Posts

Perceptron in Python and C++

Image
Intro In this post, I’m going to cover the Perceptron Algorithm and compare its implementation in Python and C++. Perceptron Explanation The perceptron algorithm performs binary classifications by using a signed weighted sum of an input point to predict one of two output classes. The weights used are found by fitting a linear decision boundary to the training data. In order for the perceptron to achieve 100% accuracy, the data must be linearly separable. Formally, the perceptron prediction function f can be represented as: Predictions are made using the sign function as the activation function: Training the perceptron is an iterative process. At a high level, the algorithm: Initialize weights to be zeroes For a predefined # of epochs makes a prediction for each misclassified point updates the weight vector The update rule for the weight matrix is given by the following: Error is calculated with the equation below where y is the label for each input x
Post a Comment
Read more

board2board - collaborative whiteboard using object detection

Image
board2board Our app board2board allows multiple users to collaboratively edit a board using a colored object. Previous projects in this area have forced users to draw on their trackpad or switch to a tablet to draw. Our app differs from previous work by using the webcam on a computer to detect the motion of a colored object. We allow users to communicate with one another remotely while still retaining the dexterity of drawing by hand. Features Since users’ background environments may vary greatly, we have given the option of choosing different colors to detect. Our currently supported colors are blue and green as we have found these to deliver the best results but we plan to extend the color spectrum to any color. We allow users to draw disconnected lines by having the spacebar toggle whether the marker is drawing or not. We also allow users to clear the board using the key “c”. Users can collaboratively edit the board by visiting the same link. In the fu
Post a Comment
Read more

Q Learning Explained with Example Code

Image
Intro to Reinforcement Learning and Q Learning Q Learning is a off-policy TD reinforcement learning algorithm. Reinforcement learning consists of an agent, an environment, and a reward system. The agent performs an action in its current environment and receives a reward for it. The goal of the agent is to maximize the expected cumulative reward. We can see this visualized in an agent-environment loop depicted below. Source: https://devblogs.nvidia.com/train-reinforcement-learning-agents-openai-gym/ Q learning is a type of TD, or Temporal Difference learning, algorithm. This means that our algorithm learns at every time step (loop in the above diagram) by remembering the best possible actions for each state and action. In fact, the Q in Q learning is a function that takes in a state “s” and an action “a” and returns the expected reward for the given inputs. Q learning seeks to approximate the optimal action-value function Q*. Another important detail to note about Q
Post a Comment
Read more

Intro and Raspberry Pi Setup

Image
Greetings! I will be using this blog to keep track of all of the interesting things that I am working on. I recently acquired a Raspberry Pi. I will show an overview of how to set it up, as well as a few details that I found useful that other instructional resources online did not include. I will also show how to setup the 7" touchscreen, install a touchscreen keyboard, and how to take screenshots. The model I have is a Raspberry Pi 3 Model B V1.2. Raspberry Pi Setup: Step 1: Install the appropriate software on the SD card Connect your mini SD card to your computer. You may need an adapter. Visit this  website and select the NOOBS (New Out Of the Box Software). Then, download the ZIP file of NOOBS, not NOOBS Lite. Unzip the file and move its contents onto the SD card. Eject the SD card from your computer and insert it into the Raspberry Pi. Step 2: First time setup Here is the function of all the different parts of the Raspberry Pi: Source Plug the Pi into
Post a Comment
Read more