| def adadelta(opfunc, x, config, state=None): | |
| """ADADELTA implementation http://arxiv.org/abs/1212.5701 | |
| ARGUMENTS: | |
| - `opfunc` : a function that takes a single input (X), the point of | |
| evaluation, and returns f(X) and df/dX | |
| - `x` : the initial point | |
| - `config` : a table of hyper-parameters | |
| - `config['rho']` : interpolation parameter | |
| - `config['eps']` : for numerical stability | |
| - `config['weightDecay']` : weight decay | |
| - `state` : a table describing the state of the optimizer; after each | |
| call the state is modified | |
| - `state['paramVariance']` : vector of temporal variances of parameters | |
| - `state['accDelta']` : vector of accummulated delta of gradients | |
| RETURNS: | |
| - `x` : the new x vector | |
| - `f(x)` : the value of optimized function, evaluated before the update | |
| """ | |
| # (0) get/update state | |
| if config is None and state is None: | |
| raise ValueError("adadelta requires a dictionary to retain state between iterations") | |
| state = state if state is not None else config | |
| rho = config.get('rho', 0.9) | |
| eps = config.get('eps', 1e-6) | |
| wd = config.get('weightDecay', 0) | |
| state['evalCounter'] = state.get('evalCounter', 0) | |
| # (1) evaluate f(x) and df/dx | |
| fx, dfdx = opfunc(x) | |
| # (2) weight decay | |
| if wd != 0: | |
| dfdx.add_(wd, x) | |
| # (3) parameter update | |
| if 'paramVariance' not in state: | |
| state['paramVariance'] = x.new().resize_as_(dfdx).zero_() | |
| state['paramStd'] = x.new().resize_as_(dfdx).zero_() | |
| state['delta'] = x.new().resize_as_(dfdx).zero_() | |
| state['accDelta'] = x.new().resize_as_(dfdx).zero_() | |
| state['paramVariance'].mul_(rho).addcmul_(1 - rho, dfdx, dfdx) | |
| state['paramStd'].resize_as_(state['paramVariance']).copy_(state['paramVariance']).add_(eps).sqrt_() | |
| state['delta'].resize_as_(state['paramVariance']).copy_( | |
| state['accDelta']).add_(eps).sqrt_().div_(state['paramStd']).mul_(dfdx) | |
| x.add_(-1, state['delta']) | |
| state['accDelta'].mul_(rho).addcmul_(1 - rho, state['delta'], state['delta']) | |
| # (4) update evaluation counter | |
| state['evalCounter'] += 1 | |
| # return x*, f(x) before optimization | |
| return x, fx |