110362e-13, 6.912213183905846e-12, 3.4158142708208483e-12, -5.302215417986998e-12, -3.940839522430512e-11, 5.601949279097138e-12, 6.059488256140292e-12, 1.2729296451654576e-11, -2.242993402090489e-12, -2.152980933435733e-12, -1.0890569672870463e-11, 6.5211631388025714e-12, -1.1462381050466371e-11, 1.9558842930352128e-12]
B2:[-2.1183288382665742e-13, -3.414346540424672e-14, -9.615830865479912e-14, 7.549096578709037e-13, -3.626362079200702e-14, 2.188527821170726e-18, -4.540578726128621e-13, 3.348260170611118e-13, 8.282698764445765e-14, 6.119610064433554e-15, -1.1303759645566403e-13, 1.8436110590788725e-13, 2.469308283155262e-13, -1.5789101207134008e-13, 4.5053094646779156e-13, 2.8786924548568895e-13]
B3:[1.1275819593958992e-13, -9.81989650217869e-14, -8.436599400880492e-14, 1.3567463925421803e-13, -1.5857129364837595e-13, 3.4242661343003727e-13, 4.8308830422491504e-14, -5.695482203573613e-14]
B4:[-0.020290048805014827, -0.40347595093696037, -0.3989
@SamKirkiles
SamKirkiles / forward_pass.py
Last active June 18, 2024 17:15
def shape(x):
return (len(x), len(x[0]), len(x[0][0]), len(x[0][0][0]))
def conv_forward(x,weight,b,parameters):
pad = parameters['pad']
stride = parameters['stride']
(m, n_h, n_w, n_C_prev) = shape(x)
@SamKirkiles
SamKirkiles / gist:5f9ccb501074558aba508e30cc814183
Created May 28, 2024 22:01
CNN
def conv_forward_naive(x,weight,b,parameters):
pad = parameters['pad']
stride = parameters['stride']
(m, n_h, n_w, n_C_prev) = (len(x), len(x[0]), len(x[0][0]), len(x[0][0][0]))
(f,f, n_C_prev, n_C) = (len(weight), len(weight[0]), len(weight[0][0]), len(weight[0][0][0]))
n_H = int(1 + (n_h + 2 * pad - f) / stride)
@SamKirkiles
SamKirkiles / endian_viz.c
Created November 9, 2018 02:13
A C program to visualize little and big endian representations of an integer.
#include <stdio.h>
#include <stdlib.h>
void convertToBinary(unsigned int n, char ** representation, int littleEndian){
int spot = 0;
int c;
if (littleEndian)
c = 0;
@SamKirkiles
SamKirkiles / GradeSchool.java
Created September 8, 2018 19:30
CS 250 - Recursive Addition
import java.util.*;
public class GradeSchool {
public static void main(String[] args){
// Test add
int[] a = new int[] {9,4,3,2};
int[] b = new int[] {3,1,8,5};
@SamKirkiles
SamKirkiles / REINFORCE-PG-SIMPLE.py
Last active February 11, 2023 10:19
REINFORCE Policy Gradients from scratch in Numpy. Code from: https://medium.com/samkirkiles/reinforce-policy-gradients-from-scratch-in-numpy-6a09ae0dfe12
import gym
import numpy as np
import matplotlib.pyplot as plt
import copy
#Hyperparameters
NUM_EPISODES = 10000
LEARNING_RATE = 0.000025
GAMMA = 0.99
@SamKirkiles
SamKirkiles / REINFORCE-PG-KERNEL.py
Last active October 30, 2019 21:30
import gym
import numpy as np
import sklearn.pipeline
from sklearn.kernel_approximation import RBFSampler
import matplotlib.pyplot as plt
import copy
#Hyperparameters
NUM_EPISODES = 10000
LEARNING_RATE = 0.000025