#SUSTAIN-MATLAB
This set of scripts runs a minimal version of the SUSATIN model of category learning (Love, Medin, & Gureckis, 2004) for application in traditional artificial classification learning (TACL) experiments. It is written in MATLAB, and is currently set up to simulate the six-types problem of Shepard, Hovland, & Jenkins (1961), though it should generalize to any dataset.
#####Keeping things as simple as possible:
This version of SUSTAIN is set up exclusively for class prediction within continuous (or at least binary) spaces. Feature prediction capabilities are not included, and neither is support for nominal dimensions. Todd Gureckis has a [nice implementation](https://github.com/NYUCCL/sustain_python) in python if you are looking for these capabilities.
There are a variety of utility scripts, and a few important ones:
- **start.m** can be used to set up and run SUSTAIN.
- **SUSTAIN.m** uses a provided architecture to train a network on a set of inputs and category assignments.
- **FORWARDPASS.m** and **UPDATE.m** are used to propagate activations forward through the model, and update the network's knowledge, respectively.
Simulations are run by executing the *start.m* script. All simulations begin by passing a model struct to the *SUSTAIN.m* function. At a minimum, 'model' needs to include:
| Field | Description | Type |
| ---------------| ------------------------------------------| :-----------------------------: |
| `exemplars` | Matrix of training stimuli coordinates | Item-by-feature matrix |
| `numblocks` | Number of iterations over the exemplars | Integer (>0) |
| `numinitials` | Number of random initializations | Integer (>0) |
| `assignments` | class assignment for each exemplar | Integer vector (>0) |
| `params` | [attn, comp, decision, lrate] | Float vector (0 - Inf) |
For almost all situations, inputs should be scaled to [-1 +1]. SUSTAIN.m will train the network and return a result struct. As-is, 'result' contains only training accuracy (averaged across inits) at each training block. Additional measures, such as test phase classification, can be added.
Written by [Nolan Conaway](http://bingweb.binghamton.edu/~nconawa1/).
Updated May 2, 2016
没有合适的资源?快使用搜索试试~ 我知道了~
温馨提示
matlab算法,工具源码,适合毕业设计、课程设计作业,所有源码均经过严格测试,可以直接运行,可以放心下载使用。有任何使用问题欢迎随时与博主沟通,第一时间进行解答! matlab算法,工具源码,适合毕业设计、课程设计作业,所有源码均经过严格测试,可以直接运行,可以放心下载使用。有任何使用问题欢迎随时与博主沟通,第一时间进行解答! matlab算法,工具源码,适合毕业设计、课程设计作业,所有源码均经过严格测试,可以直接运行,可以放心下载使用。有任何使用问题欢迎随时与博主沟通,第一时间进行解答! matlab算法,工具源码,适合毕业设计、课程设计作业,所有源码均经过严格测试,可以直接运行,可以放心下载使用。有任何使用问题欢迎随时与博主沟通,第一时间进行解答! matlab算法,工具源码,适合毕业设计、课程设计作业,所有源码均经过严格测试,可以直接运行,可以放心下载使用。有任何使用问题欢迎随时与博主沟通,第一时间进行解答! matlab算法,工具源码,适合毕业设计、课程设计作业,所有源码均经过严格测试,可以直接运行,可以放心下载使用。有任何使用问题欢迎随
资源推荐
资源详情
资源评论
收起资源包目录
毕业设计&课设-用于SUSTAIN仿真的matlab脚本.zip (11个子文件)
matlab_code
start.m 533B
utils
shuffletrials.m 862B
blockrows.m 1003B
v2struct.m 16KB
FORWARDPASS.m 2KB
RESPONSERULE.m 688B
UPDATE.m 2KB
SUSTAIN.m 4KB
.gitignore 400B
shj.mat 265B
README.md 2KB
共 11 条
- 1
资源评论
白话机器学习
- 粉丝: 8978
- 资源: 7681
上传资源 快速赚钱
- 我的内容管理 展开
- 我的资源 快来上传第一个资源
- 我的收益 登录查看自己的收益
- 我的积分 登录查看自己的积分
- 我的C币 登录后查看C币余额
- 我的收藏
- 我的下载
- 下载帮助
安全验证
文档复制为VIP权益,开通VIP直接复制
信息提交成功